From 5e10909fe9da3c9e2f5f329b39251eff877e69db Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 04:01:36 +0000 Subject: [PATCH 01/12] Bytes after the value: a payload is one MessagePack value (C#) (#125) --- .../cultmesh/content-acceptance-vectors.json | 18 +- packages/cultnet-ts/CHANGELOG.md | 4 + packages/cultnet-ts/README.md | 4 +- packages/cultnet-ts/src/content.ts | 4 +- .../CultPayloadDepth.cs | 13 +- .../Contracts/cultcache-persistence-format.md | 2 + .../CultMeshProviderSessionWire.cs | 5 +- .../CultNetDocumentRegistry.cs | 12 +- .../BytesAfterVectorTests.cs | 211 ++++++++++++++++++ .../PayloadDepthTests.cs | 21 +- .../BytesAfterPathTests.cs | 61 +++++ .../CultMeshContentVectorTests.cs | 6 +- .../CultMeshProviderSessionWireTests.cs | 27 +++ .../BytesAfterPathTests.cs | 62 +++++ tests/vectors/union-parity/README.md | 19 ++ .../bytes-after-deep-tail.msgpack | Bin 0 -> 601 bytes .../bytes-after-never-used.msgpack | Bin 0 -> 279 bytes .../union-parity/bytes-after-nil.msgpack | Bin 0 -> 278 bytes .../union-parity/bytes-after-outcomes.msgpack | 1 + .../bytes-after-second-document.msgpack | Bin 0 -> 554 bytes unity/org.gamecult.cultlib/CHANGELOG.md | 1 + 21 files changed, 441 insertions(+), 30 deletions(-) create mode 100644 tests/GameCult.Caching.Tests/BytesAfterVectorTests.cs create mode 100644 tests/GameCult.Mesh.Tests/BytesAfterPathTests.cs create mode 100644 tests/GameCult.Networking.Tests/BytesAfterPathTests.cs create mode 100644 tests/vectors/union-parity/bytes-after-deep-tail.msgpack create mode 100644 tests/vectors/union-parity/bytes-after-never-used.msgpack create mode 100644 tests/vectors/union-parity/bytes-after-nil.msgpack create mode 100644 tests/vectors/union-parity/bytes-after-outcomes.msgpack create mode 100644 tests/vectors/union-parity/bytes-after-second-document.msgpack diff --git a/contracts/cultmesh/content-acceptance-vectors.json b/contracts/cultmesh/content-acceptance-vectors.json index d9868a63..b4d92cde 100644 --- a/contracts/cultmesh/content-acceptance-vectors.json +++ b/contracts/cultmesh/content-acceptance-vectors.json @@ -899,23 +899,23 @@ { "label": "trailing_nil", "inputBase64": "mqNhL2KlYXNzZXShMdlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQCvYXBwbGljYXRpb24veC10oXSRlNlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQAAoWuRoXiBoWuhdsA=", - "accepted": true, - "reencodedBase64": "mqNhL2KlYXNzZXShMdlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQCvYXBwbGljYXRpb24veC10oXSRldlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQAAoWugkaF4gaFroXY=", - "recordKey": "mesh:cdn:artifact:a-b:1:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855" + "accepted": false, + "reencodedBase64": null, + "recordKey": null }, { "label": "trailing_reserved_code", "inputBase64": "mqNhL2KlYXNzZXShMdlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQCvYXBwbGljYXRpb24veC10oXSRlNlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQAAoWuRoXiBoWuhdsE=", - "accepted": true, - "reencodedBase64": "mqNhL2KlYXNzZXShMdlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQCvYXBwbGljYXRpb24veC10oXSRldlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQAAoWugkaF4gaFroXY=", - "recordKey": "mesh:cdn:artifact:a-b:1:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855" + "accepted": false, + "reencodedBase64": null, + "recordKey": null }, { "label": "trailing_truncated_array", "inputBase64": "mqNhL2KlYXNzZXShMdlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQCvYXBwbGljYXRpb24veC10oXSRlNlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQAAoWuRoXiBoWuhdpI=", - "accepted": true, - "reencodedBase64": "mqNhL2KlYXNzZXShMdlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQCvYXBwbGljYXRpb24veC10oXSRldlAZTNiMGM0NDI5OGZjMWMxNDlhZmJmNGM4OTk2ZmI5MjQyN2FlNDFlNDY0OWI5MzRjYTQ5NTk5MWI3ODUyYjg1NQAAoWugkaF4gaFroXY=", - "recordKey": "mesh:cdn:artifact:a-b:1:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855" + "accepted": false, + "reencodedBase64": null, + "recordKey": null }, { "label": "top_nil", diff --git a/packages/cultnet-ts/CHANGELOG.md b/packages/cultnet-ts/CHANGELOG.md index 56a15c7b..cd5a32f2 100644 --- a/packages/cultnet-ts/CHANGELOG.md +++ b/packages/cultnet-ts/CHANGELOG.md @@ -5,6 +5,10 @@ All notable changes to `@gamecult/cultnet-ts` are recorded here, following ## Unreleased +### Breaking + +- `decodeManifest` refuses a manifest with bytes after its first MessagePack value (`CDN artifact manifest has N bytes after its first MessagePack value.`), as the C# reference does; before, the tail was ignored. `contracts/cultmesh/content-acceptance-vectors.json` now holds `trailing_nil`, `trailing_reserved_code` and `trailing_truncated_array` as refused. + ### Added - Fork card (`contracts/cultnet/governance/fork-card.md`): `raised_at`, a required key holding an RFC3339 instant string or nil (the question's first admission time), outside `options_hash`; the card vectors are re-pinned with the key, three negatives (epoch integer, date-only string, missing key) and a nil-`raised_at` valid card; every `optionsHash` digest is unchanged. diff --git a/packages/cultnet-ts/README.md b/packages/cultnet-ts/README.md index 7d1b9a6a..92231d52 100644 --- a/packages/cultnet-ts/README.md +++ b/packages/cultnet-ts/README.md @@ -144,8 +144,8 @@ imports no `node:` module, so a browser bundle can take it. past the shared depth limit are refused. - The decoder reads what MessagePack-CSharp reads and refuses what it refuses: nil in a string, tag or metadata slot is `null` and is written back as nil, - a missing trailing slot takes the reference's default, extra slots and trailing - bytes are skipped (a chunk reference's fifth slot is its element id, a string + a missing trailing slot takes the reference's default, extra slots are skipped and bytes + after the manifest's first value are refused (a chunk reference's fifth slot is its element id, a string or nil, `""` when absent; `packContent` writes the decimal offset as the id, as the reference does), a float-encoded integer is refused, and invalid UTF-8 decodes to U+FFFD. A nil `chunks` is refused, as the reference's diff --git a/packages/cultnet-ts/src/content.ts b/packages/cultnet-ts/src/content.ts index 6f9fd873..683de8a0 100644 --- a/packages/cultnet-ts/src/content.ts +++ b/packages/cultnet-ts/src/content.ts @@ -249,7 +249,7 @@ class Reader { export function decodeManifest(bytes: Uint8Array): CultMeshCdnArtifactManifest { checkPayloadDepth(bytes, "CDN artifact manifest"); const reader = new Reader(bytes); - // A slot the array does not reach keeps the property initialiser; a slot past the tenth is skipped. Trailing bytes are not read. + // A slot the array does not reach keeps the property initialiser; a slot past the tenth is skipped. const count = reader.array("manifest"); let index = 0; const slot = (read: () => T, initial: T): T => index++ < count ? read() : initial; @@ -267,6 +267,8 @@ export function decodeManifest(bytes: Uint8Array): CultMeshCdnArtifactManifest { [reader.string("metadata") ?? wrongType("metadata"), reader.string("metadata")] as const), []), }; for (; index < count; index += 1) reader.skip(); + // A manifest is one record payload: exactly one MessagePack value, as the reference's one-value rule has it. + if (reader.position < bytes.length) throw new Error(`CDN artifact manifest has ${bytes.length - reader.position} bytes after its first MessagePack value.`); validateManifest(manifest); return manifest; } diff --git a/src/GameCult.Caching.MessagePack/CultPayloadDepth.cs b/src/GameCult.Caching.MessagePack/CultPayloadDepth.cs index 258663e8..6179f8be 100644 --- a/src/GameCult.Caching.MessagePack/CultPayloadDepth.cs +++ b/src/GameCult.Caching.MessagePack/CultPayloadDepth.cs @@ -7,11 +7,12 @@ namespace GameCult.Caching.MessagePack; internal enum PayloadDirection { Read, Write } /// -/// The one payload nesting count. Walks the MessagePack headers of a document payload iteratively, counting every array and +/// The one walk of a payload's shape: its depth and its one value. Walks the MessagePack headers of a document payload iteratively, counting every array and /// map (read by a formatter or skipped past it; the outermost array is depth 1) and refusing the first container past /// . Scalars are skipped by length, nothing is allocated per /// element, and the count stack is bounded by the limit. A refusal names the type, the depth reached and the limit, -/// never payload bytes. A payload that ends early is the deserializer's to refuse. +/// never payload bytes. A payload that ends early is the deserializer's to refuse; bytes after the first value are refused +/// with their count and never walked. /// internal static class CultPayloadDepth { @@ -73,7 +74,13 @@ internal static class CultPayloadDepth // One element is complete; close every container it was the last element of. while (depth > 0 && --remaining[depth - 1] == 0) depth--; - if (depth == 0) return; + if (depth == 0) + { + if (position < bytes.Length) + throw new MessagePackSerializationException( + $"{type.FullName} payload has {bytes.Length - position} bytes after its first MessagePack value (refused on {(direction == PayloadDirection.Read ? "read" : "write")})."); + return; + } } } diff --git a/src/GameCult.Caching/Contracts/cultcache-persistence-format.md b/src/GameCult.Caching/Contracts/cultcache-persistence-format.md index 5ac6c3b5..ad3c06f5 100644 --- a/src/GameCult.Caching/Contracts/cultcache-persistence-format.md +++ b/src/GameCult.Caching/Contracts/cultcache-persistence-format.md @@ -403,6 +403,8 @@ runtime enforces the same bound on read and on write, and the refusal names the the depth and the limit, never payload bytes. The vectors are in `tests/vectors/union-parity` (`bounds-outcomes.msgpack`). +A record's payload is exactly one MessagePack value: every runtime refuses a payload with bytes after its first value wherever it reads a payload under the nesting bound, and on write where a write takes bytes the library did not encode itself (a C# type's own formatter, a Rust raw envelope), without walking those bytes, so a tail past the limit is refused as bytes after the value and not as depth; the refusal names the type and the count of bytes after the value, never the bytes (vectors: `tests/vectors/union-parity/bytes-after-outcomes.msgpack`). A CDN manifest is stored and keyed as a record and follows the same rule (`contracts/cultmesh/content-acceptance-vectors.json`: `trailing_nil`, `trailing_reserved_code` and `trailing_truncated_array` are refused). Streams of values (CultNet frames, a store's record array) are different contracts and are not covered. + ## Variant Records A variant record is a record whose document is a base record's document with some diff --git a/src/GameCult.Mesh/CultMeshProviderSessionWire.cs b/src/GameCult.Mesh/CultMeshProviderSessionWire.cs index 03f3a70c..ad9ae66e 100644 --- a/src/GameCult.Mesh/CultMeshProviderSessionWire.cs +++ b/src/GameCult.Mesh/CultMeshProviderSessionWire.cs @@ -284,7 +284,10 @@ namespace GameCult.Mesh throw new ArgumentException($"Expected provider command schema '{CultMeshProviderSessionWireContract.CommandSchema}'.", nameof(document)); if (!string.Equals(document.PayloadEncoding, "messagepack", StringComparison.Ordinal)) throw new ArgumentException("Provider command document must use MessagePack.", nameof(document)); - var command = MessagePackSerializer.Deserialize(document.Payload, Options); + // A raw record's payload is one MessagePack value (ruling trailing-bytes-refuse): a tail is refused by count, never walked. + var command = MessagePackSerializer.Deserialize(new ReadOnlyMemory(document.Payload), Options, out var read); + if (read < document.Payload.Length) + throw new MessagePackSerializationException($"Provider command document: {document.Payload.Length - read} bytes after its first MessagePack value (refused on read)."); if (command == null) throw new MessagePackSerializationException("Provider command decoded to null."); Validate(command); return command; diff --git a/src/GameCult.Networking/CultNetDocumentRegistry.cs b/src/GameCult.Networking/CultNetDocumentRegistry.cs index 19f17c8c..bf17a91f 100644 --- a/src/GameCult.Networking/CultNetDocumentRegistry.cs +++ b/src/GameCult.Networking/CultNetDocumentRegistry.cs @@ -641,23 +641,19 @@ namespace GameCult.Networking string storedAt, CultNetDocumentMessageOptions? options = null) { - var binding = GetByDocumentType(descriptor.DocumentType) ?? - new CultNetDocumentBinding( - descriptor.DocumentType, - descriptor.SchemaId, - value => CultDocumentMessagePackSerialization.SerializeUntyped(value, value.GetType(), _documents), - payload => CultDocumentMessagePackSerialization.DeserializeUntyped(descriptor.DocumentType, payload, _documents)); + var binding = GetByDocumentType(descriptor.DocumentType); return new CultNetRawDocumentRecord { - SchemaId = binding.SchemaId, + SchemaId = binding?.SchemaId ?? descriptor.SchemaId, SchemaName = descriptor.SchemaName, SchemaVersion = descriptor.SchemaVersion, SchemaContentHash = descriptor.ContentHash, RecordKey = key.Value, StoredAt = storedAt, PayloadEncoding = "messagepack", - Payload = binding.PayloadSerializer(document), + Payload = binding?.PayloadSerializer(document) ?? + CultDocumentMessagePackSerialization.SerializeUntyped(document, document.GetType(), _documents), SourceRuntimeId = options?.SourceRuntimeId, SourceAgentId = options?.SourceAgentId, SourceRole = options?.SourceRole, diff --git a/tests/GameCult.Caching.Tests/BytesAfterVectorTests.cs b/tests/GameCult.Caching.Tests/BytesAfterVectorTests.cs new file mode 100644 index 00000000..86d677e6 --- /dev/null +++ b/tests/GameCult.Caching.Tests/BytesAfterVectorTests.cs @@ -0,0 +1,211 @@ +#nullable enable +using System; +using System.IO; +using System.Linq; +using GameCult.Caching.MessagePack; +using MessagePack; +using NUnit.Framework; +using UnionSurface = GameCult.Caching.Tests.UnionParityVectorTests.UnionSurface; + +namespace GameCult.Caching.Tests +{ + // A record's payload is exactly one MessagePack value. The vectors are union-surface.msgpack followed by a tail; the C# + // reference is their only writer (VectorFiles, CULTCACHE_WRITE_VECTORS=1) and every other runtime is held to bytes-after-outcomes.msgpack. + public class BytesAfterVectorTests + { + private const string Vector = "union-parity"; + private const string Canary = "CANARY-bytes-after-7d1e"; + private const string Clause = "bytes after its first MessagePack value"; + private const int Limit = CultMessagePackSecurity.MaximumPayloadDepth; + + private static readonly CultDocumentRegistry Registry = CultDocumentRegistry.ForTypes(new[] { typeof(UnionSurface) }); + + private static byte[] Base => VectorFiles.Committed(Vector, "union-surface.msgpack"); + + private static byte[] DeepTail() + { + var text = System.Text.Encoding.ASCII.GetBytes(Canary); + return Enumerable.Repeat((byte)0x91, 300).Concat(new[] { (byte)(0xa0 | text.Length) }).Concat(text).ToArray(); + } + + // file, tail + private static (string File, byte[] Tail)[] Cases() => new (string, byte[])[] + { + ("bytes-after-nil.msgpack", new byte[] { 0xc0 }), + ("bytes-after-never-used.msgpack", new byte[] { 0xc1, 0xc1 }), + ("bytes-after-deep-tail.msgpack", DeepTail()), + ("bytes-after-second-document.msgpack", Base), + }; + + // The bounds row shape: [file, outcome, union, key|nil, detail]. + private static byte[] OutcomesFile() + { + var buffer = new System.Buffers.ArrayBufferWriter(); + var w = new MessagePackWriter(buffer); + var cases = Cases(); + w.WriteArrayHeader(cases.Length); + foreach (var (file, _) in cases) + { + w.WriteArrayHeader(5); + w.Write(file); + w.Write("refused"); + w.Write("UnionSurface"); + w.WriteNil(); + w.Write("bytes-after"); + } + + w.Flush(); + return buffer.WrittenSpan.ToArray(); + } + + private static string Message(Action action) + { + try { action(); } + catch (Exception exception) + { + var message = ""; + for (var e = exception; e != null; e = e.InnerException) message += e.Message + "\n"; + return message; + } + + throw new AssertionException("the payload was accepted"); + } + + [Test] + public void VectorsMatchAndRefuse() + { + var baseBytes = Base; + foreach (var (file, tail) in Cases()) VectorFiles.Check(Vector, file, baseBytes.Concat(tail).ToArray()); + VectorFiles.Check(Vector, "bytes-after-outcomes.msgpack", OutcomesFile()); + + // From here only the committed files speak. + var reader = new MessagePackReader(VectorFiles.Committed(Vector, "bytes-after-outcomes.msgpack")); + Assert.That(reader.ReadArrayHeader(), Is.EqualTo(4)); + for (var i = 0; i < 4; i++) + { + Assert.That(reader.ReadArrayHeader(), Is.EqualTo(5)); + var file = reader.ReadString()!; + Assert.That(reader.ReadString(), Is.EqualTo("refused")); + Assert.That(reader.ReadString(), Is.EqualTo("UnionSurface")); + Assert.That(reader.TryReadNil(), Is.True); + Assert.That(reader.ReadString(), Is.EqualTo("bytes-after")); + + var payload = VectorFiles.Committed(Vector, file); + Assert.That(payload.Take(baseBytes.Length), Is.EqualTo(baseBytes), $"{file} begins with union-surface.msgpack"); + var count = payload.Length - baseBytes.Length; + var message = Message(() => CultDocumentMessagePackSerialization.DeserializeUntyped(typeof(UnionSurface), payload, Registry)); + Assert.That(message, Does.Contain(typeof(UnionSurface).FullName).And.Contain($"{count} {Clause}").And.Contain("refused on read"), file); + Assert.That(message, Does.Not.Contain(Canary).And.Not.Contain("the limit is"), file); + } + } + + // The codec the store and the inspector read through is a path of its own: DecodeDocument deserializes by the codec's + // delegate, so a tail it did not refuse would be read as a document. + [Test] + public void CodecDecodeDocumentRefusesATail() + { + var codec = CultCacheMessagePack.CreateCodec(Registry); + Assert.That(codec.DecodeDocument(new CultRecordKey("surface"), Base), Is.Not.Null); + var message = Message(() => codec.DecodeDocument(new CultRecordKey("surface"), Base.Concat(new byte[] { 0xc0 }).ToArray())); + Assert.That(message, Does.Contain(typeof(UnionSurface).FullName).And.Contain($"1 {Clause}").And.Contain("refused on read")); + } + + [Test] + public void StoreLoadRefuses() + { + var work = Path.Combine(Path.GetTempPath(), $"cultlib-bytesafter-{Guid.NewGuid():N}"); + Directory.CreateDirectory(work); + try + { + var snapshot = CultDocumentMessagePackSerialization.DeserializeSnapshot(VectorFiles.Committed(Vector, "union-surface.cc")); + snapshot.Records[0].Payload = Base.Concat(new byte[] { 0xc0 }).ToArray(); + var scratch = Path.Combine(work, "tail.cc"); + File.WriteAllBytes(scratch, CultDocumentMessagePackSerialization.SerializeSnapshot(snapshot)); + var load = Message(() => CultCacheMessagePack.Create(scratch, new CultCacheOpenOptions { Registry = Registry }).Dispose()); + Assert.That(load, Does.Contain("surface").And.Contain($"1 {Clause}")); + } + finally + { + Directory.Delete(work, recursive: true); + } + } + + // A directory store names a page by the SHA-256 of its bytes and its manifest commits to that hash: the page holding the + // payload is rewritten with a tail, renamed, and the manifest told its new hash, so the store is well formed and only the payload is wrong. + [Test] + public async System.Threading.Tasks.Task DirectoryLoadPageRefuses() + { + var folder = Path.Combine(Path.GetTempPath(), $"cultlib-bytesafter-dir-{Guid.NewGuid():N}"); + Directory.CreateDirectory(folder); + try + { + var path = Path.Combine(folder, "tail.cc"); + var options = new CultCacheOpenOptions { Registry = Registry, UseDirectoryStore = true, StoreFlushOnDispose = true }; + using (var writer = CultCacheMessagePack.Create(path, options)) + { + var value = CultDocumentMessagePackSerialization.DeserializeUntyped(typeof(UnionSurface), Base, Registry); + await writer.UpsertAsync(typeof(UnionSurface), value, new CultRecordKey("surface")); + await writer.FlushAsync(); + } + + var rehashed = new System.Collections.Generic.List<(byte[] Old, byte[] New)>(); + foreach (var file in Directory.EnumerateFiles(folder, "*.msgpack", SearchOption.AllDirectories).ToArray()) + { + var bytes = File.ReadAllBytes(file); + if (bytes.AsSpan().IndexOf(Base) < 0) continue; + var page = CultDocumentMessagePackSerialization.DeserializePersistedRecord(bytes); + page.Payload = Base.Concat(new byte[] { 0xc0 }).ToArray(); + var forged = CultDocumentMessagePackSerialization.SerializePersistedRecord(page); + File.Delete(file); + File.WriteAllBytes(Path.Combine(Path.GetDirectoryName(file)!, Convert.ToHexString(System.Security.Cryptography.SHA256.HashData(forged)).ToLowerInvariant() + ".msgpack"), forged); + rehashed.Add((System.Security.Cryptography.SHA256.HashData(bytes), System.Security.Cryptography.SHA256.HashData(forged))); + } + + Assert.That(rehashed, Is.Not.Empty, "the record page was found"); + foreach (var file in Directory.EnumerateFiles(folder, "*", SearchOption.AllDirectories).Where(f => !f.EndsWith(".msgpack", StringComparison.Ordinal)).ToArray()) + { + var bytes = File.ReadAllBytes(file); + foreach (var (old, @new) in rehashed) + for (var at = bytes.AsSpan().IndexOf(old); at >= 0; at = bytes.AsSpan().IndexOf(old)) @new.CopyTo(bytes, at); + File.WriteAllBytes(file, bytes); + } + + var thrown = Assert.CatchAsync(async () => { using var cache = await CultCacheMessagePack.OpenAsync(path, new CultCacheOpenOptions { Registry = Registry, UseDirectoryStore = true, ReadOnly = true }); }); + var chain = ""; + for (var e = thrown; e != null; e = e.InnerException) chain += e.Message + "\n"; + Assert.That(chain, Does.Contain($"1 {Clause}")); + } + finally + { + Directory.Delete(folder, recursive: true); + } + } + + public sealed class TwoValueFormatter : global::MessagePack.Formatters.IMessagePackFormatter + { + public void Serialize(ref MessagePackWriter writer, TwoValues value, MessagePackSerializerOptions options) + { + writer.WriteArrayHeader(0); + writer.Write(7); + } + + public TwoValues Deserialize(ref MessagePackReader reader, MessagePackSerializerOptions options) { reader.Skip(); return new TwoValues(); } + } + + [CultDocument("tests.two_values", "tests.two_values.v1")] + [MessagePackObject] + [MessagePackFormatter(typeof(TwoValueFormatter))] + public sealed class TwoValues + { + [Key(0)] [CultName] public string Id = string.Empty; + } + + [Test] + public void OwnFormatterWritingTwoValuesRefusedOnWrite() + { + var registry = CultDocumentRegistry.ForTypes(new[] { typeof(TwoValues) }); + var message = Message(() => CultDocumentMessagePackSerialization.SerializeUntyped(new TwoValues(), typeof(TwoValues), registry)); + Assert.That(message, Does.Contain(typeof(TwoValues).FullName).And.Contain($"1 {Clause}").And.Contain("refused on write")); + } + } +} diff --git a/tests/GameCult.Caching.Tests/PayloadDepthTests.cs b/tests/GameCult.Caching.Tests/PayloadDepthTests.cs index 91aa9a6f..1d3db1d0 100644 --- a/tests/GameCult.Caching.Tests/PayloadDepthTests.cs +++ b/tests/GameCult.Caching.Tests/PayloadDepthTests.cs @@ -148,10 +148,25 @@ namespace GameCult.Caching.Tests } [Test] - public void OnlyTheFirstValueIsThePayload() + public void OnlyOneValueIsAPayload() { - Accepts(Cat(Bytes(0x90), Chain(Limit + 10)), "bytes after the first value are not walked"); - Accepts(Cat(Bytes(0x01), Chain(Limit + 10)), "nor after a scalar"); + var tails = new[] + { + (Cat(Bytes(0x90), Chain(Limit + 10)), Chain(Limit + 10).Length), + (Cat(Bytes(0x01), Chain(Limit + 10)), Chain(Limit + 10).Length), + (Bytes(0x90, 0xc1, 0xc1), 2), + }; + foreach (var (payload, count) in tails) + foreach (var direction in new[] { PayloadDirection.Read, PayloadDirection.Write }) + { + var message = Assert.Throws(() => CultPayloadDepth.Check(payload, typeof(PayloadDepthTests), direction))!.Message; + var word = direction == PayloadDirection.Read ? "read" : "write"; + Assert.That(message, Is.EqualTo($"{typeof(PayloadDepthTests).FullName} payload has {count} bytes after its first MessagePack value (refused on {word}).")); + } + + // The tail is not walked: a tail nested past the limit is refused as bytes after the value, not as depth. + var deep = Assert.Throws(() => CultPayloadDepth.Check(Cat(Bytes(0x90), Chain(Limit + 10)), typeof(object), PayloadDirection.Read))!; + Assert.That(deep.Message, Does.Not.Contain("the limit is")); } } } diff --git a/tests/GameCult.Mesh.Tests/BytesAfterPathTests.cs b/tests/GameCult.Mesh.Tests/BytesAfterPathTests.cs new file mode 100644 index 00000000..ed9b904b --- /dev/null +++ b/tests/GameCult.Mesh.Tests/BytesAfterPathTests.cs @@ -0,0 +1,61 @@ +using System; +using System.Linq; +using System.Reflection; +using FluentAssertions; +using GameCult.Caching; +using GameCult.Caching.MessagePack; +using GameCult.Networking; +using NUnit.Framework; + +namespace GameCult.Mesh.Tests; + +// The mesh's typed reads of a raw record decode the payload through the one-value rule: a tail is refused, never ignored. +public class BytesAfterPathTests +{ + private const string Clause = "bytes after its first MessagePack value"; + + private static CultNetRawDocumentRecord WithTail(CultMeshCdnArtifactManifest manifest) + { + var payload = CultDocumentMessagePackSerialization.SerializeUntyped(manifest, typeof(CultMeshCdnArtifactManifest)).Concat(new byte[] { 0xc0 }).ToArray(); + return new CultNetRawDocumentRecord + { + SchemaId = CultDocumentRegistry.Shared.GetRequired().SchemaId, + RecordKey = "k", + StoredAt = "1970-01-01T00:00:00.000Z", + Payload = payload, + }; + } + + private static CultMeshCdnArtifactManifest Manifest() => new() { ArtifactId = "a", Version = "1", ContentHash = "h" }; + + private static string Refusal(Action read) + { + var thrown = Assert.Catch(read)!; + return thrown is TargetInvocationException { InnerException: { } inner } ? inner.ToString() : thrown.ToString(); + } + + [Test] + public void ASnapshotDecodeThroughABindingRefusesATail() + { + var registry = CultMeshCdn.CreateDocumentRegistry(); + var snapshot = new CultNetSnapshotResponseRawMessage { Documents = new[] { WithTail(Manifest()) } }; + Refusal(() => CultMesh.DecodeSnapshotDocuments(snapshot, registry)).Should().Contain(Clause); + } + + [Test] + public void ASnapshotDecodeWithNoBindingRefusesATail() + { + var snapshot = new CultNetSnapshotResponseRawMessage { Documents = new[] { WithTail(Manifest()) } }; + Refusal(() => CultMesh.DecodeSnapshotDocuments(snapshot, new CultNetDocumentRegistry())).Should().Contain(Clause); + } + + [Test] + public void ATypedReadOfOneSnapshotRecordRefusesATail() + { + var record = WithTail(Manifest()); + var snapshot = new CultNetSnapshotResponseRawMessage { MessageId = "m", Documents = new[] { record } }; + var read = typeof(CultMesh).GetMethod("ReadDocumentFromSnapshotResponse", BindingFlags.NonPublic | BindingFlags.Static)! + .MakeGenericMethod(typeof(CultMeshCdnArtifactManifest)); + Refusal(() => read.Invoke(null, new object[] { snapshot, record.SchemaId, record.RecordKey })).Should().Contain(Clause); + } +} diff --git a/tests/GameCult.Mesh.Tests/CultMeshContentVectorTests.cs b/tests/GameCult.Mesh.Tests/CultMeshContentVectorTests.cs index da764cc5..0f627cfb 100644 --- a/tests/GameCult.Mesh.Tests/CultMeshContentVectorTests.cs +++ b/tests/GameCult.Mesh.Tests/CultMeshContentVectorTests.cs @@ -462,12 +462,12 @@ public sealed class CultMeshContentVectorTests private static byte[] Reencode(CultMeshContentChunkResponseMessage message) => CultNetSchemaMessageSerialization.Serialize(message); - /// The one reader of manifest bytes: the document reader production stores and reads manifests with (CultPayloadDepth 256 included). + /// The one reader of manifest bytes: the production CultNet binding of the manifest (CultPayloadDepth 256 and the one-value rule included). private static CultMeshCdnArtifactManifest ReadManifest(byte[] bytes) => - (CultMeshCdnArtifactManifest)CultDocumentMessagePackSerialization.DeserializeUntyped(typeof(CultMeshCdnArtifactManifest), bytes); + (CultMeshCdnArtifactManifest)CultMeshCdn.CreateDocumentRegistry().GetByDocumentType(typeof(CultMeshCdnArtifactManifest))!.PayloadDeserializer(bytes); private static byte[] Reencode(CultMeshCdnArtifactManifest manifest) => - CultDocumentMessagePackSerialization.SerializeUntyped(manifest, typeof(CultMeshCdnArtifactManifest)); + CultMeshCdn.CreateDocumentRegistry().GetByDocumentType(typeof(CultMeshCdnArtifactManifest))!.PayloadSerializer(manifest); private static CultMeshContentChunkRequestMessage Request(string id, string hash, string key, int size) => new() { diff --git a/tests/GameCult.Mesh.Tests/CultMeshProviderSessionWireTests.cs b/tests/GameCult.Mesh.Tests/CultMeshProviderSessionWireTests.cs index 75125148..ca4346aa 100644 --- a/tests/GameCult.Mesh.Tests/CultMeshProviderSessionWireTests.cs +++ b/tests/GameCult.Mesh.Tests/CultMeshProviderSessionWireTests.cs @@ -1,5 +1,6 @@ using System; using System.Collections.Generic; +using System.Linq; using GameCult.Networking; using MessagePack; using NUnit.Framework; @@ -168,6 +169,32 @@ public sealed class CultMeshProviderSessionWireTests }); } + [Test] + public void DecodeCommandDocumentRefusesBytesAfterItsFirstValue() + { + var command = new CultMeshProviderCommandWire + { + CommandId = "cmd-7", CommandKind = "pilot.thrust", ProviderId = "aetheria", ServiceInstanceId = "aetheria-42", + Payload = new Dictionary { ["thrust"] = 1 } + }; + var payload = Convert.FromBase64String(CultMeshProviderSessionWire.EncodePayload(command)); + CultNetRawDocumentRecord Record(byte[] bytes) => new() + { + SchemaId = CultMeshProviderSessionWireContract.CommandSchema, + RecordKey = "provider-command:aetheria:aetheria-42:cmd-7", + PayloadEncoding = "messagepack", + Payload = bytes + }; + + Assert.That(CultMeshProviderSessionWire.DecodeCommandDocument(Record(payload)).CommandId, Is.EqualTo("cmd-7")); + var thrown = Assert.Throws(() => + CultMeshProviderSessionWire.DecodeCommandDocument(Record(payload.Concat(new byte[] { 0xc0, 0xc0 }).ToArray())))!; + Assert.That(thrown.Message, Does.Contain("2 bytes after its first MessagePack value")); + var one = Assert.Throws(() => + CultMeshProviderSessionWire.DecodeCommandDocument(Record(payload.Concat(new byte[] { 0xc0 }).ToArray())))!; + Assert.That(one.Message, Does.Contain("1 bytes after its first MessagePack value")); + } + [Test] public void LeaseResponse_RoundTrips_WithApplicationAcceptance() { diff --git a/tests/GameCult.Networking.Tests/BytesAfterPathTests.cs b/tests/GameCult.Networking.Tests/BytesAfterPathTests.cs new file mode 100644 index 00000000..6a56ceb1 --- /dev/null +++ b/tests/GameCult.Networking.Tests/BytesAfterPathTests.cs @@ -0,0 +1,62 @@ +using System; +using System.Linq; +using GameCult.Caching; +using GameCult.Caching.MessagePack; +using MessagePack; +using NUnit.Framework; + +namespace GameCult.Networking.Tests +{ + // A record's payload is one MessagePack value: each CultNet path that decodes a payload refuses a tail, by the one rule in + // CultPayloadDepth.Check. One test per path family, so that pointing a path at a tolerant reader turns it red. + public class BytesAfterPathTests + { + private const string Clause = "1 bytes after its first MessagePack value"; + + [CultDocument("tests.bytes_after_net", "tests.bytes_after_net.v1")] + [MessagePackObject] + public sealed class TailDoc + { + [Key(0)] [CultName] public string Id = "a"; + [Key(1)] public int N = 3; + } + + private static readonly CultDocumentRegistry Documents = CultDocumentRegistry.ForTypes(new[] { typeof(TailDoc) }); + + private static byte[] WithTail() => + CultDocumentMessagePackSerialization.SerializeUntyped(new TailDoc(), typeof(TailDoc), Documents).Concat(new byte[] { 0xc0 }).ToArray(); + + private static CultNetRawDocumentRecord Raw(CultNetDocumentRegistry registry, byte[] payload) => new() + { + SchemaId = Documents.GetRequired(typeof(TailDoc)).SchemaId, + RecordKey = "a", + StoredAt = "1970-01-01T00:00:00.000Z", + Payload = payload, + }; + + private static void Refuses(TestDelegate read) => + Assert.That(Assert.Catch(read)!.ToString(), Does.Contain(Clause)); + + [Test] + public void TheGenericBindingRefusesATail() => + Refuses(() => CultNetDocumentBinding.ForDocument(Documents).PayloadDeserializer(WithTail())); + + [Test] + public void TheRuntimeTypeBindingRefusesATail() => + Refuses(() => CultNetDocumentBinding.ForDocument(typeof(TailDoc), Documents).PayloadDeserializer(WithTail())); + + [Test] + public void ARawEnvelopeThroughARegisteredBindingRefusesATail() + { + var registry = new CultNetDocumentRegistry(Documents).Register(CultNetDocumentBinding.ForDocument(Documents)); + Refuses(() => registry.DeserializeRawDocument(Raw(registry, WithTail()))); + } + + [Test] + public void ARawEnvelopeWithNoBindingRefusesATail() + { + var registry = new CultNetDocumentRegistry(Documents); + Refuses(() => registry.DeserializeRawDocument(Raw(registry, WithTail()))); + } + } +} diff --git a/tests/vectors/union-parity/README.md b/tests/vectors/union-parity/README.md index fbf9c6c1..525f5642 100644 --- a/tests/vectors/union-parity/README.md +++ b/tests/vectors/union-parity/README.md @@ -170,3 +170,22 @@ arrays around a canary string, which no formatter reads. | `depth-skipped-past-limit.msgpack` | 257 | refused, depth | refuse on read though the typed depth is under 10 | A walk of the payload must not recurse: 100000 nested arrays in an extra slot are refused on a 256 KiB stack. + +## Bytes after the value (bytes-after-outcomes.msgpack) + +A record's payload is exactly one MessagePack value. Each file is `union-surface.msgpack` followed by a tail the first value does not own; the C# reference writes them and every other runtime is held to the same refusal. + +| File | Tail after the value | outcome, detail | +|---|---|---| +| `bytes-after-nil.msgpack` | `c0` (1 byte) | refused, bytes-after | +| `bytes-after-never-used.msgpack` | `c1 c1` (2 bytes, the never-used code) | refused, bytes-after | +| `bytes-after-deep-tail.msgpack` | 300 nested fixarray(1) around a canary string (324 bytes), nested past the limit | refused, bytes-after | +| `bytes-after-second-document.msgpack` | a second copy of `union-surface.msgpack` | refused, bytes-after | + +`bytes-after-outcomes.msgpack` is an array of `[file, "refused", "UnionSurface", nil, "bytes-after"]` rows, the row shape of `bounds-outcomes.msgpack` but in its own file, so the bounds walkers (which assert 12 rows) are unchanged. + +Must: refuse on read, naming the type and the count of bytes after the first value (`1`, `2`, the tail length) in a clause that says "bytes after its first MessagePack value"; never echo the canary; never report the depth refusal ("the limit is") for any of them, because the tail is not walked; and a single-file store holding `bytes-after-nil.msgpack` as a record payload refuses on open. + +What a runtime walks: only the first value of the payload, to find where it ends (a leaf counts as complete at its own length, a container when its last child completes). It compares that end with the payload length; a payload that ends early is not bytes after the value and stays the decoder's to refuse. It never walks the tail, so the tail's own shape (a stray `c1`, 300 levels) is never judged. + +A CDN manifest read as a record follows the same rule; its vectors are `contracts/cultmesh/content-acceptance-vectors.json` (`trailing_*` are refusals). diff --git a/tests/vectors/union-parity/bytes-after-deep-tail.msgpack b/tests/vectors/union-parity/bytes-after-deep-tail.msgpack new file mode 100644 index 0000000000000000000000000000000000000000..2dcc82f9cf5235956c37ed6c9e429bfca556b629 GIT binary patch literal 601 zcmbQfytuR|EipNjc@o2%6!aU5${fLW866AE~O}j2j2oI3a+MgjvPa?*y267{?Fp%UW~* literal 0 HcmV?d00001 diff --git a/tests/vectors/union-parity/bytes-after-nil.msgpack b/tests/vectors/union-parity/bytes-after-nil.msgpack new file mode 100644 index 0000000000000000000000000000000000000000..3bbdaa3321d1d4def22b1d20c9bb3a044b211d1c GIT binary patch literal 278 zcmZ9GK@P$o5Jk0W+<2D)-Fcf4&^Cc!D2*ON2NSm%(QZ5hN0Wsy;H+kT{(pb2V?Wp} za5=*XzR7Pm@i;96`fR0A1}mT27zS?iz6=Y3WmwSr5f_&%DA+jIZe7<9AE4Y{nWUjt w#>EmV*{Ou^<_p=xn@`?wfX-zlK+_w`Twqw8^SeogR|{H8yGm==NwD)6Ka^HkvH$=8 literal 0 HcmV?d00001 diff --git a/tests/vectors/union-parity/bytes-after-outcomes.msgpack b/tests/vectors/union-parity/bytes-after-outcomes.msgpack new file mode 100644 index 00000000..8237924e --- /dev/null +++ b/tests/vectors/union-parity/bytes-after-outcomes.msgpack @@ -0,0 +1 @@ +”•·bytes-after-nil.msgpack§refused¬UnionSurfaceÀ«bytes-after•¾bytes-after-never-used.msgpack§refused¬UnionSurfaceÀ«bytes-after•½bytes-after-deep-tail.msgpack§refused¬UnionSurfaceÀ«bytes-after•Ù#bytes-after-second-document.msgpack§refused¬UnionSurfaceÀ«bytes-after \ No newline at end of file diff --git a/tests/vectors/union-parity/bytes-after-second-document.msgpack b/tests/vectors/union-parity/bytes-after-second-document.msgpack new file mode 100644 index 0000000000000000000000000000000000000000..80d136fa80a0680a2d93d8a5b7a42a3adbc660e1 GIT binary patch literal 554 zcmbQfytuR|EipNjc@o2GE Dx2>_` literal 0 HcmV?d00001 diff --git a/unity/org.gamecult.cultlib/CHANGELOG.md b/unity/org.gamecult.cultlib/CHANGELOG.md index a2a1efae..a4a2bc11 100644 --- a/unity/org.gamecult.cultlib/CHANGELOG.md +++ b/unity/org.gamecult.cultlib/CHANGELOG.md @@ -6,6 +6,7 @@ All notable changes to this package are documented in this file. ### Breaking +- A document payload with bytes after its first MessagePack value is refused on read and on write (`MessagePackSerializationException` naming the type and the count of bytes after the value, never the bytes); the tail is not walked, so a tail nested past the limit is refused as bytes after the value, not as depth. Before, the tail was ignored on read and dropped on the next write. A payload that ends early is still the deserializer's to refuse. A CDN manifest is a record payload and follows the same rule: the acceptance vectors `trailing_nil`, `trailing_reserved_code` and `trailing_truncated_array` in `contracts/cultmesh/content-acceptance-vectors.json` are now refusals. `CultMeshProviderSessionWire.DecodeCommandDocument` refuses a command record whose payload has bytes after its first value (`MessagePackSerializationException` with the count); before, the tail was ignored. Semver: breaking for C# consumers, a major bump. - A single-file store holding two records under one key is unreadable: the reader refuses it with `CultStoreUnreadableException` naming the key (keys compare by ordinal code units, no case folding or normalisation), so a cache, `ReadStoreSnapshot` and a flush refuse the file and leave it as it is. Before, the later record silently replaced the earlier. A directory store reads its manifest through the same reader and refuses a duplicate key the same way. Repair the duplicate with the runtime that wrote it before reading the store here. - A record whose payload carries a slot past the slots its reading type declares, at any depth (the record, a nested object, a list element, a union arm or a unit arm), is loaded read-only: it is served, and every staged write or removal of its key (`Upsert`, `UpsertVariant`, `Remove`, a batch, and `Flatten` of a variant) is refused with `CultSchemaConflictException` before anything is staged, and the store keeps the record byte for byte. Before, the extra slots were skipped on read and dropped on the next write. Declaring an id compatible consents only to depth-0 slots the schema comparison reports ignored under that id; any other extra slot keeps the record read-only whatever is declared, including under the reader's own id. A variant overriding a slot its type does not have is read-only too (it used to be shed on flush). Write such a record from the build that declares those slots. Missing trailing slots still default and stay writable. - `Flatten` (and `FlattenAsync`) of a variant whose base chain holds a record its store does not own (a record held read-only for extra slots, or undeclared drift that drops slots) is refused with `CultSchemaConflictException` naming that base record, and the store is left as it was. Before, the flatten wrote the base's decoded view into a plain record and the base's unread data was lost for that variant. Writing and removing the variant itself are still accepted: they persist only its delta. From 892765ba97824b8de8f16b7a74ec32a2020eeda8 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 04:24:07 +0000 Subject: [PATCH 02/12] Python: a record with unread slots is held read-only (slot-readonly-py) (#126) --- packages/cultcache-py/CHANGELOG.md | 2 + packages/cultcache-py/README.md | 11 + .../cultcache-py/src/cultcache_py/__init__.py | 2 + .../cultcache-py/src/cultcache_py/cache.py | 113 +++- .../src/cultcache_py/documents.py | 90 +++- .../cultcache-py/src/cultcache_py/stores.py | 16 +- .../cultcache-py/tests/test_slot_readonly.py | 492 ++++++++++++++++++ .../cultcache-py/tests/test_union_parity.py | 8 +- 8 files changed, 704 insertions(+), 30 deletions(-) create mode 100644 packages/cultcache-py/tests/test_slot_readonly.py diff --git a/packages/cultcache-py/CHANGELOG.md b/packages/cultcache-py/CHANGELOG.md index 95f82042..08fb4d1f 100644 --- a/packages/cultcache-py/CHANGELOG.md +++ b/packages/cultcache-py/CHANGELOG.md @@ -9,6 +9,8 @@ ### Breaking +- A record whose payload carries a slot past its reader's declared slots, at any depth (the document, a nested record, a list element, a union arm, a unit arm), is read, held read-only, and every write or removal of its key (`put`, `put_global`, `update`, `put_envelope`, `put_envelopes`, `delete`, `delete_global`) is refused with `SchemaConflictError` before any store is touched (`cultlib-gaps:ruling:extra-trailing-slots`). Reading dropped those slots silently and the next write shed them. `SchemaConflictError.unread_slots` is an `UnreadSlots(type_name, slot_count, declared_count, occurrences, depth0_only)` and the message names the type and both counts, never a slot value. A depth-0 mark is consented, and the record writable, only when the record is stored under an id the document lists in `compatible_schema_ids` and that id's catalog entry has slots past the document's, covering the mark's slot count (`cultlib-gaps:ruling:slot-mark-consent`); a nested mark, or a mark under the document's own id, keeps the record read-only whatever is declared. Missing trailing slots keep their declared defaults. An untouched marked record keeps its stored bytes through every other write. Added `CultCache.is_writable(document, key)` (advisory; the write still decides: it also answers False for a foreign-carried key, for a key held as a type the document does not own, and for a global document at a non-global key; `delete` asks the same judges, so removing a key held as another type is refused with `SchemaConflictError` unless the document owns that record's schema id (its own or one it lists compatible), in which case the removal removes that record, its file row and its lookups, as a write of the key would replace it; it returned silently and removed nothing), `DocumentDefinition.decode_payload_marked`, and `UnreadSlots`. A document with a custom `decode` (no slot codec) never marks. A store holding a variant record is still unreadable here. Implies a minor bump while the version is 0.x (0.3.0 to 0.4.0): code that wrote back a record from a newer writer now gets `SchemaConflictError`. + - A cache holds one record per key whatever its type (`variants:ruling:cache-refuses-retype`). A `put`, `put_envelope`, `put_envelopes` or `put_global` of a second type under a key the cache holds, or under a key a store carries foreign, is refused with `SchemaConflictError` naming the key, even with no store, and changes nothing; retype a key by deleting it, then putting the new type. A write may replace another registered type's record only when the written type owns its schema id (declares it or lists it compatible) and both types share one home store. A load that reads one key from two stores, whatever the types, is refused with `CultCacheError` naming the key and both types, and admits nothing. A store that holds a record no registered type claims now loads and is carried: the cache does not serve it, the store keeps its bytes through every write, `delete` or `put` at its key is refused with `SchemaConflictError`, and registering a type that declares its schema id claims it at the next load. It raised `CultCacheError` "Unknown persisted document type". Two global types share the key `__global__`, so a second global type in one cache is now refused, and a store that already holds two global types at `__global__` (a JSONL store written by caches that never pulled could produce one) no longer loads: the pull raises `CultCacheError` and admits nothing. C# refuses a second persisted global on pull the same way (`cultcache-store-composition.md`, "Globals never invented"); repair such a file by deleting one of the two records with the runtime that wrote it. - Added `foreign_records()` and the frozen `ForeignRecord(key, schema_id, schema_name, stored_at)`, sorted by key in UTF-16 code-unit order. `stores.retyped` and `stores.written_owns` lose their underscore: the cache judge and the single-file store's commit guard ask one predicate. diff --git a/packages/cultcache-py/README.md b/packages/cultcache-py/README.md index 43e6c90e..72d51562 100644 --- a/packages/cultcache-py/README.md +++ b/packages/cultcache-py/README.md @@ -97,6 +97,7 @@ store. - `update_global(...)` - `delete(...)` - `delete_global(...)` +- `is_writable(document, key)` (advisory: would a write or removal of the key be refused; see below) - `snapshot()` ## Indexes And Roles @@ -180,6 +181,16 @@ The payload is a MessagePack array. Field keys are durable slot indexes: - newly added fields should use new keys - fields with defaults tolerate missing or nil slots when older payloads are read +A payload carrying a slot past the slots a reader declares, at any depth (the document, a nested record, a list element, a +union arm, a unit arm), was written by a newer build. It is read, and the record is held read-only: `put`, `update`, +`put_envelope(s)`, `delete` and their global forms refuse its key with `SchemaConflictError` before any store is touched, and +`SchemaConflictError.unread_slots` is an `UnreadSlots(type_name, slot_count, declared_count, occurrences, depth0_only)` naming the +first type read with extra slots (never a slot value). `CultCache.is_writable(document, key)` answers the same question before +the write: False for such a record, for a key a store carries foreign, and for a key held as another type's record; the write +still decides. `DocumentDefinition.decode_payload_marked(payload)` returns the value and its `UnreadSlots` (or None). Declaring the +stored id in `compatible_schema_ids` consents to the loss of depth-0 slots that id's catalog entry reports; a nested mark never is. +A store keeps a read-only record's bytes through every other write. A document with a custom `decode` never marks. + That matches the Rust `#[derive(DatabaseEntry)]` formatter shape and the C# `[Key(n)]` intent: schema evolution comes from stable field slots, not source member order. diff --git a/packages/cultcache-py/src/cultcache_py/__init__.py b/packages/cultcache-py/src/cultcache_py/__init__.py index dc73ab1b..4873b6ed 100644 --- a/packages/cultcache-py/src/cultcache_py/__init__.py +++ b/packages/cultcache-py/src/cultcache_py/__init__.py @@ -9,6 +9,7 @@ from .documents import ( DatabaseEntryField, DocumentDefinition, UnionArm, + UnreadSlots, database_entry_field, define_database_entry_type, define_document_registry, @@ -38,6 +39,7 @@ __all__ = [ "SlotCodec", "StoreUnreadableError", "UnionArm", + "UnreadSlots", "leaves", "database_entry_field", "define_database_entry_type", diff --git a/packages/cultcache-py/src/cultcache_py/cache.py b/packages/cultcache-py/src/cultcache_py/cache.py index 315e64ac..85b01d89 100644 --- a/packages/cultcache-py/src/cultcache_py/cache.py +++ b/packages/cultcache-py/src/cultcache_py/cache.py @@ -8,7 +8,7 @@ from dataclasses import dataclass, field, replace from typing import Any, Callable, Generic, Iterable, TypeVar from .backing_store import BackingStore, CultCacheEnvelope -from .documents import DocumentDefinition, extract_value +from .documents import DocumentDefinition, UnreadSlots, extract_value from .stores import SchemaConflictError, retyped, written_owns T = TypeVar("T") @@ -123,6 +123,10 @@ class _State: # The records the last pull read that no registered type claims, sorted by key. Derived from the pull and swapped with the # rest of the state, written nowhere else; it decides only refusals at its keys. A store keeps their bytes. foreign: list["ForeignRecord"] = field(default_factory=list) + # The records held read-only, by key (a key holds one record) with the type that holds it: the payload carried slots past the + # declared ones and nothing consents to the loss. Derived from the load or the write that admitted the record and swapped with + # the rest of the state; it decides only refusals at its keys, and any other key is absent. + unread: dict[str, tuple[str, UnreadSlots]] = field(default_factory=dict) @dataclass(frozen=True) @@ -337,6 +341,7 @@ class CultCache: routes, generic = self._state.stores_by_type, self._state.generic_store seen_globals: set[str] = set() foreign: list[ForeignRecord] = [] + loaded_unread: dict[str, tuple[str, UnreadSlots]] = {} type_at_key: dict[str, tuple[str, int]] = {} # one record per key across every store: its type (or schema name) and store stores: list[BackingStore] = [] for candidate in [*self._state.stores_by_type.values(), self._state.generic_store]: @@ -376,8 +381,10 @@ class CultCache: loaded_legacy_global_keys[envelope.type] = envelope.key envelope = replace(envelope, key=GLOBAL_KEY) self._claim_key(type_at_key, envelope.key, envelope.type, position) - envelope = self._stamped(document, envelope) - value = document.decode_payload(envelope.payload) + arrived, envelope = envelope, self._stamped(document, envelope) + value, unread = self._decode_marked(document, arrived) + if unread is not None: + loaded_unread[envelope.key] = (envelope.type, unread) loaded_values.setdefault(envelope.type, {})[envelope.key] = value read_values[(envelope.type, envelope.key)] = value loaded_envelopes.setdefault(envelope.type, {})[envelope.key] = envelope @@ -393,6 +400,7 @@ class CultCache: self._state.lookup_keys, self._state.legacy_global_keys, self._state.foreign, + self._state.unread, ) = ( loaded_values, loaded_envelopes, @@ -401,6 +409,7 @@ class CultCache: loaded_lookup_keys, loaded_legacy_global_keys, sorted(foreign, key=lambda record: _utf16_order(record.key)), + loaded_unread, ) @staticmethod @@ -491,9 +500,9 @@ class CultCache: def put_envelope(self, document: DocumentDefinition[T], envelope: CultCacheEnvelope) -> T: self._assert_registered(document) self._check_envelope(document, envelope) - envelope = self._stamped(document, envelope) - value = document.decode_payload(envelope.payload) - self._write(document, [(envelope, value)], batch=False) + arrived, envelope = envelope, self._stamped(document, envelope) + value, unread = self._decode_marked(document, envelope) + self._write(document, [(envelope, value)], batch=False, unread={envelope.key: unread}) return value @_mutating @@ -501,13 +510,33 @@ class CultCache: self._assert_registered(document) values: list[T] = [] stamped: list[CultCacheEnvelope] = [] + unread: dict[str, UnreadSlots | None] = {} for envelope in envelopes: self._check_envelope(document, envelope) stamped.append(self._stamped(document, envelope)) - values.append(document.decode_payload(envelope.payload)) - self._write(document, list(zip(stamped, values)), batch=True) + value, unread[envelope.key] = self._decode_marked(document, stamped[-1]) + values.append(value) + self._write(document, list(zip(stamped, values)), batch=True, unread=unread) return values + @staticmethod + def _decode_marked(document: DocumentDefinition[Any], held_as: CultCacheEnvelope) -> tuple[Any, UnreadSlots | None]: + """The payload decoded under the document, with the mark of any slot it carried past the declared ones, or no mark when the + loss is consented. `held_as` is the envelope under the id the record is stored under: a load's arrival, a write's stamped + envelope, so a key is writable in memory exactly when it is after a reload. A depth-0 mark is consented only when that id is + one the document lists compatible and its catalog entry has slots past the document's that cover the mark's slot count (the + loss a compatible-id declaration consents to, cultlib-gaps:ruling:slot-mark-consent). Any other mark keeps the record read-only.""" + value, slots = document.decode_payload_marked(held_as.payload) + if slots is None: + return value, None + entry, stored = document.catalog_entry(), held_as.catalog_entry + if slots.depth0_only and stored is not None and held_as.schema_id in entry.compatible_schema_ids and held_as.schema_id != entry.schema_id: + declared = {member.slot for member in entry.members} + reported = [member.slot for member in stored.members if member.slot not in declared] + if reported and slots.slot_count <= max(reported) + 1: + return value, None + return value, slots + @staticmethod def _stamped(document: DocumentDefinition[Any], envelope: CultCacheEnvelope) -> CultCacheEnvelope: """The record under its registered schema's id, whatever id the envelope arrived under, so the catalog entry a store @@ -527,7 +556,12 @@ class CultCache: raise CultCacheError(f"Global document {document.type} must use key {GLOBAL_KEY}") def _write( - self, document: DocumentDefinition[Any], records: list[tuple[CultCacheEnvelope, Any]], *, batch: bool + self, + document: DocumentDefinition[Any], + records: list[tuple[CultCacheEnvelope, Any]], + *, + batch: bool, + unread: dict[str, UnreadSlots | None] | None = None, ) -> None: """Validates every record (home and name and index extractors included) before the store is touched; once the store accepts, applying to the cache has no fallible step.""" @@ -535,6 +569,7 @@ class CultCache: # A batch naming a key twice lands the last record for it, as C# CultCacheBatch stages one operation per key. The # judge, the store and the apply all read this one list. records = list({envelope.key: (envelope, value) for envelope, value in records}.values()) + self._refuse_unread([envelope.key for envelope, _ in records]) name_extractor = self._state.name_extractors.get(type) index_extractors = self._state.index_extractors.get(type, {}) admitted = [ @@ -566,6 +601,9 @@ class CultCache: self._remove_lookups(type, envelope.key) values[envelope.key] = value envelopes[envelope.key] = envelope + # A held mark refused this write above, so a key reaching here carries none to clear. + if (mark := (unread or {}).get(envelope.key)) is not None: + self._state.unread[envelope.key] = (type, mark) lookup_keys[envelope.key] = keys if keys.name is not None: self._state.names[(type, keys.name)] = envelope.key @@ -592,6 +630,51 @@ class CultCache: replaced.append((other, envelope.key)) return replaced + @_locked + def is_writable(self, document: DocumentDefinition[Any], key: str) -> bool: + """Whether a write or removal of the key by this document would pass the judges that run before any store is touched: False for + a record held read-only (its payload carried slots past the declared ones, at any depth, and nothing consents to the loss, so + writing it would drop data a newer writer stored), for a key a store carries foreign, and for a key held as another type's + record this document does not own. Advisory (the write still decides): it answers at call time, and a key nothing holds is + writable.""" + self._assert_registered(document) + try: + self._refuse_at(document, key) + except (SchemaConflictError, CultCacheError): + return False + return True + + def _refuse_at(self, document: DocumentDefinition[Any], key: str) -> list[tuple[str, str]]: + """The judges a write or removal of the key by this document passes before any store is touched, asked by `is_writable` and + by `delete`: a global document writes only its global key, a record held read-only is never replaced, and a key held as + another type's record or carried foreign is refused (`_refuse_retype`).""" + if document.global_document and key != GLOBAL_KEY: + raise CultCacheError(f"Global document {document.type} must use key {GLOBAL_KEY}") + self._refuse_unread([key]) + entry = document.catalog_entry() + probe = CultCacheEnvelope.create(key=key, type=document.type, payload=b"", schema_id=entry.schema_id, catalog_entry=entry) + return self._refuse_retype(document, [probe]) + + def _refuse_unread(self, keys: list[str]) -> None: + """The one check every write and removal asks before any store is touched.""" + for key in keys: + held = self._state.unread.get(key) + if held is None: + continue + type, slots = held + raise SchemaConflictError( + self._state.envelopes[type][key].schema_id or type, + [self._state.documents[type].catalog_entry().schema_name], + key, + message=( + f"{self._state.documents[type].catalog_entry().schema_name} record {key} is read-only: {slots.type_name} was read with " + f"{slots.slot_count} slots where this build declares {slots.declared_count} ({slots.occurrences} object(s) in the " + "record). Writing it would drop data a newer writer stored; read it with types that declare those slots, or " + "write it from that build." + ), + unread_slots=slots, + ) + @staticmethod def _foreign_refusal(record: ForeignRecord) -> SchemaConflictError: return SchemaConflictError( @@ -644,16 +727,18 @@ class CultCache: def _delete(self, document: DocumentDefinition[Any], key: str) -> None: self._assert_registered(document) - if key not in self._state.envelopes.get(document.type, {}): - carried = next((record for record in self._state.foreign if record.key == key), None) - if carried is not None: - raise self._foreign_refusal(carried) + # A key held as another type's record this document owns is removed, as a write of the key would replace it (C# Remove is by key). + owned = self._refuse_at(document, key) store = self._store_for_type(document.type) legacy_key = self._state.legacy_global_keys.get(document.type) if key == GLOBAL_KEY else None if store is not None: - store.delete(document.type, key if legacy_key is None else legacy_key) + # One record per key: the owned record when this document does not hold the key, else its own, in one store call. + holder = owned[0][0] if owned else document.type + store.delete(holder, key if legacy_key is None else legacy_key) if legacy_key is not None: del self._state.legacy_global_keys[document.type] + for other, _ in owned: + self._drop(other, key) self._drop(document.type, key) def _drop(self, type: str, key: str) -> None: diff --git a/packages/cultcache-py/src/cultcache_py/documents.py b/packages/cultcache-py/src/cultcache_py/documents.py index 6101c136..4f7dfe7d 100644 --- a/packages/cultcache-py/src/cultcache_py/documents.py +++ b/packages/cultcache-py/src/cultcache_py/documents.py @@ -1,6 +1,7 @@ from __future__ import annotations import json +import threading from dataclasses import MISSING, dataclass, field from collections.abc import Mapping, Sequence, Set from itertools import chain @@ -30,6 +31,49 @@ MAX_SLOT_KEY = 65535 _STACK_FAULTS: tuple[type[BaseException], ...] = (_MsgpackStackError,) +@dataclass(frozen=True) +class UnreadSlots: + """What a payload decode found past the slots this runtime declares, at any depth: the first type read with extra slots, how + many slots it carried against how many it declares, and how many objects in the payload did. It holds no slot value, so a + refusal can name it. `depth0_only` is True when every mark sat in the document's own record (the half of consent the payload + decides). A record carrying it is read-only unless the schema comparison explains every mark (see `CultCache.is_writable`).""" + + type_name: str + slot_count: int + declared_count: int + occurrences: int + depth0_only: bool + + +class _SlotScope: + """Opened by `DocumentDefinition.decode_payload_marked`; the record codec and the union codec note into it. It is thread-local, + so no codec signature carries it, and a codec decoded outside a payload decode notes nothing.""" + + def __init__(self) -> None: + self.depth = 0 # records entered: 0 while the document's own record is read + self.first: tuple[str, int, int] | None = None + self.occurrences = 0 + self.depth0_only = True + + def note(self, type_name: str, slot_count: int, declared_count: int, depth0: bool) -> None: + if self.first is None: + self.first = (type_name, slot_count, declared_count) + self.occurrences += 1 + self.depth0_only = self.depth0_only and depth0 + + def result(self) -> UnreadSlots | None: + if self.first is None: + return None + return UnreadSlots(*self.first, self.occurrences, self.depth0_only) + + +_scope = threading.local() + + +def _current_scope() -> _SlotScope | None: + return getattr(_scope, "open", None) + + _CONTAINERS = (Mapping, Sequence, Set) _SCALAR_SEQUENCES = (str, bytes, bytearray, memoryview) @@ -185,15 +229,25 @@ class DocumentDefinition(Generic[T]): return self.payload_encoder(tree) def decode_payload(self, payload: bytes) -> T: + return self.decode_payload_marked(payload)[0] + + def decode_payload_marked(self, payload: bytes) -> tuple[T, UnreadSlots | None]: + """The value and what the decode found past the declared slots, at any depth. The payload depth check runs first: a + payload past the limit is refused, never marked.""" try: tree = self.payload_decoder(payload) except _STACK_FAULTS: raise CultCachePayloadDepthError(self.type, "read") from None _refuse_past_depth_limit(tree, self.type, "read") - value = self.decode(tree) + outer, opened = _current_scope(), _SlotScope() + _scope.open = opened + try: + value = self.decode(tree) + finally: + _scope.open = outer if self.validate: self.validate(value) - return value + return value, opened.result() def __post_init__(self) -> None: # Judged here so every constructor passes it: the factories, the dataclass constructor and dataclasses.replace. @@ -418,6 +472,9 @@ def define_union( raise ValueError(f"Union {name!r} value is malformed: expected [armKey, [armSlots]]") arm_name, codec = by_key[key] if codec is None: + scope = _current_scope() + if scope is not None and raw[1]: + scope.note(arm_name, len(raw[1]), 0, False) return UnionArm(arm_name, None) try: return UnionArm(arm_name, codec.decode(raw[1])) @@ -479,17 +536,26 @@ def _record_codec( def decode(raw: Any) -> Any: if not isinstance(raw, list): raise ValueError(f"DatabaseEntry document type {type!r} expected a MessagePack array") + scope = _current_scope() + if scope is not None and len(raw) > max_key + 1: + scope.note(type, len(raw), max_key + 1, scope.depth == 0) values: dict[str, Any] = {} - for field in normalized: - if field.key < len(raw): - try: - values[field.name] = field.codec.decode(raw[field.key]) - except ValueError as error: - raise ValueError(f"{type} slot {field.key} ({field.name}): {error}") from None - elif field.default is MISSING: - values[field.name] = None - else: - values[field.name] = field.default() if callable(field.default) else field.default + if scope is not None: + scope.depth += 1 + try: + for field in normalized: + if field.key < len(raw): + try: + values[field.name] = field.codec.decode(raw[field.key]) + except ValueError as error: + raise ValueError(f"{type} slot {field.key} ({field.name}): {error}") from None + elif field.default is MISSING: + values[field.name] = None + else: + values[field.name] = field.default() if callable(field.default) else field.default + finally: + if scope is not None: + scope.depth -= 1 if cls is None: return values return cls(**values) diff --git a/packages/cultcache-py/src/cultcache_py/stores.py b/packages/cultcache-py/src/cultcache_py/stores.py index 61841956..c15ac32e 100644 --- a/packages/cultcache-py/src/cultcache_py/stores.py +++ b/packages/cultcache-py/src/cultcache_py/stores.py @@ -13,6 +13,7 @@ from typing import Any, Callable, Iterator, TypeVar import msgpack +from .documents import UnreadSlots from .backing_store import ( CultCacheEnvelope, CultCacheSchemaCatalogEntry, @@ -46,9 +47,19 @@ class SchemaConflictError(ValueError): """A write the catalog cannot describe: records of different types under one schema id, two arrived schemas that share an id and disagree on the schema name, or a record no chosen entry publishes. Nothing is written. It names the id, the names involved (`schema_names`) and a record key (`record_key`). A cache raises it too, with an empty record key, when a - document registers a schema id another registered document already carries.""" + document registers a schema id another registered document already carries. `unread_slots` is None except for the refusal to + write or remove a record held read-only because its payload carries slots past the declared ones; its message names the type + and both slot counts and never holds a slot value.""" - def __init__(self, schema_id: str, schema_names: list[str], record_key: str, *, message: str | None = None) -> None: + def __init__( + self, + schema_id: str, + schema_names: list[str], + record_key: str, + *, + message: str | None = None, + unread_slots: UnreadSlots | None = None, + ) -> None: super().__init__( message or f"Schema id {schema_id!r} cannot describe record {record_key!r}: {', '.join(repr(name) for name in schema_names)}" @@ -56,6 +67,7 @@ class SchemaConflictError(ValueError): self.schema_id = schema_id self.schema_names = schema_names self.record_key = record_key + self.unread_slots = unread_slots class JsonLinesBackingStore: diff --git a/packages/cultcache-py/tests/test_slot_readonly.py b/packages/cultcache-py/tests/test_slot_readonly.py new file mode 100644 index 00000000..77d4a989 --- /dev/null +++ b/packages/cultcache-py/tests/test_slot_readonly.py @@ -0,0 +1,492 @@ +from __future__ import annotations + +import dataclasses +import shutil +import threading +import tempfile +import unittest +from pathlib import Path + +import msgpack +from cultcache_py.cache import CultCacheError + +from cultcache_py import ( + CultCache, + CultCachePayloadDepthError, + SlotCodec, + SchemaConflictError, + SingleFileMessagePackBackingStore, + UnionArm, + UnreadSlots, + database_entry_field, + define_database_entry_type, + define_record_codec, + leaves as t, +) +import test_union_parity as _parity +from test_union_parity import ( + VECTORS, + _surface_value, + _vector, + surface_document, + unit_outcome, +) + +SLOT = "extra-slot" # the canary: the value of every extra slot in the vectors; no refusal may carry it + +# The vectors name the C# types; this runtime names a type by its codec's name, a document by its document type. +_NAMES = {"UnionSurface": surface_document.type, "SurfaceDeclaring": surface_document.type, "Denied2": "Denied"} + +slot_unit_document = define_database_entry_type( + "tests.slot_unit", + [database_entry_field("name", 0, codec=t.string), database_entry_field("outcome", 1, codec=unit_outcome)], + name="name", +) +# The reader of extra-slot-consent.cc: the surface document's slots under a new schema id that lists the old one compatible. +declaring_document = define_database_entry_type( + "tests.union_surface", + [ + database_entry_field("id", 0, codec=t.string), + database_entry_field("version", 1, codec=t.int32), + database_entry_field("root", 2, codec=_parity._node), + database_entry_field("outcomes", 3, codec=t.list(_parity.outcome)), + database_entry_field("part", 4, codec=t.nullable(_parity.item)), + database_entry_field("crafted", 5, codec=t.nullable(_parity.crafted)), + database_entry_field("missing", 6, codec=_parity._node), + ], + name="id", + schema_id="tests.union_surface.v2", + schema_name="tests.union_surface", + compatible_schema_ids=("tests.union_surface.wide",), +) + +# file -> (type, slots carried, slots declared, every mark at depth 0): the README table. +_PAYLOADS = { + "extra-slot-record.msgpack": ("UnionSurface", 8, 7, True), + "extra-slot-nested-record.msgpack": ("Composition", 3, 2, False), + "extra-slot-arm.msgpack": ("Box", 3, 2, False), + "extra-slot-list-element.msgpack": ("Text", 3, 2, False), + "depth-skipped-at-limit.msgpack": ("UnionSurface", 8, 7, True), +} + + +def _copy_store_without_variants(source: Path, target: Path) -> None: + """This runtime holds no variant records (a five-slot record makes a store unreadable here), so the store vector is + walked with its variant row removed; every other byte is the C# writer's.""" + [snapshot] = list(msgpack.Unpacker(open(source, "rb"), raw=False)) + snapshot[2] = [record for record in snapshot[2] if len(record) <= 4] + target.write_bytes(msgpack.packb(snapshot, use_bin_type=True)) + + +def _open(path: Path, *documents) -> CultCache: + builder = CultCache.builder() + for document in documents: + builder.register_document_type(document) + cache = builder.add_generic_store(SingleFileMessagePackBackingStore(path)).build() + cache.pull_all_backing_stores() + return cache + + +def _rows(): + return msgpack.unpackb(_vector("slot-outcomes.msgpack"), raw=False) + + +class PayloadMarkTests(unittest.TestCase): + def test_every_read_only_payload_decodes_and_names_its_mark(self) -> None: + for file, (type_name, carried, declared, depth0) in _PAYLOADS.items(): + with self.subTest(file=file): + value, mark = surface_document.decode_payload_marked(_vector(file)) + self.assertEqual(value, _surface_value()) + self.assertEqual(mark, UnreadSlots(_NAMES.get(type_name, type_name), carried, declared, 1, depth0)) + self.assertEqual(surface_document.decode_payload(_vector(file)), _surface_value()) + + def test_the_outcomes_manifest_still_counts_nine_rows_and_four_read_only(self) -> None: + rows = msgpack.unpackb(_vector("outcomes.msgpack"), raw=False) + self.assertEqual(len(rows), 9) + self.assertEqual([row[0] for row in rows if row[1] == "read-only"], [f for f in _PAYLOADS if f != "depth-skipped-at-limit.msgpack"]) + + def test_a_clean_payload_and_a_short_one_carry_no_mark(self) -> None: + self.assertEqual(surface_document.decode_payload_marked(_vector("union-surface.msgpack")), (_surface_value(), None)) + slots = msgpack.unpackb(_vector("union-surface.msgpack"), raw=False) + slots[2] = [1, ["only-id"]] # a Text arm with a missing trailing slot: defaults, no mark + _, mark = surface_document.decode_payload_marked(msgpack.packb(slots, use_bin_type=True)) + self.assertIsNone(mark) + + def test_a_retired_slot_below_the_highest_key_does_not_mark(self) -> None: + slots = msgpack.unpackb(_vector("union-surface.msgpack"), raw=False) + # Image's declared slots are 0, 1, 3 (2 retired): four carried, four declared. + image = slots[2][1][1][2] + self.assertEqual(image[0], 5) + image[1][2] = "retired-slot-value" + _, mark = surface_document.decode_payload_marked(msgpack.packb(slots, use_bin_type=True)) + self.assertIsNone(mark) + + def test_the_depth_check_runs_before_the_mark(self) -> None: + with self.assertRaises(CultCachePayloadDepthError): + surface_document.decode_payload_marked(_vector("depth-skipped-past-limit.msgpack")) + + def test_the_unit_arm_with_extra_slots_is_read_past_and_marked(self) -> None: + payload = msgpack.packb(["unit-arm", [2, [SLOT]]], use_bin_type=True) + value, mark = slot_unit_document.decode_payload_marked(payload) + self.assertEqual(value, {"name": "unit-arm", "outcome": UnionArm("Denied")}) + self.assertEqual(mark, UnreadSlots("Denied", 1, 0, 1, False)) + self.assertEqual(slot_unit_document.decode_payload_marked(msgpack.packb(["ok", [2, []]]))[1], None) + + def test_a_codec_decoded_outside_a_payload_decode_notes_nothing(self) -> None: + self.assertEqual(unit_outcome.decode([2, [SLOT]]), UnionArm("Denied")) + + def test_the_scope_belongs_to_its_thread(self) -> None: + def decode_elsewhere(raw): + thread = threading.Thread(target=lambda: unit_outcome.decode([2, [SLOT]])) + thread.start() + thread.join() + return raw + + outer = define_database_entry_type("tests.threaded", [database_entry_field("a", 0, codec=SlotCodec(lambda v: v, decode_elsewhere))]) + self.assertIsNone(outer.decode_payload_marked(msgpack.packb(["x"]))[1]) + + def test_a_nested_payload_decode_restores_the_outer_scope(self) -> None: + def decode_clean_inside(raw): + surface_document.decode_payload_marked(_vector("union-surface.msgpack")) + return raw + + inner = define_record_codec("Inner", [database_entry_field("x", 0, codec=t.string)]) + outer = define_database_entry_type( + "tests.nested", + [ + database_entry_field("a", 0, codec=SlotCodec(lambda v: v, decode_clean_inside)), + database_entry_field("b", 1, codec=inner), + ], + ) + _, mark = outer.decode_payload_marked(msgpack.packb(["x", ["v", SLOT]])) + self.assertEqual(mark, UnreadSlots("Inner", 2, 1, 1, False)) + + +class StoreRowTests(unittest.TestCase): + def setUp(self) -> None: + self.tmp = Path(tempfile.mkdtemp()) + self.addCleanup(shutil.rmtree, self.tmp, True) + + def _store(self, name: str) -> tuple[Path, CultCache]: + path = self.tmp / name + _copy_store_without_variants(VECTORS / name, path) + documents = (declaring_document,) if name == "extra-slot-consent.cc" else (surface_document, slot_unit_document) + return path, _open(path, *documents) + + def _document(self, store: str, key: str): + if key == "unit-arm": + return slot_unit_document + return declaring_document if store == "extra-slot-consent.cc" else surface_document + + def test_every_row_is_writable_exactly_where_the_vector_says_and_refuses_with_its_mark(self) -> None: + walked = 0 + for store, key, writable, explained, type_name, carried, declared, occurrences, flatten in _rows(): + if flatten is not None: + continue # a variant: this runtime holds none (deviation) + walked += 1 + with self.subTest(store=store, key=key): + path, cache = self._store(store) + document = self._document(store, key) + before = path.read_bytes() + self.assertEqual(cache.is_writable(document, key), writable) + held = cache.get(document, key) + self.assertIsNotNone(held) + if writable: + continue + self.assertEqual(explained, False) + for attempt in ( + lambda: cache.put(document, key, held), + lambda: cache.delete(document, key), + lambda: cache.put_envelope(document, cache.get_required_envelope(document, key)), + lambda: cache.put_envelopes(document, [cache.get_required_envelope(document, key)]), + lambda: cache.update(document, key, lambda value: value), + ): + with self.assertRaises(SchemaConflictError) as caught: + attempt() + refusal = caught.exception + self.assertEqual( + refusal.unread_slots, + UnreadSlots(_NAMES.get(type_name, type_name), carried, declared, occurrences, refusal.unread_slots.depth0_only), + ) + self.assertEqual(refusal.record_key, key) + message = str(refusal) + self.assertNotIn(SLOT, message) + self.assertIn(refusal.unread_slots.type_name, message) + self.assertIn(f"{carried} slots", message) + self.assertIn(f"declares {declared}", message) + self.assertEqual(path.read_bytes(), before) + self.assertTrue(cache.get(document, key) is held) + self.assertEqual(walked, 7) # eight rows, the variant not walked + + def test_a_key_no_store_holds_and_another_key_are_writable_while_a_marked_one_is_held(self) -> None: + path, cache = self._store("extra-slot-store.cc") + self.assertFalse(cache.is_writable(surface_document, "marked")) + self.assertTrue(cache.is_writable(surface_document, "nothing-here")) + cache.put(surface_document, "other", cache.get_required(surface_document, "clean")) + self.assertEqual(cache.get(surface_document, "other"), _surface_value()) + + def test_consent_lands_an_explained_mark_and_refuses_the_others(self) -> None: + path, cache = self._store("extra-slot-consent.cc") + self.assertTrue(cache.is_writable(declaring_document, "consented")) + cache.put(declaring_document, "consented", cache.get_required(declaring_document, "consented")) + [stored] = [e for e in SingleFileMessagePackBackingStore(path).pull_all() if e.key == "consented"] + self.assertEqual(stored.schema_id, "tests.union_surface.v2") + self.assertEqual(len(msgpack.unpackb(stored.payload, raw=False)), 7) # the consented slot is gone + self.assertTrue(cache.is_writable(declaring_document, "consented")) + for key in ("over-consent", "nested"): + self.assertFalse(cache.is_writable(declaring_document, key), key) + with self.assertRaises(SchemaConflictError): + cache.put(declaring_document, key, cache.get_required(declaring_document, key)) + + def test_a_reload_marks_again_and_a_write_of_a_clean_envelope_over_a_marked_key_is_refused(self) -> None: + path, cache = self._store("extra-slot-store.cc") + envelope = cache.get_required_envelope(surface_document, "clean") + with self.assertRaises(SchemaConflictError): + cache.put_envelope(surface_document, type(envelope)(**{**envelope.__dict__, "key": "marked"})) + cache.pull_all_backing_stores() + self.assertFalse(cache.is_writable(surface_document, "marked")) + + def test_a_written_envelope_that_carries_extra_slots_is_held_read_only(self) -> None: + path, cache = self._store("extra-slot-store.cc") + envelope = cache.get_required_envelope(surface_document, "clean") + carrying = type(envelope)(**{**envelope.__dict__, "key": "arrived", "payload": _vector("extra-slot-arm.msgpack")}) + cache.put_envelope(surface_document, carrying) + self.assertFalse(cache.is_writable(surface_document, "arrived")) + with self.assertRaises(SchemaConflictError): + cache.delete(surface_document, "arrived") + cache.pull_all_backing_stores() + self.assertFalse(cache.is_writable(surface_document, "arrived")) + + def test_a_batch_of_envelopes_that_carry_extra_slots_is_held_read_only(self) -> None: + path, cache = self._store("extra-slot-store.cc") + envelope = cache.get_required_envelope(surface_document, "clean") + batch = [ + type(envelope)(**{**envelope.__dict__, "key": key, "payload": _vector(file)}) + for key, file in (("b1", "extra-slot-arm.msgpack"), ("b2", "union-surface.msgpack")) + ] + cache.put_envelopes(surface_document, batch) + self.assertEqual([cache.is_writable(surface_document, key) for key in ("b1", "b2")], [False, True]) + + def test_a_mark_under_an_id_the_reader_does_not_list_compatible_is_not_consented(self) -> None: + unlisted = define_database_entry_type( + "tests.union_surface", + [database_entry_field("id", 0, codec=t.string)] + [database_entry_field(n, i, codec=c) for n, i, c in ( + ("version", 1, t.int32), ("root", 2, _parity._node), ("outcomes", 3, t.list(_parity.outcome)), + ("part", 4, t.nullable(_parity.item)), ("crafted", 5, t.nullable(_parity.crafted)), ("missing", 6, _parity._node))], + name="id", + schema_id="tests.union_surface.v2", + schema_name="tests.union_surface", + ) + path = self.tmp / "unlisted.cc" + _copy_store_without_variants(VECTORS / "extra-slot-consent.cc", path) + cache = _open(path, unlisted) + self.assertFalse(cache.is_writable(unlisted, "consented")) + + def test_a_written_envelope_is_judged_on_the_id_it_is_stored_under_so_a_reload_agrees(self) -> None: + path, cache = self._store("extra-slot-consent.cc") + arrived = next(e for e in SingleFileMessagePackBackingStore(path).pull_all() if e.key == "consented") + self.assertEqual(arrived.schema_id, "tests.union_surface.wide") + cache.put_envelope(declaring_document, dataclasses.replace(arrived, key="via-envelope")) + cache.put_envelopes(declaring_document, [dataclasses.replace(arrived, key="via-batch")]) + in_memory = [cache.is_writable(declaring_document, k) for k in ("via-envelope", "via-batch")] + cache.pull_all_backing_stores() + self.assertEqual([cache.is_writable(declaring_document, k) for k in ("via-envelope", "via-batch")], in_memory) + self.assertEqual(in_memory, [False, False]) + + def test_is_writable_agrees_with_the_write_for_a_key_a_store_carries_foreign(self) -> None: + path = self.tmp / "foreign.cc" + _copy_store_without_variants(VECTORS / "extra-slot-store.cc", path) + cache = _open(path, surface_document) # tests.slot_unit is not registered: its record is carried foreign + self.assertEqual([r.key for r in cache.foreign_records()], ["unit-arm"]) + self.assertFalse(cache.is_writable(surface_document, "unit-arm")) + with self.assertRaises(SchemaConflictError): + cache.put(surface_document, "unit-arm", _surface_value()) + with self.assertRaises(SchemaConflictError): + cache.delete(surface_document, "unit-arm") + + def test_every_key_of_a_batch_is_checked_wherever_the_marked_one_stands(self) -> None: + path, cache = self._store("extra-slot-store.cc") + before = path.read_bytes() + envelope = cache.get_required_envelope(surface_document, "clean") + fresh = dataclasses.replace(envelope, key="fresh") + marked = dataclasses.replace(envelope, key="marked") + for batch in ([fresh, marked], [marked, fresh], [fresh, dataclasses.replace(envelope, key="f2"), marked]): + with self.assertRaises(SchemaConflictError) as caught: + cache.put_envelopes(surface_document, batch) + self.assertEqual(caught.exception.record_key, "marked") + self.assertEqual(path.read_bytes(), before) + self.assertIsNone(cache.get(surface_document, "fresh")) + + def test_only_the_marks_that_hold_a_record_read_only_are_kept_by_key(self) -> None: + _, store = self._store("extra-slot-store.cc") + self.assertEqual({k: v[0] for k, v in store._state.unread.items()}, {"marked": surface_document.type, "unit-arm": "tests.slot_unit"}) + _, consent = self._store("extra-slot-consent.cc") + self.assertEqual(sorted(consent._state.unread), ["nested", "over-consent"]) # the consented mark is no state + + def test_an_arrival_with_no_catalog_entry_under_a_listed_id_is_not_consented(self) -> None: + wide = next(e for e in SingleFileMessagePackBackingStore(VECTORS / "extra-slot-consent.cc").pull_all() if e.key == "consented") + bare = dataclasses.replace(wide, catalog_entry=None) + + class Held: + def pull_all(self): + return [bare] + + def push(self, envelope): ... + + def push_all(self, envelopes): ... + + def delete(self, type, key): ... + + cache = CultCache.builder().register_document_type(declaring_document).add_generic_store(Held()).build() + cache.pull_all_backing_stores() + self.assertFalse(cache.is_writable(declaring_document, "consented")) + + def test_a_removal_is_refused_where_is_writable_says_no(self) -> None: + path, cache = self._store("extra-slot-store.cc") + before = path.read_bytes() + other = define_database_entry_type("tests.other", [database_entry_field("a", 0, codec=t.string)], name="a") + cache2 = _open(self.tmp / "extra-slot-store.cc", surface_document, slot_unit_document, other) + self.assertFalse(cache2.is_writable(other, "clean")) + with self.assertRaises(SchemaConflictError): + cache2.delete(other, "clean") + self.assertEqual(path.read_bytes(), before) + self.assertIsNotNone(cache2.get(surface_document, "clean")) + self.assertTrue(cache2.is_writable(other, "nothing-here")) + + def test_a_global_document_is_not_writable_at_a_key_other_than_the_global_one(self) -> None: + solo = define_database_entry_type("tests.solo", [database_entry_field("a", 0, codec=t.string)], global_document=True) + cache = CultCache.builder().register_document_type(solo).build() + self.assertFalse(cache.is_writable(solo, "elsewhere")) + with self.assertRaises(CultCacheError): + cache.put(solo, "elsewhere", {"a": "x"}) + with self.assertRaises(CultCacheError): + cache.delete(solo, "elsewhere") + self.assertTrue(cache.is_writable(solo, CultCache.GLOBAL_KEY)) + + def test_a_reload_replaces_the_marks_wholesale(self) -> None: + path, cache = self._store("extra-slot-store.cc") + self.assertFalse(cache.is_writable(surface_document, "marked")) + # The store is repaired by its owner: the marked record is rewritten clean. The reload must forget the old mark. + [snapshot] = list(msgpack.Unpacker(open(path, "rb"), raw=False)) + clean = next(r for r in snapshot[2] if r[0] == "clean") + for record in snapshot[2]: + if record[0] in ("marked", "unit-arm"): + record[3] = clean[3] if record[0] == "marked" else record[3] + snapshot[2] = [r for r in snapshot[2] if r[0] != "unit-arm"] + path.write_bytes(msgpack.packb(snapshot, use_bin_type=True)) + cache.pull_all_backing_stores() + self.assertTrue(cache.is_writable(surface_document, "marked")) + self.assertEqual(cache._state.unread, {}) + + def test_a_listed_id_whose_catalog_reports_no_extra_slot_does_not_consent(self) -> None: + wide = next(e for e in SingleFileMessagePackBackingStore(VECTORS / "extra-slot-consent.cc").pull_all() if e.key == "consented") + same = dataclasses.replace(wide, catalog_entry=declaring_document.catalog_entry()) # the stored entry has no slot past the reader's + + class Held: + def pull_all(self): + return [same] + + def push(self, envelope): ... + + def push_all(self, envelopes): ... + + def delete(self, type, key): ... + + cache = CultCache.builder().register_document_type(declaring_document).add_generic_store(Held()).build() + cache.pull_all_backing_stores() + self.assertFalse(cache.is_writable(declaring_document, "consented")) + + def test_a_record_stored_under_the_readers_own_id_is_never_consented(self) -> None: + own = dataclasses.replace( + next(e for e in SingleFileMessagePackBackingStore(VECTORS / "extra-slot-consent.cc").pull_all() if e.key == "consented"), + schema_id="tests.union_surface.v2", + ) + wider = dataclasses.replace(own.catalog_entry, schema_id="tests.union_surface.v2", compatible_schema_ids=("tests.union_surface.v2",)) + listing = dataclasses.replace( + declaring_document, schema_id="tests.union_surface.v2", compatible_schema_ids=("tests.union_surface.wide", "tests.union_surface.v2") + ) + + class Held: + def pull_all(self): + return [dataclasses.replace(own, catalog_entry=wider)] + + def push(self, envelope): ... + + def push_all(self, envelopes): ... + + def delete(self, type, key): ... + + cache = CultCache.builder().register_document_type(listing).add_generic_store(Held()).build() + cache.pull_all_backing_stores() + self.assertFalse(cache.is_writable(listing, "consented")) + + def test_a_removal_by_the_owner_of_a_key_held_as_another_type_removes_it_as_a_write_would_replace_it(self) -> None: + a = define_database_entry_type( + "tests.owned_a", [database_entry_field("a", 0, codec=t.string)], name="a", indexes={"by_a": "a"} + ) + b = define_database_entry_type( + "tests.owned_b", + [database_entry_field("a", 0, codec=t.string)], + compatible_schema_ids=(a.catalog_entry().schema_id,), + ) + path = self.tmp / "owned.cc" + writer = _open(path, a) + writer.put(a, "k", {"a": "x"}) + writer.put(a, "other", {"a": "y"}) + cache = _open(path, a, b) + self.assertTrue(cache.is_writable(b, "k")) + cache.delete(b, "k") + self.assertIsNone(cache.get(a, "k")) + self.assertIsNone(cache.get_by_index(a, "by_a", "x")) + self.assertEqual([e.key for e in SingleFileMessagePackBackingStore(path).pull_all()], ["other"]) + self.assertTrue(cache.is_writable(b, "k")) + + def test_an_owners_removal_is_one_store_call_so_the_cache_and_the_store_agree_when_it_fails(self) -> None: + a = define_database_entry_type("tests.one_a", [database_entry_field("a", 0, codec=t.string)], name="a") + b = define_database_entry_type( + "tests.one_b", [database_entry_field("a", 0, codec=t.string)], compatible_schema_ids=(a.catalog_entry().schema_id,) + ) + path = self.tmp / "one.cc" + _open(path, a).put(a, "k", {"a": "x"}) + + class SecondDeleteFails(SingleFileMessagePackBackingStore): + calls = 0 + + def delete(self, type, key): + self.calls += 1 + if self.calls > 1: + raise OSError("second delete") + super().delete(type, key) + + store = SecondDeleteFails(path) + cache = CultCache.builder().register_document_type(a).register_document_type(b).add_generic_store(store).build() + cache.pull_all_backing_stores() + cache.delete(b, "k") + self.assertEqual(store.calls, 1) + self.assertIsNone(cache.get(a, "k")) + self.assertEqual(SingleFileMessagePackBackingStore(path).pull_all(), []) + + def test_writing_another_record_and_flushing_carries_the_marked_one_unchanged(self) -> None: + path, cache = self._store("extra-slot-store.cc") + original = {e.key: e for e in SingleFileMessagePackBackingStore(path).pull_all()} + cache.put(surface_document, "clean", {**_surface_value(), "version": 9}) + after = {e.key: e for e in SingleFileMessagePackBackingStore(path).pull_all()} + for key in ("marked", "unit-arm"): + self.assertEqual(after[key].payload, original[key].payload, key) + self.assertEqual(after[key].schema_id, original[key].schema_id, key) + self.assertEqual(after[key].stored_at, original[key].stored_at, key) + self.assertNotEqual(after["clean"].payload, original["clean"].payload) + self.assertEqual(after["marked"].catalog_entry, original["marked"].catalog_entry) + + def test_the_vector_store_loads_under_the_reference_documents_and_marks_nothing_else(self) -> None: + path, cache = self._store("extra-slot-store.cc") + self.assertTrue(cache.is_writable(surface_document, "clean")) + self.assertFalse(cache.is_writable(surface_document, "marked")) + self.assertFalse(cache.is_writable(slot_unit_document, "unit-arm")) + self.assertEqual( + cache.get(slot_unit_document, "unit-arm"), {"name": "unit-arm", "outcome": UnionArm("Denied")} + ) + + +if __name__ == "__main__": + unittest.main() diff --git a/packages/cultcache-py/tests/test_union_parity.py b/packages/cultcache-py/tests/test_union_parity.py index 451cd030..1e19ce8d 100644 --- a/packages/cultcache-py/tests/test_union_parity.py +++ b/packages/cultcache-py/tests/test_union_parity.py @@ -261,7 +261,9 @@ class UnionSurfaceVectorTests(unittest.TestCase): else: read_only += 1 self.assertEqual(outcome_kind, "read-only", file) - self.assertEqual(surface_document.decode_payload(payload), _surface_value(), file) + value, mark = surface_document.decode_payload_marked(payload) + self.assertEqual(value, _surface_value(), file) + self.assertIsNotNone(mark, file) # read past and marked: tests/test_slot_readonly.py pins what the mark holds self.assertEqual((refused, read_only), (5, 4)) @@ -526,7 +528,9 @@ class BoundsVectorTests(unittest.TestCase): surface_document.encode_payload(surface_document.decode_payload(payload)), _vector(detail), file ) elif outcome_kind == "read-only": - self.assertEqual(surface_document.decode_payload(payload), _surface_value(), file) + value, mark = surface_document.decode_payload_marked(payload) + self.assertEqual(value, _surface_value(), file) + self.assertIsNotNone(mark, file) else: with self.assertRaises(ValueError, msg=file) as caught: surface_document.decode_payload(payload) From 09b825d68939b66a633b2ad05654bf970e15f64a Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 04:44:03 +0000 Subject: [PATCH 03/12] Rust: a record with unread slots is held read-only (slot-readonly-rs) (#127) --- packages/cultcache-rs/CHANGELOG.md | 8 + packages/cultcache-rs/README.md | 11 + .../cultcache-rs-derive/src/lib.rs | 18 +- packages/cultcache-rs/src/lib.rs | 252 ++++++++- packages/cultcache-rs/tests/common/mod.rs | 235 +++++++++ packages/cultcache-rs/tests/slot_readonly.rs | 493 ++++++++++++++++++ packages/cultcache-rs/tests/union_parity.rs | 234 +-------- packages/cultmesh-rs/src/lib.rs | 53 ++ packages/cultnet-rs/src/replication.rs | 1 + 9 files changed, 1063 insertions(+), 242 deletions(-) create mode 100644 packages/cultcache-rs/tests/common/mod.rs create mode 100644 packages/cultcache-rs/tests/slot_readonly.rs diff --git a/packages/cultcache-rs/CHANGELOG.md b/packages/cultcache-rs/CHANGELOG.md index ed7cd60a..353facea 100644 --- a/packages/cultcache-rs/CHANGELOG.md +++ b/packages/cultcache-rs/CHANGELOG.md @@ -8,6 +8,13 @@ only in git history. ### Breaking +Semver: this is a breaking change in the 0.x line (0.3.x to 0.4.0). + +- A record whose payload carries unread slots is read-only. A payload with a slot past its type's declared slots, at any depth (record, nested record, list element, union arm, unit arm), still decodes, and every write or removal of its key is now refused before any store is touched: `put`, `put_envelope`, `put_raw_envelope`, `put_prepared_batch` (the whole batch) and `delete`, so `update` too. The refusal is `SchemaConflictError` with the new `unread_slots`; its message names the type and both slot counts and never a slot value. Holders of such a record used to overwrite it and drop the extra slots silently. The payload nesting refusal still runs first. A held payload this build cannot read at all (an arm it lacks from a newer writer, or a corrupt record) is read-only too: every write or removal is refused with an error naming the type and key (a newer writer's record is read with that build; a corrupt one is restored from a backup or removed at the store itself, as no cache call removes it), and `is_writable` answers false. `load_soa` is a write of every held key it omits or changes, so it refuses to drop or replace a read-only record. `is_writable` also answers false for a key a store carries as a foreign record and for a type other than the one holding the key, as the writes do. Each write over a held key decodes the held payload once to ask (about 40% of a `put` of a 200,000-element record, measured in release); keys nothing holds pay nothing. +- cultmesh-rs `pull_rudp_catalog_snapshot` applies documents one by one with `put_raw_envelope`, so a snapshot document for a key the node holds read-only ends the pull at that document and the documents before it stay applied. C# applies a raw snapshot as one cache commit, so there a refusal applies nothing: this is a divergence, owed by follow-up `mesh-snapshot-atomic` (stage the pull through `put_prepared_batch`). +- `SchemaConflictError` gains the public field `unread_slots: Option`; code that builds it with a struct literal must set it (`None`). +- A unit arm with slots (`[N, [extra...]]`) decodes as the unit arm and marks the record, where the derive refused it. +- The registration map's value changes from the schema name to a private struct holding the schema name and the unread-slots check. - A single-file store written as `cultcache.store.v3` is read and written: a store the C# runtime marked v3 (it holds a record with an element id: an owned list element with a `[CultElementId]` member) keeps its header here, and is carried through every rewrite with its ids; this runtime never drops the mark. A runtime from before this change refuses a v3 store by name. Rolling back to an older release after a store has been written as v3 is one-way: there is no carry-back, so restore from a copy taken before upgrading. - A write over a file the reader refuses (a zero-byte file, a truncated store, a stamp this runtime does not read) refuses with `CultCacheStoreUnreadable` and leaves the file as it is. A whole-store write used to replace such a file. This reverses 0.3.1, where `push_all` replaced a file with no readable stamp (healing it); the stamp rule now has one owner, the reader every write asks under the lock, and it accepts v1 and v3 (0.3.1 refused v3). - `#[cultcache(index)]` means unique, as in C# `RefuseIndexSharing`: within one type and one alias no two records hold one non-blank value. A put, batch, push or load that would make two records hold a value is refused, naming both keys and the alias, and a store that already holds a duplicate does not open. A grouping index is no longer expressible; a consumer that relied on shared values must drop the `index` and look up by another way. @@ -22,6 +29,7 @@ only in git history. ### Added +- `UnreadSlots { type_name, slot_count, declared_count, occurrences }`, `decode_payload_marked` (the decode that returns the mark) and `CultCache::is_writable(type, key)`. The derives' slot visitors note into a thread-local scope the decode entry opens (`__enter_slots`, `__note_unread_slots`, `__UnitArmSlots`, doc-hidden); a decode outside that entry notes nothing. - `CultCache::foreign_records()` and `ForeignRecord { key, schema_id, schema_name, stored_at }`: the records the last pull found that no registered type claims, sorted by key. - Lookups, matching the C# `GetByName`, `GetByIndex` and `GetGlobal`. `#[cultcache(name)]` and `#[cultcache(index)]` / `#[cultcache(index = "alias")]` on `DatabaseEntry` fields and `#[cultcache(global)]` on the struct give `CultCache::get_by_name`, `get_by_index` and `get_global`. An empty name and a blank index value are not found; records of one type sharing a name find the one with the smallest key (an index value is unique, below). A second record of a global type under another key is refused on every write path and on load with `CultCacheSecondGlobal`. Two `name` members, or two members with one index alias, are refused at compile time. A member's text comes from `CultLookupValue` (`String`, `Option` of one, integers, `bool`, `char`). - `SchemaConflictError`, the typed refusal of a write the catalog cannot describe: records of different types under one schema id, or an arrived schema and a registered one that share an id and disagree on the schema name. A whole-store write (`push_all`) refuses with it, names the schema id, the names involved and a record key, writes nothing and leaves the file as it is. A caller tells it apart with `error.downcast_ref::()`. diff --git a/packages/cultcache-rs/README.md b/packages/cultcache-rs/README.md index 8a8b8f5e..ed9701d5 100644 --- a/packages/cultcache-rs/README.md +++ b/packages/cultcache-rs/README.md @@ -236,6 +236,17 @@ assert_eq!(rmp_serde::from_slice::(&bytes)?, tree); # Ok::<(), Box>(()) ``` +## Records written by a newer build + +A payload that carries a slot past the slots its type declares, in a record, a nested record, a list element or a union +arm (a unit arm declares none), still decodes: the extra slots are skipped. The cache then holds that record read-only, +because writing it would drop what the newer writer stored. `put`, `put_envelope`, `put_raw_envelope`, +`put_prepared_batch` and `delete` of its key are refused with `SchemaConflictError`, whose `unread_slots` names the type +and the slot counts and never a value; `CultCache::is_writable(type, key)` answers the same check. An untouched +read-only record keeps its bytes through every other write. A short payload (missing trailing slots) is not marked, and +a payload nested past `MAX_PAYLOAD_DEPTH` is refused before anything is read. `decode_payload_marked` is the decode that +returns the `UnreadSlots` mark. This runtime has no compatible schema ids, so no mark is consented. + ## Structure Of Arrays The C# CultCache hot path exposes CPU-local SoA scans as: diff --git a/packages/cultcache-rs/cultcache-rs-derive/src/lib.rs b/packages/cultcache-rs/cultcache-rs-derive/src/lib.rs index c7f9a3a2..b426ae01 100644 --- a/packages/cultcache-rs/cultcache-rs-derive/src/lib.rs +++ b/packages/cultcache-rs/cultcache-rs-derive/src/lib.rs @@ -381,6 +381,7 @@ fn slot_codec(ident: &syn::Ident, slots: &[CultCacheSlot]) -> proc_macro2::Token where __A: ::serde::de::SeqAccess<'de>, { + let __ticket = ::cultcache_rs::__enter_slots(); #(#option_declarations)* let mut index = 0usize; while index <= #max_slot { @@ -390,7 +391,15 @@ fn slot_codec(ident: &syn::Ident, slots: &[CultCacheSlot]) -> proc_macro2::Token } index += 1; } - while let ::std::option::Option::Some(_) = seq.next_element::<::serde::de::IgnoredAny>()? {} + // The slots past the declared ones are skipped and counted: the decode entry marks the record + // that carried them read-only. + let mut __extra = 0usize; + while seq.next_element::<::serde::de::IgnoredAny>()?.is_some() { + __extra += 1; + } + if __extra > 0 { + ::cultcache_rs::__note_unread_slots(__ticket, stringify!(#ident), #max_slot + 1 + __extra, #max_slot + 1); + } Ok(#ident { #(#build_fields),* }) @@ -567,10 +576,13 @@ fn union_codec(input: DeriveInput) -> syn::Result { quote! { #key => { match seq - .next_element::<::std::option::Option<[(); 0]>>() + .next_element::<::std::option::Option<::cultcache_rs::__UnitArmSlots>>() .map_err(|_| ::serde::de::Error::custom(#arity_message))? { - ::std::option::Option::Some(::std::option::Option::Some(_)) => #ident::#variant, + ::std::option::Option::Some(::std::option::Option::Some(slots)) => { + slots.note(stringify!(#variant)); + #ident::#variant + } ::std::option::Option::Some(::std::option::Option::None) => { return ::std::result::Result::Err(::serde::de::Error::custom( format!(#nil_slots_message, #key), diff --git a/packages/cultcache-rs/src/lib.rs b/packages/cultcache-rs/src/lib.rs index 686547f4..59ffa1af 100644 --- a/packages/cultcache-rs/src/lib.rs +++ b/packages/cultcache-rs/src/lib.rs @@ -800,6 +800,154 @@ pub fn decode_payload(payload: &[u8]) -> Result< decode_grown(payload).map_err(opaque_decode_error) } +/// What a payload decode found past the slots this build declares, at any depth: the first type read with extra slots, +/// how many slots it carried against how many it declares, and how many objects in the payload did. It holds no slot +/// value, so a refusal can name it. A record whose payload carries it is read-only in the cache +/// ([`CultCache::is_writable`]); this runtime has no compatible schema ids, so every mark is unconsented. +#[derive(Clone, Debug, PartialEq, Eq)] +pub struct UnreadSlots { + /// The first type read with extra slots (a unit arm is named by its variant). + pub type_name: String, + /// The slots that object carried. + pub slot_count: usize, + /// The slots this build declares for it. + pub declared_count: usize, + /// The objects in the payload that carried extra slots. + pub occurrences: usize, +} + +/// The open mark scope of a decode: how many slot objects it has entered, and the mark of the first entered one that +/// carried extra slots (the reference names the first type read, which is the outermost: it is entered before its members). +#[derive(Default)] +struct MarkScope { + entered: usize, + mark: Option<(usize, UnreadSlots)>, +} + +thread_local! { + /// The open mark scope of a decode on this thread: `None` when no decode entry has one open, so a decode through + /// serde alone notes nothing. The slot visitors the derives write note into it. + static UNREAD_SLOTS: std::cell::RefCell> = const { std::cell::RefCell::new(None) }; +} + +/// The derives' sensor, first half: a slot object is entered. Its ticket orders it among the objects of the decode. +#[doc(hidden)] +pub fn __enter_slots() -> usize { + UNREAD_SLOTS.with(|scope| { + scope.borrow_mut().as_mut().map_or(0, |open| { + open.entered += 1; + open.entered + }) + }) +} + +/// The derives' sensor, second half: the object entered as `ticket`, of `type_name`, carried `carried` slots where this +/// build declares `declared`. +#[doc(hidden)] +pub fn __note_unread_slots(ticket: usize, type_name: &str, carried: usize, declared: usize) { + UNREAD_SLOTS.with(|scope| { + let mut scope = scope.borrow_mut(); + let Some(open) = scope.as_mut() else { + return; + }; + let occurrences = open.mark.as_ref().map_or(0, |(_, found)| found.occurrences) + 1; + if open.mark.as_ref().is_none_or(|(first, _)| ticket < *first) { + let named = UnreadSlots { type_name: type_name.to_string(), slot_count: carried, declared_count: declared, occurrences }; + open.mark = Some((ticket, named)); + } else if let Some((_, found)) = open.mark.as_mut() { + found.occurrences = occurrences; + } + }); +} + +/// A unit arm is a zero-slot record: its slots array is read and counted, not refused. +#[doc(hidden)] +pub struct __UnitArmSlots { + ticket: usize, + extra: usize, +} + +impl __UnitArmSlots { + /// Notes the arm if its array held any element. + pub fn note(self, variant: &str) { + if self.extra > 0 { + __note_unread_slots(self.ticket, variant, self.extra, 0); + } + } +} + +impl<'de> serde::Deserialize<'de> for __UnitArmSlots { + fn deserialize>(deserializer: D) -> std::result::Result { + struct Count; + impl<'de> serde::de::Visitor<'de> for Count { + type Value = __UnitArmSlots; + fn expecting(&self, formatter: &mut std::fmt::Formatter) -> std::fmt::Result { + formatter.write_str("a slots array") + } + fn visit_seq>(self, mut seq: A) -> std::result::Result { + let ticket = __enter_slots(); + let mut extra = 0usize; + while seq.next_element::()?.is_some() { + extra += 1; + } + Ok(__UnitArmSlots { ticket, extra }) + } + } + deserializer.deserialize_seq(Count) + } +} + +/// The scope one marked decode opens on this thread; closed when it ends, whatever the decode did. A marked decode is +/// never re-entered (only the cache's own checks open one), so scopes do not nest. +struct UnreadScope; + +impl UnreadScope { + fn open() -> Self { + UNREAD_SLOTS.with(|scope| *scope.borrow_mut() = Some(MarkScope::default())); + Self + } + + fn mark(&self) -> Option { + UNREAD_SLOTS.with(|scope| scope.borrow_mut().as_mut().and_then(|open| open.mark.take()).map(|(_, mark)| mark)) + } +} + +impl Drop for UnreadScope { + fn drop(&mut self) { + UNREAD_SLOTS.with(|scope| *scope.borrow_mut() = None); + } +} + +/// [`decode_payload`] with the mark: the same depth check and grown decode, inside a scope the derives' slot +/// visitors note into. A payload past the nesting limit is refused before the decode, never marked. +pub fn decode_payload_marked(payload: &[u8]) -> Result<(T, Option)> { + check_payload_depth(payload)?; + let scope = UnreadScope::open(); + let value = decode_grown(payload).map_err(opaque_decode_error)?; + Ok((value, scope.mark())) +} + +/// The registered check of a held payload: monomorphized by [`CultCache::register_entry_type`] so the untyped write +/// entries can decode a payload they cannot name. +/// A payload the type cannot decode at all (an arm this build lacks, a corrupt record) is an error here, and the held +/// record is read-only: the cache cannot tell what a writer of another build stored in it. +fn unread_slots_of(payload: &[u8]) -> Result> { + decode_payload_marked::(payload).map(|(_, mark)| mark) +} + +/// What the cache finds in a held payload when asked to write or remove its key. +enum HeldFinding { + Free, + Marked(UnreadSlots), + Unreadable(anyhow::Error), +} + +/// What a registered type is to the cache: its schema name and how to read the slots a held payload carries. +struct Definition { + schema_name: &'static str, + unread_slots: fn(&[u8]) -> Result>, +} + /// The nesting bound for an envelope admitted without a typed decode, from the same walk that /// [`decode_payload`] runs first. Names the entry's type and key, never the payload. fn require_payload_depth(entry: &CultCacheEnvelope) -> Result<()> { @@ -938,10 +1086,26 @@ pub struct SchemaConflictError { pub schema_id: String, pub schema_names: Vec, pub record_key: String, + /// Set when the refusal is a write or removal of a record held read-only for the slots its payload carries past + /// this build's; `None` for every other refusal. + pub unread_slots: Option, } impl std::fmt::Display for SchemaConflictError { fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + if let Some(unread) = &self.unread_slots { + return write!( + formatter, + "{} record {} is read-only: {} was read with {} slots where this build declares {} ({} object(s) in the record). \ + Writing it would drop data a newer writer stored; read it with types that declare those slots, or write it from that build.", + self.schema_names.first().map_or("", String::as_str), + self.record_key, + unread.type_name, + unread.slot_count, + unread.declared_count, + unread.occurrences + ); + } write!( formatter, "schema id {:?} cannot describe record {:?}: {:?}", @@ -2738,7 +2902,7 @@ struct CultCacheStoreRegistration { } pub struct CultCache { - definitions: BTreeMap, + definitions: BTreeMap, /// The registered types declared `#[cultcache(global)]`. globals: BTreeSet, schema_name_definitions: BTreeMap, @@ -2825,8 +2989,8 @@ impl CultCache { "CultCache entry types must declare a non-empty type" )); } - if let Some(existing_schema) = self.definitions.get(T::TYPE) - && *existing_schema != T::SCHEMA_NAME + if let Some(existing) = self.definitions.get(T::TYPE) + && existing.schema_name != T::SCHEMA_NAME { return Err(anyhow!( "CultCache already has a different definition registered for type {:?}", @@ -2875,7 +3039,7 @@ impl CultCache { } } } - self.definitions.insert(T::TYPE.to_string(), T::SCHEMA_NAME); + self.definitions.insert(T::TYPE.to_string(), Definition { schema_name: T::SCHEMA_NAME, unread_slots: unread_slots_of:: }); self.schema_name_definitions .insert(T::SCHEMA_NAME.to_string(), T::TYPE.to_string()); Ok(()) @@ -3036,6 +3200,60 @@ impl CultCache { Ok(Some(payload)) } + /// Whether a write or removal of `(type_id, key)` would be admitted: false when the held payload carries slots past + /// its registered type's declared slots, at any depth, or cannot be read by this build at all (the record is + /// read-only), when a store carries a foreign record at the key, or when another registered type holds it. True for a + /// key nothing holds. An unregistered type is an error. Advisory: the write entries ask the same checks themselves. + pub fn is_writable(&self, type_id: &str, key: &str) -> Result { + if !matches!(self.held_finding(type_id, key)?, HeldFinding::Free) { + return Ok(false); + } + let probe = CultCacheEnvelope { + key: key.to_string(), + r#type: type_id.to_string(), + payload: Vec::new(), + stored_at: String::new(), + schema_id: None, + }; + Ok(self.home_index(type_id).is_ok() && self.refuse_retype(&self.entries, &self.foreign, &probe).is_ok()) + } + + /// The one decode behind every write entry and [`CultCache::is_writable`]: the held envelope at `(type_id, key)` + /// read under the registered type's mark. + fn held_finding(&self, type_id: &str, key: &str) -> Result { + let Some(definition) = self.definitions.get(type_id) else { + bail!("No entry type registered for CultCache envelope type {type_id:?}"); + }; + let Some(held) = self.entries.get(&entry_id_parts(type_id, key)) else { + return Ok(HeldFinding::Free); + }; + Ok(match (definition.unread_slots)(&held.payload) { + Ok(None) => HeldFinding::Free, + Ok(Some(unread)) => HeldFinding::Marked(unread), + Err(error) => HeldFinding::Unreadable(error), + }) + } + + /// Refuses a write or removal of a record held read-only, before any store is touched: for the slots its payload + /// carries past this build's, or because this build cannot read the payload at all. + fn refuse_unread_slots(&self, type_id: &str, key: &str) -> Result<()> { + match self.held_finding(type_id, key)? { + HeldFinding::Free => Ok(()), + HeldFinding::Unreadable(error) => Err(error.context(format!( + "CultCache entry {type_id:?} at key {key:?} is held read-only: this build cannot read its payload. A newer writer's record is read with that build; a corrupt one is restored from a backup or removed at the store itself, not through the cache" + ))), + HeldFinding::Marked(unread) => { + let held = &self.entries[&entry_id_parts(type_id, key)]; + Err(anyhow::Error::new(SchemaConflictError { + schema_id: schema_identity(held).to_string(), + schema_names: vec![self.definitions[type_id].schema_name.to_string()], + record_key: key.to_string(), + unread_slots: Some(unread), + })) + } + } + } + pub fn get_required(&self, key: &str) -> Result { self.get::(key)? .ok_or_else(|| anyhow!("CultCache has no {:?} entry at key {:?}", T::TYPE, key)) @@ -3166,6 +3384,7 @@ impl CultCache { pub fn put(&mut self, key: impl Into, value: &T) -> Result { let (entry, parsed) = self.prepare_entry(key, value)?; + self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_second_global(&self.entries, &entry.r#type, &entry.key)?; self.refuse_retype(&self.entries, &self.foreign, &entry)?; let change = self.judge_unique(&[&entry])?; @@ -3266,6 +3485,7 @@ impl CultCache { )); } require_payload_depth(entry)?; + self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_claimed_id(entry)?; let Some(home) = self.home_index(&entry.r#type)? else { continue; @@ -3308,6 +3528,7 @@ impl CultCache { pub fn put_envelope(&mut self, entry: CultCacheEnvelope) -> Result { let parsed = self.validate_envelope::(&entry)?; + self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_claimed_id(&entry)?; self.refuse_second_global(&self.entries, &entry.r#type, &entry.key)?; self.refuse_retype(&self.entries, &self.foreign, &entry)?; @@ -3345,6 +3566,7 @@ impl CultCache { )); } require_payload_depth(&entry)?; + self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_claimed_id(&entry)?; self.refuse_second_global(&self.entries, &entry.r#type, &entry.key)?; self.refuse_retype(&self.entries, &self.foreign, &entry)?; @@ -3400,6 +3622,7 @@ impl CultCache { pub fn delete(&mut self, key: &str) -> Result { self.require_entry_type::()?; + self.refuse_unread_slots(T::TYPE, key)?; let id = entry_id_parts(T::TYPE, key); let Some(entry) = self.entries.get(&id).cloned() else { if let Some((_, foreign)) = self.foreign.iter().find(|(_, foreign)| foreign.key == key) { @@ -3427,7 +3650,9 @@ impl CultCache { /// Replaces the cache's view with the table's records. Every record is /// checked (registered type, and a home store once any store is attached) /// into a local image first, so a refused load admits nothing and leaves - /// the prior view intact. Loading writes no store. + /// the prior view intact. Loading writes no store, but a held record that is read-only (unread slots, or a payload + /// this build cannot read) may not be omitted or replaced by the table: the next batch would lay the image over the + /// store. pub fn load_soa(&mut self, table: CultCacheSoaTable) -> Result<()> { let mut loaded = BTreeMap::new(); for entry in table.into_envelopes()? { @@ -3442,6 +3667,12 @@ impl CultCache { self.refuse_retype(&loaded, &self.foreign, &entry)?; loaded.insert(entry_id(&entry), entry); } + // A load is a write of every held key it omits or changes: the one check refuses those held read-only. + for ((type_id, key), held) in &self.entries { + if loaded.get(&(type_id.clone(), key.clone())).is_none_or(|incoming| incoming.payload != held.payload) { + self.refuse_unread_slots(type_id, key)?; + } + } self.refuse_second_globals(&loaded)?; let holders = self.judge_image(&loaded)?; self.entries = loaded; @@ -3518,7 +3749,7 @@ impl CultCache { } for (slot, keys) in &ordered { if let [first, second, ..] = keys.as_slice() { - let schema = self.definitions.get(&slot.0).copied().unwrap_or(&slot.0); + let schema = self.definitions.get(&slot.0).map_or(slot.0.as_str(), |definition| definition.schema_name); let violation = [slot.1.as_str(), first.as_str(), second.as_str(), schema]; if least.is_none_or(|best| violation_cmp(&violation, &best).is_lt()) { least = Some(violation); @@ -3553,7 +3784,7 @@ impl CultCache { .find(|held_key| held_key.as_str() != key); match held_other { Some(existing_key) => Err(CultCacheSecondGlobal { - schema_name: self.definitions.get(type_id).copied().unwrap_or(type_id).to_string(), + schema_name: self.definitions.get(type_id).map_or(type_id, |definition| definition.schema_name).to_string(), existing_key: existing_key.clone(), key: key.to_string(), } @@ -3597,6 +3828,7 @@ impl CultCache { schema_id: schema_identity(entry).to_string(), schema_names: vec![resolved, entry.r#type.clone()], record_key: entry.key.clone(), + unread_slots: None, })), _ => Ok(()), } @@ -3611,7 +3843,7 @@ impl CultCache { fn require_entry_type(&self) -> Result<()> { match self.definitions.get(T::TYPE) { - Some(schema_name) if *schema_name == T::SCHEMA_NAME => Ok(()), + Some(definition) if definition.schema_name == T::SCHEMA_NAME => Ok(()), _ => Err(anyhow!( "CultCache entry type {:?} is not registered on this cache instance", T::TYPE @@ -3693,6 +3925,7 @@ fn encode_store_snapshot(entries: &[CultCacheEnvelope], disk: &DiskStore) -> Res schema_id, schema_names: vec![existing_type, entry.r#type.clone()], record_key: entry.key.clone(), + unread_slots: None, })); } } @@ -3710,6 +3943,7 @@ fn encode_store_snapshot(entries: &[CultCacheEnvelope], disk: &DiskStore) -> Res schema_id, schema_names: vec![arrived_type, document_type], record_key: key, + unread_slots: None, })); } } @@ -3846,6 +4080,7 @@ fn foreign_conflict(carried: &CultCacheEnvelope, written_type: &str) -> anyhow:: schema_id: schema_identity(carried).to_string(), schema_names: vec![carried.r#type.clone(), written_type.to_string()], record_key: carried.key.clone(), + unread_slots: None, }) } @@ -3856,6 +4091,7 @@ fn retype_conflict(held: &CultCacheEnvelope, written: &CultCacheEnvelope) -> any schema_id: schema_identity(written).to_string(), schema_names: vec![held.r#type.clone(), written.r#type.clone()], record_key: written.key.clone(), + unread_slots: None, }) } diff --git a/packages/cultcache-rs/tests/common/mod.rs b/packages/cultcache-rs/tests/common/mod.rs new file mode 100644 index 00000000..3f893c0f --- /dev/null +++ b/packages/cultcache-rs/tests/common/mod.rs @@ -0,0 +1,235 @@ +//! The union-parity vectors' types and constructed value, shared by the test files that walk them. +#![allow(dead_code)] + +use std::path::PathBuf; + +use cultcache_rs::CultSlots; +use cultcache_rs::CultUnion; +use cultcache_rs::DatabaseEntry; + + +#[derive(Clone, Debug, PartialEq, CultUnion)] +pub enum Node { + #[cultcache(key = 0)] + Box(BoxArm), + #[cultcache(key = 1)] + Text(TextArm), + #[cultcache(key = 2)] + Use(UseArm), + #[cultcache(key = 5)] + Image(ImageArm), + #[cultcache(key = 6)] + Wrap(WrapArm), +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct WrapArm { + #[cultcache(key = 0)] + pub inner: Option>, + #[cultcache(key = 1)] + pub id: String, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct BoxArm { + #[cultcache(key = 0)] + pub id: String, + #[cultcache(key = 1)] + pub children: Vec, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct TextArm { + #[cultcache(key = 0)] + pub id: String, + #[cultcache(key = 1)] + pub content: String, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct UseArm { + #[cultcache(key = 0)] + pub id: String, + #[cultcache(key = 1)] + pub composition: Composition, + #[cultcache(key = 2)] + pub slots: Vec>, +} + +// Slot 2 is retired: a gap written nil. +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct ImageArm { + #[cultcache(key = 0)] + pub id: String, + #[cultcache(key = 1)] + pub uri: String, + #[cultcache(key = 3)] + pub alt: String, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct Composition { + #[cultcache(key = 0)] + pub name: String, + #[cultcache(key = 1)] + pub body: Option>, +} + +#[derive(Clone, Debug, PartialEq, CultUnion)] +pub enum Outcome { + #[cultcache(key = 0)] + Pending(PendingArm), + #[cultcache(key = 1)] + Completed(CompletedArm), + #[cultcache(key = 2)] + Denied(IdArm), + #[cultcache(key = -2147483648)] + Floor(IdArm), + #[cultcache(key = 2147483647)] + Ceiling(IdArm), +} + +// The codec's unit arm (no members, written [key, []]), kept apart from the golden: a document list's arms carry ids. +#[derive(Clone, Debug, PartialEq, CultUnion)] +pub enum UnitOutcome { + #[cultcache(key = 2)] + Denied, + #[cultcache(key = -2147483648)] + Floor, + #[cultcache(key = 2147483647)] + Ceiling, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct PendingArm { + #[cultcache(key = 0)] + pub message: String, + #[cultcache(key = 1)] + pub id: String, +} + +// An arm with no members of its own: just its element id. +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct IdArm { + #[cultcache(key = 0)] + pub id: String, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct CompletedArm { + #[cultcache(key = 0)] + pub source_version: i32, + #[cultcache(key = 1)] + pub id: String, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct NamedArm { + #[cultcache(key = 0)] + pub name: String, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct UsesArm { + #[cultcache(key = 0)] + pub name: String, + #[cultcache(key = 1)] + pub uses: i32, +} + +#[derive(Clone, Debug, PartialEq, CultSlots)] +pub struct CountArm { + #[cultcache(key = 0)] + pub name: String, + #[cultcache(key = 1)] + pub count: i32, +} + +#[derive(Clone, Debug, PartialEq, CultUnion)] +pub enum Item { + #[cultcache(key = 0)] + Gem(NamedArm), + #[cultcache(key = 1)] + Tool(UsesArm), + #[cultcache(key = 2)] + Part(CountArm), +} + +// The same arms as Item under its own keys: key 2 is unknown to it. +#[derive(Clone, Debug, PartialEq, CultUnion)] +pub enum CraftedItem { + #[cultcache(key = 0)] + Tool(UsesArm), + #[cultcache(key = 1)] + Part(CountArm), +} + +#[derive(Clone, Debug, PartialEq, DatabaseEntry)] +#[cultcache(type = "tests.union_surface", schema = "tests.union_surface.v1")] +pub struct UnionSurface { + #[cultcache(key = 0)] + pub id: String, + #[cultcache(key = 1)] + pub version: i32, + #[cultcache(key = 2)] + pub root: Option, + #[cultcache(key = 3)] + pub outcomes: Vec, + #[cultcache(key = 4)] + pub part: Option, + #[cultcache(key = 5)] + pub crafted: Option, + #[cultcache(key = 6)] + pub missing: Option, +} + +pub fn text(id: &str, content: &str) -> Node { + Node::Text(TextArm { id: id.into(), content: content.into() }) +} + +pub fn image(id: &str, uri: &str, alt: &str) -> Node { + Node::Image(ImageArm { id: id.into(), uri: uri.into(), alt: alt.into() }) +} + +pub fn boxed(id: &str, children: Vec) -> Node { + Node::Box(BoxArm { id: id.into(), children }) +} + +pub fn constructed() -> UnionSurface { + let inner = boxed("0000000000d0", vec![text("0000000000d1", "x"), image("0000000000d2", "uri://a", "alt")]); + UnionSurface { + id: "surface".into(), + version: 3, + root: Some(boxed( + "0000000000c0", + vec![ + text("0000000000c1", "hello"), + Node::Use(UseArm { + id: "0000000000c2".into(), + composition: Composition { name: "comp".into(), body: Some(std::boxed::Box::new(inner)) }, + slots: vec![Some(text("0000000000e1", "slot")), None, Some(boxed("0000000000e2", vec![]))], + }), + image("0000000000c3", "uri://b", ""), + ], + )), + outcomes: vec![ + Outcome::Pending(PendingArm { message: "wait".into(), id: "0000000000a1".into() }), + Outcome::Completed(CompletedArm { source_version: 7, id: "0000000000a2".into() }), + Outcome::Denied(id_arm("0000000000a3")), + ], + part: Some(Item::Tool(UsesArm { name: "hammer".into(), uses: 12 })), + crafted: Some(CraftedItem::Part(CountArm { name: "gear".into(), count: 4 })), + missing: None, + } +} + +pub fn id_arm(id: &str) -> IdArm { + IdArm { id: id.into() } +} + +pub fn vector(name: &str) -> Vec { + let path = PathBuf::from(env!("CARGO_MANIFEST_DIR")) + .join("../../tests/vectors/union-parity") + .join(name); + std::fs::read(&path).unwrap_or_else(|error| panic!("{}: {error}", path.display())) +} diff --git a/packages/cultcache-rs/tests/slot_readonly.rs b/packages/cultcache-rs/tests/slot_readonly.rs new file mode 100644 index 00000000..5bffcd9a --- /dev/null +++ b/packages/cultcache-rs/tests/slot_readonly.rs @@ -0,0 +1,493 @@ +//! A record whose payload carries unread slots at any depth is read-only: the C# reference's union-parity vectors +//! (README "Extra trailing slot" and "Slot stores") walked through the cache. This runtime has no compatible schema +//! ids, so every mark is unconsented and no `explained` outcome can arise. + +mod common; + +use std::path::Path; + +use common::*; +use cultcache_rs::CultCache; +use cultcache_rs::CultCacheEnvelope; +use cultcache_rs::DatabaseEntry; +use cultcache_rs::SchemaConflictError; +use cultcache_rs::SingleFileMessagePackBackingStore; +use cultcache_rs::UnreadSlots; +use cultcache_rs::decode_payload_marked; + +/// A row of `slot-outcomes.msgpack`: store, key, writable, explained, type, slots, declared, occurrences, flattenRefusedBy. +type SlotRow = (String, String, bool, Option, Option, Option, Option, Option, Option); + +const SLOT_VALUE: &str = "extra-slot"; + +/// `tests.slot_unit`, a document `[Name, Outcome: UnitOutcome]`. +#[derive(Clone, Debug, PartialEq, DatabaseEntry)] +#[cultcache(type = "tests.slot_unit", schema = "tests.slot_unit.v1")] +struct SlotUnit { + #[cultcache(key = 0)] + name: String, + #[cultcache(key = 1)] + outcome: UnitOutcome, +} + +/// Two declared slots, the second defaulted: a payload of one slot is short, not marked. +#[derive(Clone, Debug, PartialEq, DatabaseEntry)] +#[cultcache(type = "tests.shortened", schema = "tests.shortened.v1")] +struct Shortened { + #[cultcache(key = 0)] + id: String, + #[cultcache(key = 1, default)] + note: String, +} + +fn registered() -> CultCache { + let mut cache = CultCache::new(); + cache.register_entry_type::().unwrap(); + cache.register_entry_type::().unwrap(); + cache +} + +/// Any MessagePack value a store holds, enough to drop one record from a store and write the rest back. +#[derive(serde::Serialize, serde::Deserialize)] +#[serde(untagged)] +enum Packed { + Nil(()), + Bool(bool), + Int(i64), + Text(String), + Bin(#[serde(with = "serde_bytes")] Vec), + Array(Vec), +} + +/// The committed store without its `marked-variant` record. A variant is a five-slot record, which this runtime's +/// store reader refuses whole (`the_variant_row_...`), so the other rows are walked from a copy without it. +fn without_variant_row(store: &[u8]) -> Vec { + let Packed::Array(mut parts) = rmp_serde::from_slice(store).unwrap() else { panic!("a store is an array") }; + let Packed::Array(records) = parts.last_mut().unwrap() else { panic!("a store ends in its records") }; + let before = records.len(); + records.retain(|record| !matches!(record, Packed::Array(fields) if matches!(fields.first(), Some(Packed::Text(key)) if key == "marked-variant"))); + assert_eq!(records.len(), before - 1); + rmp_serde::to_vec(&parts).unwrap() +} + +/// A copy of a committed store, opened with both types registered; the vector itself is never written. The slot store +/// is read without its variant row. +fn open_copy(file: &str, dir: &Path) -> (CultCache, std::path::PathBuf) { + let path = dir.join(file); + let bytes = if file == "extra-slot-store.cc" { without_variant_row(&vector(file)) } else { vector(file) }; + std::fs::write(&path, bytes).unwrap(); + let mut cache = registered(); + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + cache.pull_all_backing_stores().unwrap(); + (cache, path) +} + +fn envelope(key: &str, payload: Vec) -> CultCacheEnvelope { + CultCacheEnvelope { + key: key.into(), + r#type: UnionSurface::TYPE.into(), + payload, + stored_at: "2026-01-01T00:00:00Z".into(), + schema_id: None, + } +} + +fn conflict(error: &anyhow::Error) -> &SchemaConflictError { + error.downcast_ref::().unwrap_or_else(|| panic!("not a schema conflict: {error:#}")) +} + +fn mark(type_name: &str, slot_count: usize, declared_count: usize, occurrences: usize) -> UnreadSlots { + UnreadSlots { type_name: type_name.into(), slot_count, declared_count, occurrences } +} + +/// The marks the README table pins, by `detail`. The reference names an arm by its class; here an arm is the +/// `...Arm` struct the union wraps. +fn expected_mark(detail: &str) -> UnreadSlots { + match detail { + "record" | "depth" => mark("UnionSurface", 8, 7, 1), + "nested-record" => mark("Composition", 3, 2, 1), + "arm" => mark("BoxArm", 3, 2, 1), + "list-element" => mark("TextArm", 3, 2, 1), + other => panic!("unknown detail {other}"), + } +} + +fn assert_refused_with(error: &anyhow::Error, key: &str, expected: &UnreadSlots) { + let refusal = conflict(error); + assert_eq!(refusal.unread_slots.as_ref(), Some(expected)); + assert_eq!(refusal.record_key, key); + assert_eq!(refusal.schema_names.len(), 1); + let message = error.to_string(); + assert!(message.contains(&expected.type_name), "{message}"); + assert!(message.contains(&expected.slot_count.to_string()) && message.contains(&expected.declared_count.to_string()), "{message}"); + assert!(!message.contains(SLOT_VALUE), "a refusal echoed the slot value: {message}"); +} + +#[test] +fn every_read_only_vector_decodes_marked_and_is_not_writable() { + let rows: Vec<(String, String, String, Option, String)> = rmp_serde::from_slice(&vector("outcomes.msgpack")).unwrap(); + assert_eq!(rows.len(), 9); + let bounds: Vec = rmp_serde::from_slice(&vector("bounds-outcomes.msgpack")).unwrap(); + let mut cases: Vec<(String, String)> = rows + .into_iter() + .filter(|(_, outcome, ..)| outcome == "read-only") + .map(|(file, _, _, _, detail)| (file, detail)) + .collect(); + for entry in &bounds[1..] { + let fields: Vec<&str> = entry.as_array().unwrap().iter().filter_map(|field| field.as_str()).collect(); + if fields[1] == "read-only" { + cases.push((fields[0].to_string(), fields[3].to_string())); + } + } + assert_eq!(cases.len(), 5); + for (file, detail) in cases { + let payload = vector(&file); + let (value, found) = decode_payload_marked::(&payload).unwrap_or_else(|error| panic!("{file}: {error:#}")); + assert_eq!(value, constructed(), "{file}"); + assert_eq!(found, Some(expected_mark(&detail)), "{file}"); + + let mut cache = registered(); + let dir = tempfile::tempdir().unwrap(); + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(dir.path().join("s.cc"))).unwrap(); + cache.put_raw_envelope(envelope("k", payload.clone())).unwrap(); + assert_eq!(cache.get::("k").unwrap(), Some(constructed()), "{file}"); + assert!(!cache.is_writable(UnionSurface::TYPE, "k").unwrap(), "{file}"); + assert!(cache.is_writable(UnionSurface::TYPE, "unheld").unwrap(), "{file}"); + let error = cache.put("k", &constructed()).unwrap_err(); + assert_refused_with(&error, "k", &expected_mark(&detail)); + } +} + +#[test] +fn clean_and_short_payloads_are_writable_and_unmarked() { + let (value, found) = decode_payload_marked::(&vector("union-surface.msgpack")).unwrap(); + assert_eq!((value, found), (constructed(), None)); + // A payload short of its declared slots keeps its defaults and does not mark. + let (value, found) = decode_payload_marked::(&rmp_serde::to_vec(&("only",)).unwrap()).unwrap(); + assert_eq!((value.id.as_str(), value.note.as_str(), found), ("only", "", None)); +} + +#[test] +fn a_payload_past_the_nesting_limit_is_refused_never_marked() { + for file in ["depth-skipped-past-limit.msgpack", "depth-past-limit.msgpack"] { + let error = decode_payload_marked::(&vector(file)).unwrap_err(); + assert!(error.to_string().contains("the limit is 256"), "{file}: {error:#}"); + } +} + +#[test] +fn a_plain_serde_decode_marks_nothing_and_does_not_leak_a_scope() { + // No scope is open: the visitors note into nothing. + let value: UnionSurface = rmp_serde::from_slice(&vector("extra-slot-record.msgpack")).unwrap(); + assert_eq!(value, constructed()); + // A marked decode followed by a clean one: the first mark does not survive into the second. + assert!(decode_payload_marked::(&vector("extra-slot-record.msgpack")).unwrap().1.is_some()); + assert_eq!(decode_payload_marked::(&vector("union-surface.msgpack")).unwrap().1, None); +} + +#[test] +fn the_slot_store_rows_walk_through_the_cache() { + let rows: Vec = + rmp_serde::from_slice(&vector("slot-outcomes.msgpack")).unwrap(); + assert_eq!(rows.len(), 8); + let dir = tempfile::tempdir().unwrap(); + let mut walked = 0; + for (store, key, writable, _explained, type_name, slots, declared, occurrences, flatten_refused_by) in rows { + // The variant row is a record this runtime cannot hold: see the variant test. The consent store's `writable` and + // `explained` answer the compatible-id declaration, which this runtime has none of; its marks are walked below. + if flatten_refused_by.is_some() || store == "extra-slot-consent.cc" { + continue; + } + let case = dir.path().join(&key); + std::fs::create_dir_all(&case).unwrap(); + let (mut cache, _) = open_copy(&store, &case); + let type_id = if key == "unit-arm" { SlotUnit::TYPE } else { UnionSurface::TYPE }; + assert_eq!(cache.is_writable(type_id, &key).unwrap(), writable, "{store} {key}"); + let removal = if key == "unit-arm" { cache.delete::(&key) } else { cache.delete::(&key) }; + match writable { + true => assert!(removal.unwrap()), + false => { + let found = conflict(&removal.unwrap_err()).unread_slots.clone().unwrap(); + // The reference names the unit arm's class, Denied2; here the variant. + let expected_type = match type_name.as_deref().unwrap() { + "Denied2" => "Denied", + other => other, + }; + assert_eq!((found.type_name.as_str(), Some(found.slot_count), Some(found.declared_count), Some(found.occurrences)), (expected_type, slots, declared, occurrences)); + assert!(cache.snapshot_envelopes().iter().any(|held| held.key == key), "a refused removal removed {key}"); + } + } + walked += 1; + } + assert_eq!(walked, 3); +} + +#[test] +fn the_consent_store_marks_are_unconsented_here() { + let rows: Vec = + rmp_serde::from_slice(&vector("slot-outcomes.msgpack")).unwrap(); + let dir = tempfile::tempdir().unwrap(); + let mut found = 0; + for (store, key, _writable, _explained, _type_name, slots, declared, occurrences, _) in rows { + if store != "extra-slot-consent.cc" { + continue; + } + let (cache, _) = open_copy(&store, dir.path()); + let held = cache.snapshot_envelopes().into_iter().find(|held| held.key == key).unwrap(); + let (_, mark) = decode_payload_marked::(&held.payload).unwrap(); + match slots { + // The reference's `consented` row is writable there (a listed compatible id covers it); here every mark refuses. + None => assert_eq!(mark, None, "{key}"), + Some(slots) => { + let mark = mark.unwrap_or_else(|| panic!("{key} carried no mark")); + // The first type read is the document, whose reference name is SurfaceDeclaring. + assert_eq!(mark.type_name, "UnionSurface", "{key}"); + assert_eq!((Some(mark.slot_count), Some(mark.declared_count), Some(mark.occurrences)), (Some(slots), declared, occurrences), "{key}"); + assert!(!cache.is_writable(UnionSurface::TYPE, &key).unwrap(), "{key}"); + } + } + found += 1; + } + assert_eq!(found, 4); +} + +#[test] +fn every_write_entry_refuses_a_marked_key_before_any_store_is_touched() { + let dir = tempfile::tempdir().unwrap(); + let (mut cache, path) = open_copy("extra-slot-store.cc", dir.path()); + let expected = mark("Composition", 3, 2, 1); + let before = std::fs::read(&path).unwrap(); + let held = cache.snapshot_envelopes().into_iter().find(|held| held.key == "marked").unwrap(); + let clean = cache.snapshot_envelopes().into_iter().find(|held| held.key == "clean").unwrap(); + assert!(!cache.is_writable(UnionSurface::TYPE, "marked").unwrap()); + assert!(cache.is_writable(UnionSurface::TYPE, "clean").unwrap()); + + let replacing = envelope("marked", rmp_serde::to_vec(&constructed()).unwrap()); + assert_refused_with(&cache.put("marked", &constructed()).unwrap_err(), "marked", &expected); + assert_refused_with(&cache.put_envelope::(replacing.clone()).unwrap_err(), "marked", &expected); + assert_refused_with(&cache.put_raw_envelope(replacing.clone()).unwrap_err(), "marked", &expected); + assert_refused_with(&cache.put_prepared_batch(vec![replacing]).unwrap_err(), "marked", &expected); + assert_refused_with(&cache.delete::("marked").unwrap_err(), "marked", &expected); + assert_refused_with( + &cache.update::("marked", |current| current.unwrap()).unwrap_err(), + "marked", + &expected, + ); + + // A batch with a marked key refuses whole: its clean member is not written either. + let mut rewritten = clean.clone(); + rewritten.payload = rmp_serde::to_vec(&UnionSurface { version: 9, ..constructed() }).unwrap(); + let batch = vec![rewritten, envelope("marked", rmp_serde::to_vec(&constructed()).unwrap())]; + assert_refused_with(&cache.put_prepared_batch(batch).unwrap_err(), "marked", &expected); + + assert_eq!(std::fs::read(&path).unwrap(), before, "a refused write touched the store"); + assert_eq!(cache.snapshot_envelopes().into_iter().find(|entry| entry.key == "marked").unwrap(), held); + assert_eq!(cache.snapshot_envelopes().into_iter().find(|entry| entry.key == "clean").unwrap(), clean); + + // A write to another key succeeds, and the marked record's bytes, schema id and stored-at are carried unchanged. + cache.put("clean", &UnionSurface { version: 9, ..constructed() }).unwrap(); + cache.put("other", &constructed()).unwrap(); + assert!(cache.delete::("other").unwrap()); + let mut reopened = registered(); + reopened.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + reopened.pull_all_backing_stores().unwrap(); + let carried = reopened.snapshot_envelopes().into_iter().find(|entry| entry.key == "marked").unwrap(); + assert_eq!(carried, held); + assert_eq!(carried.payload, vector("extra-slot-nested-record.msgpack")); + assert_eq!(reopened.get::("clean").unwrap().unwrap().version, 9); +} + +#[test] +fn a_unit_arm_with_slots_decodes_as_the_unit_arm_and_marks() { + let (document, found) = decode_payload_marked::(&unit_arm_payload()).unwrap(); + assert_eq!(document, SlotUnit { name: "unit-arm".into(), outcome: UnitOutcome::Denied }); + assert_eq!(found, Some(mark("Denied", 1, 0, 1))); + + let dir = tempfile::tempdir().unwrap(); + let (mut cache, _) = open_copy("extra-slot-store.cc", dir.path()); + assert_eq!(cache.get::("unit-arm").unwrap(), Some(document.clone())); + assert!(!cache.is_writable(SlotUnit::TYPE, "unit-arm").unwrap()); + let error = cache.put("unit-arm", &document).unwrap_err(); + assert_refused_with(&error, "unit-arm", &mark("Denied", 1, 0, 1)); + assert_refused_with(&cache.delete::("unit-arm").unwrap_err(), "unit-arm", &mark("Denied", 1, 0, 1)); + + // Several elements are still one marked object, and a clean unit arm is unmarked. + let (_, found) = decode_payload_marked::(&rmp_serde::to_vec(&(2u64, ("a", "b", "c"))).unwrap()).unwrap(); + assert_eq!(found, Some(mark("Denied", 3, 0, 1))); + let (_, found) = decode_payload_marked::(&vector("unit-arm-denied.msgpack")).unwrap(); + assert_eq!(found, None); +} + +fn unit_arm_payload() -> Vec { + let dir = tempfile::tempdir().unwrap(); + let (cache, _) = open_copy("extra-slot-store.cc", dir.path()); + cache.snapshot_envelopes().into_iter().find(|held| held.key == "unit-arm").unwrap().payload +} + +#[test] +fn the_variant_row_of_the_slot_store_is_refused_whole_by_the_store_reader() { + // `marked-variant` is a five-slot record (a variant): this runtime holds no variants, and its v1 snapshot reader + // refuses such a record, so the store as written cannot load here and the other rows are walked without it. + let dir = tempfile::tempdir().unwrap(); + let path = dir.path().join("extra-slot-store.cc"); + std::fs::write(&path, vector("extra-slot-store.cc")).unwrap(); + let mut cache = registered(); + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + let error = cache.pull_all_backing_stores().unwrap_err(); + assert!(format!("{error:#}").contains("is not decodable"), "{error:#}"); +} + +fn slot_unit_envelope(key: &str, payload: Vec) -> CultCacheEnvelope { + CultCacheEnvelope { r#type: SlotUnit::TYPE.into(), ..envelope(key, payload) } +} + +#[test] +fn a_held_payload_the_type_cannot_read_is_read_only_and_refused_by_name() { + let mut cache = registered(); + // A newer writer's record: an arm this build lacks and an extra root slot. get cannot read it, and it is the data + // read-only exists to protect. + let newer = rmp_serde::to_vec(&("CANARY-name", (99u64, ("x",)), "CANARY-extra")).unwrap(); + cache.put_raw_envelope(slot_unit_envelope("n", newer)).unwrap(); + assert!(cache.get::("n").is_err()); + assert!(!cache.is_writable(SlotUnit::TYPE, "n").unwrap()); + let replacement = SlotUnit { name: "n".into(), outcome: UnitOutcome::Denied }; + let held = cache.snapshot_envelopes(); + for error in [ + cache.put("n", &replacement).unwrap_err(), + cache.put_raw_envelope(slot_unit_envelope("n", rmp_serde::to_vec(&replacement).unwrap())).unwrap_err(), + cache.delete::("n").unwrap_err(), + ] { + let message = format!("{error:#}"); + assert!(message.contains("\"n\"") && message.contains("tests.slot_unit") && message.contains("read-only"), "{message}"); + assert!(!message.contains("CANARY"), "{message}"); + } + assert_eq!(cache.snapshot_envelopes(), held); +} + +#[test] +fn a_corrupt_held_payload_is_refused_naming_the_key_never_its_bytes() { + let mut cache = registered(); + cache.put_raw_envelope(envelope("broken", rmp_serde::to_vec(&("CANARY-id", "CANARY-version")).unwrap())).unwrap(); + assert!(!cache.is_writable(UnionSurface::TYPE, "broken").unwrap()); + let error = cache.delete::("broken").unwrap_err(); + let message = format!("{error:#}"); + assert!(message.contains("\"broken\"") && message.contains("backup") && message.contains("store"), "{message}"); + assert!(!message.contains("build that wrote it"), "{message}"); + assert!(!message.contains("CANARY"), "{message}"); + assert!(cache.is_writable("tests.unregistered", "broken").is_err()); +} + +fn held_marked_and_clean() -> (CultCache, tempfile::TempDir, std::path::PathBuf) { + let dir = tempfile::tempdir().unwrap(); + let (cache, path) = open_copy("extra-slot-store.cc", dir.path()); + (cache, dir, path) +} + +#[test] +fn load_soa_does_not_replace_or_drop_a_marked_record() { + use cultcache_rs::CultCacheSoaTable; + let (mut cache, _dir, path) = held_marked_and_clean(); + let all = cache.snapshot_envelopes(); + let held = all.iter().find(|held| held.key == "marked").unwrap().clone(); + let without: Vec<_> = all.iter().filter(|held| held.key != "marked").cloned().collect(); + let mut replaced = all.clone(); + replaced.iter_mut().find(|held| held.key == "marked").unwrap().payload = rmp_serde::to_vec(&constructed()).unwrap(); + for table in [without, replaced] { + let error = cache.load_soa(CultCacheSoaTable::from_envelopes(table)).unwrap_err(); + assert_refused_with(&error, "marked", &mark("Composition", 3, 2, 1)); + assert_eq!(cache.snapshot_envelopes(), all, "a refused load changed the image"); + } + // The same record carried unchanged, with another key added, loads. + let mut kept = all.clone(); + kept.push(envelope("added", rmp_serde::to_vec(&constructed()).unwrap())); + cache.load_soa(CultCacheSoaTable::from_envelopes(kept)).unwrap(); + assert_eq!(cache.snapshot_envelopes().iter().find(|held| held.key == "marked"), Some(&held)); + // A clean record may be dropped by a load. + let no_clean: Vec<_> = cache.snapshot_envelopes().into_iter().filter(|held| held.key != "clean").collect(); + cache.load_soa(CultCacheSoaTable::from_envelopes(no_clean)).unwrap(); + // The marked record is still on disk after the next batch. + let rewrite = envelope("added", rmp_serde::to_vec(&UnionSurface { version: 5, ..constructed() }).unwrap()); + cache.put_prepared_batch(vec![rewrite]).unwrap(); + let mut reopened = registered(); + reopened.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + reopened.pull_all_backing_stores().unwrap(); + assert_eq!(reopened.snapshot_envelopes().iter().find(|held| held.key == "marked"), Some(&held)); +} + +#[test] +fn is_writable_agrees_with_the_writes_for_foreign_keys_and_other_types() { + let dir = tempfile::tempdir().unwrap(); + // Only UnionSurface registered: the store's unit-arm record is foreign here. + let path = dir.path().join("extra-slot-store.cc"); + std::fs::write(&path, without_variant_row(&vector("extra-slot-store.cc"))).unwrap(); + let mut cache = CultCache::new(); + cache.register_entry_type::().unwrap(); + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + cache.pull_all_backing_stores().unwrap(); + assert!(cache.put("unit-arm", &constructed()).is_err()); + assert!(!cache.is_writable(UnionSurface::TYPE, "unit-arm").unwrap()); + assert!(cache.is_writable(UnionSurface::TYPE, "unheld").unwrap()); + + // Another registered type at a key a record of this one holds. + let (mut cache, _dir, _) = held_marked_and_clean(); + let other = SlotUnit { name: "clean".into(), outcome: UnitOutcome::Denied }; + assert!(cache.put("clean", &other).is_err()); + assert!(!cache.is_writable(SlotUnit::TYPE, "clean").unwrap()); + assert!(!cache.is_writable(SlotUnit::TYPE, "marked").unwrap()); + assert!(cache.is_writable(UnionSurface::TYPE, "clean").unwrap()); +} + +#[test] +fn a_marked_key_refuses_a_batch_wherever_it_sits() { + let (mut cache, _dir, path) = held_marked_and_clean(); + let before = std::fs::read(&path).unwrap(); + let marked = envelope("marked", rmp_serde::to_vec(&constructed()).unwrap()); + let clean = |key: &str| envelope(key, rmp_serde::to_vec(&constructed()).unwrap()); + for batch in [ + vec![marked.clone(), clean("a"), clean("b")], + vec![clean("a"), marked.clone(), clean("b")], + vec![clean("a"), clean("b"), marked.clone()], + ] { + let error = cache.put_prepared_batch(batch).unwrap_err(); + assert_refused_with(&error, "marked", &mark("Composition", 3, 2, 1)); + assert_eq!(std::fs::read(&path).unwrap(), before); + assert!(cache.snapshot_envelopes().iter().all(|held| held.key != "a" && held.key != "b")); + } +} + +#[test] +fn a_marked_decode_leaves_no_scope_open() { + assert_eq!(cultcache_rs::__enter_slots(), 0); + decode_payload_marked::(&vector("extra-slot-record.msgpack")).unwrap(); + assert_eq!(cultcache_rs::__enter_slots(), 0, "a marked decode left its scope open on this thread"); + let _ = decode_payload_marked::(&vector("depth-skipped-past-limit.msgpack")); + assert_eq!(cultcache_rs::__enter_slots(), 0); +} + +#[test] +fn is_writable_is_false_for_a_type_with_no_home_store() { + let dir = tempfile::tempdir().unwrap(); + let mut cache = registered(); + // One store, routed to SlotUnit only: UnionSurface has no home. + cache.add_backing_store(SingleFileMessagePackBackingStore::new(dir.path().join("s.cc")), [SlotUnit::TYPE]).unwrap(); + assert!(cache.put("k", &constructed()).is_err()); + assert!(!cache.is_writable(UnionSurface::TYPE, "k").unwrap()); + assert!(cache.is_writable(SlotUnit::TYPE, "k").unwrap()); +} + +#[test] +fn load_soa_refuses_a_same_length_replacement_of_a_marked_record() { + use cultcache_rs::CultCacheSoaTable; + let (mut cache, _dir, _) = held_marked_and_clean(); + let all = cache.snapshot_envelopes(); + let marked = all.iter().find(|held| held.key == "marked").unwrap(); + // The clean record's bytes padded to the marked record's length: same length, different bytes. + let mut same_length = vec![0xc0; marked.payload.len()]; + same_length[..2].copy_from_slice(&[0x91, 0xa1]); + assert_eq!(same_length.len(), marked.payload.len()); + let mut replaced = all.clone(); + replaced.iter_mut().find(|held| held.key == "marked").unwrap().payload = same_length; + let error = cache.load_soa(CultCacheSoaTable::from_envelopes(replaced)).unwrap_err(); + assert_refused_with(&error, "marked", &mark("Composition", 3, 2, 1)); + assert_eq!(cache.snapshot_envelopes(), all); +} diff --git a/packages/cultcache-rs/tests/union_parity.rs b/packages/cultcache-rs/tests/union_parity.rs index 24fe999a..c51bd3d8 100644 --- a/packages/cultcache-rs/tests/union_parity.rs +++ b/packages/cultcache-rs/tests/union_parity.rs @@ -1,244 +1,17 @@ //! The Rust runtime held to tests/vectors/union-parity, written by the C# reference. //! Bytes and outcomes are read from the committed files; only the constructed value is spelled here. -use std::path::PathBuf; - use cultcache_rs::CultCache; use cultcache_rs::CacheBackingStore; use cultcache_rs::CultCacheEnvelope; use cultcache_rs::MAX_PAYLOAD_DEPTH; use cultcache_rs::PayloadTooDeep; use cultcache_rs::check_payload_depth; -use cultcache_rs::CultSlots; -use cultcache_rs::CultUnion; use cultcache_rs::DatabaseEntry; use cultcache_rs::SingleFileMessagePackBackingStore; -#[derive(Clone, Debug, PartialEq, CultUnion)] -enum Node { - #[cultcache(key = 0)] - Box(BoxArm), - #[cultcache(key = 1)] - Text(TextArm), - #[cultcache(key = 2)] - Use(UseArm), - #[cultcache(key = 5)] - Image(ImageArm), - #[cultcache(key = 6)] - Wrap(WrapArm), -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct WrapArm { - #[cultcache(key = 0)] - inner: Option>, - #[cultcache(key = 1)] - id: String, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct BoxArm { - #[cultcache(key = 0)] - id: String, - #[cultcache(key = 1)] - children: Vec, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct TextArm { - #[cultcache(key = 0)] - id: String, - #[cultcache(key = 1)] - content: String, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct UseArm { - #[cultcache(key = 0)] - id: String, - #[cultcache(key = 1)] - composition: Composition, - #[cultcache(key = 2)] - slots: Vec>, -} - -// Slot 2 is retired: a gap written nil. -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct ImageArm { - #[cultcache(key = 0)] - id: String, - #[cultcache(key = 1)] - uri: String, - #[cultcache(key = 3)] - alt: String, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct Composition { - #[cultcache(key = 0)] - name: String, - #[cultcache(key = 1)] - body: Option>, -} - -#[derive(Clone, Debug, PartialEq, CultUnion)] -enum Outcome { - #[cultcache(key = 0)] - Pending(PendingArm), - #[cultcache(key = 1)] - Completed(CompletedArm), - #[cultcache(key = 2)] - Denied(IdArm), - #[cultcache(key = -2147483648)] - Floor(IdArm), - #[cultcache(key = 2147483647)] - Ceiling(IdArm), -} - -// The codec's unit arm (no members, written [key, []]), kept apart from the golden: a document list's arms carry ids. -#[derive(Clone, Debug, PartialEq, CultUnion)] -enum UnitOutcome { - #[cultcache(key = 2)] - Denied, - #[cultcache(key = -2147483648)] - Floor, - #[cultcache(key = 2147483647)] - Ceiling, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct PendingArm { - #[cultcache(key = 0)] - message: String, - #[cultcache(key = 1)] - id: String, -} - -// An arm with no members of its own: just its element id. -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct IdArm { - #[cultcache(key = 0)] - id: String, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct CompletedArm { - #[cultcache(key = 0)] - source_version: i32, - #[cultcache(key = 1)] - id: String, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct NamedArm { - #[cultcache(key = 0)] - name: String, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct UsesArm { - #[cultcache(key = 0)] - name: String, - #[cultcache(key = 1)] - uses: i32, -} - -#[derive(Clone, Debug, PartialEq, CultSlots)] -struct CountArm { - #[cultcache(key = 0)] - name: String, - #[cultcache(key = 1)] - count: i32, -} - -#[derive(Clone, Debug, PartialEq, CultUnion)] -enum Item { - #[cultcache(key = 0)] - Gem(NamedArm), - #[cultcache(key = 1)] - Tool(UsesArm), - #[cultcache(key = 2)] - Part(CountArm), -} - -// The same arms as Item under its own keys: key 2 is unknown to it. -#[derive(Clone, Debug, PartialEq, CultUnion)] -enum CraftedItem { - #[cultcache(key = 0)] - Tool(UsesArm), - #[cultcache(key = 1)] - Part(CountArm), -} - -#[derive(Clone, Debug, PartialEq, DatabaseEntry)] -#[cultcache(type = "tests.union_surface", schema = "tests.union_surface.v1")] -struct UnionSurface { - #[cultcache(key = 0)] - id: String, - #[cultcache(key = 1)] - version: i32, - #[cultcache(key = 2)] - root: Option, - #[cultcache(key = 3)] - outcomes: Vec, - #[cultcache(key = 4)] - part: Option, - #[cultcache(key = 5)] - crafted: Option, - #[cultcache(key = 6)] - missing: Option, -} - -fn text(id: &str, content: &str) -> Node { - Node::Text(TextArm { id: id.into(), content: content.into() }) -} - -fn image(id: &str, uri: &str, alt: &str) -> Node { - Node::Image(ImageArm { id: id.into(), uri: uri.into(), alt: alt.into() }) -} - -fn boxed(id: &str, children: Vec) -> Node { - Node::Box(BoxArm { id: id.into(), children }) -} - -fn constructed() -> UnionSurface { - let inner = boxed("0000000000d0", vec![text("0000000000d1", "x"), image("0000000000d2", "uri://a", "alt")]); - UnionSurface { - id: "surface".into(), - version: 3, - root: Some(boxed( - "0000000000c0", - vec![ - text("0000000000c1", "hello"), - Node::Use(UseArm { - id: "0000000000c2".into(), - composition: Composition { name: "comp".into(), body: Some(std::boxed::Box::new(inner)) }, - slots: vec![Some(text("0000000000e1", "slot")), None, Some(boxed("0000000000e2", vec![]))], - }), - image("0000000000c3", "uri://b", ""), - ], - )), - outcomes: vec![ - Outcome::Pending(PendingArm { message: "wait".into(), id: "0000000000a1".into() }), - Outcome::Completed(CompletedArm { source_version: 7, id: "0000000000a2".into() }), - Outcome::Denied(id_arm("0000000000a3")), - ], - part: Some(Item::Tool(UsesArm { name: "hammer".into(), uses: 12 })), - crafted: Some(CraftedItem::Part(CountArm { name: "gear".into(), count: 4 })), - missing: None, - } -} - -fn id_arm(id: &str) -> IdArm { - IdArm { id: id.into() } -} - -fn vector(name: &str) -> Vec { - let path = PathBuf::from(env!("CARGO_MANIFEST_DIR")) - .join("../../tests/vectors/union-parity") - .join(name); - std::fs::read(&path).unwrap_or_else(|error| panic!("{}: {error}", path.display())) -} +mod common; +use common::*; fn decode(bytes: &[u8]) -> Result { rmp_serde::from_slice(bytes).map_err(|error| error.to_string()) @@ -325,11 +98,10 @@ fn an_unknown_arm_lists_the_declared_keys_and_never_echoes_the_payload() { #[test] fn a_union_that_is_not_a_key_and_slots_pair_is_refused_naming_the_enum() { - let cases: [(&str, Vec); 4] = [ + let cases: [(&str, Vec); 3] = [ ("an empty array", vec![0x90]), ("a trailing element", rmp_serde::to_vec(&(1u64, (7u8,), 0u8)).unwrap()), ("a string key", rmp_serde::to_vec(&("CANARY-key", (7u8,))).unwrap()), - ("a unit arm with slots", rmp_serde::to_vec(&(2u64, ("CANARY-slot",))).unwrap()), ]; for (label, bytes) in cases { let message = rmp_serde::from_slice::(&bytes).unwrap_err().to_string(); diff --git a/packages/cultmesh-rs/src/lib.rs b/packages/cultmesh-rs/src/lib.rs index 33cb4858..18a02fcb 100644 --- a/packages/cultmesh-rs/src/lib.rs +++ b/packages/cultmesh-rs/src/lib.rs @@ -732,6 +732,18 @@ impl CultMeshNode { let cultnet_rs::CultNetMessage::SnapshotResponseRaw { documents, .. } = response else { anyhow::bail!("expected cultnet.snapshot_response_raw.v0"); }; + self.apply_raw_snapshot_documents(documents) + } + + /// Applies each snapshot document of a registered schema with `put_raw_envelope`. The first refusal ends the pull + /// and the documents before it stay applied; a key this node holds read-only (unread slots from a newer peer, or a + /// payload this build cannot read) is refused like any other. C# applies a raw snapshot as one cache commit, so a + /// refusal there applies nothing: this is a divergence, owed by follow-up mesh-snapshot-atomic (stage the pull through + /// `put_prepared_batch`). + fn apply_raw_snapshot_documents( + &mut self, + documents: Vec, + ) -> Result { let mut applied = 0usize; for document in documents { let Some(document_type) = @@ -1199,6 +1211,47 @@ mod tests { Note => "cultmesh.test.note.v0", }); + fn raw_note(key: &str, payload: Vec) -> cultnet_rs::CultNetRawDocumentRecord { + cultnet_rs::CultNetRawDocumentRecord { + schema_id: "cultmesh.test.note.v0".to_string(), + record_key: key.to_string(), + stored_at: "2026-01-01T00:00:00Z".to_string(), + payload_encoding: cultnet_rs::CultNetRawPayloadEncoding::Messagepack, + payload, + source_runtime_id: None, + source_agent_id: None, + source_role: None, + tags: None, + } + } + + #[test] + fn a_snapshot_document_for_a_read_only_key_ends_the_pull_unlike_csharp_which_applies_nothing() -> Result<()> { + let temp = tempfile::tempdir()?; + let mut node = CultMesh::create_node(temp.path().join("m.cc"), TestDocuments, Default::default())?; + // A newer peer's note: one slot past the declared one. + let newer = rmp_serde::to_vec(&("newer body", "extra slot"))?; + let plain = |body: &str| rmp_serde::to_vec(&(body,)); + assert_eq!(node.apply_raw_snapshot_documents(vec![raw_note("held", newer)])?, 1); + assert_eq!(node.get_required::("held")?.body, "newer body"); + + let error = node + .apply_raw_snapshot_documents(vec![ + raw_note("before", plain("b")?), + raw_note("held", plain("overwrite")?), + raw_note("after", plain("a")?), + ]) + .unwrap_err(); + let conflict = error.downcast_ref::().expect("typed refusal"); + assert_eq!(conflict.record_key, "held"); + assert!(conflict.unread_slots.is_some()); + assert!(!error.to_string().contains("extra slot")); + assert_eq!(node.get_required::("before")?.body, "b"); + assert_eq!(node.get_required::("held")?.body, "newer body"); + assert!(node.get::("after")?.is_none()); + Ok(()) + } + #[test] fn node_round_trips_registered_documents() -> Result<()> { let temp = tempfile::tempdir()?; diff --git a/packages/cultnet-rs/src/replication.rs b/packages/cultnet-rs/src/replication.rs index 743a28fd..7816e7dd 100644 --- a/packages/cultnet-rs/src/replication.rs +++ b/packages/cultnet-rs/src/replication.rs @@ -146,6 +146,7 @@ impl CultNetDocumentRegistry { schema_id: binding.schema_id.clone(), schema_names: vec![holder.document_type.clone(), binding.document_type.clone()], record_key: String::new(), + unread_slots: None, })); } self.bindings_by_schema_id From 676f611c1fe6c057b59e718e592e201aca2b4906 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 05:03:29 +0000 Subject: [PATCH 04/12] TypeScript: a record with unread slots is held read-only (slot-readonly-ts) (#128) --- packages/cultcache-ts/CHANGELOG.md | 1 + packages/cultcache-ts/README.md | 7 +- packages/cultcache-ts/src/browser.ts | 1 + packages/cultcache-ts/src/codec.ts | 70 ++- packages/cultcache-ts/src/cult-cache.ts | 179 +++++-- .../cultcache-ts/src/document-formatter.ts | 9 +- packages/cultcache-ts/src/document.ts | 17 +- packages/cultcache-ts/src/store-format.ts | 7 +- packages/cultcache-ts/src/types.ts | 20 + .../cultcache-ts/test/slot-readonly.test.ts | 461 ++++++++++++++++++ .../cultcache-ts/test/union-parity.test.ts | 202 +------- .../test/union-surface-fixture.ts | 240 +++++++++ 12 files changed, 976 insertions(+), 238 deletions(-) create mode 100644 packages/cultcache-ts/test/slot-readonly.test.ts create mode 100644 packages/cultcache-ts/test/union-surface-fixture.ts diff --git a/packages/cultcache-ts/CHANGELOG.md b/packages/cultcache-ts/CHANGELOG.md index e71821c8..b91a0535 100644 --- a/packages/cultcache-ts/CHANGELOG.md +++ b/packages/cultcache-ts/CHANGELOG.md @@ -20,6 +20,7 @@ All notable changes to `@gamecult/cultcache-ts` are recorded here, following - A store whose canonical-schema catalog slot is not a string is refused as `StoreUnreadableError`. - The cache holds one record per key (variants:ruling:cache-refuses-retype). A `put` or `putEnvelope` that lands a second type under a key the cache holds, or under a key a store carries foreign, is refused with `SchemaConflictError` naming the key, even with no store (a memory-only cache used to hold both types at the key). A type that owns the held record's schema id (as for the store) replaces it, unless the held record is in a store other than the written type's home: that rename is refused with `SchemaConflictError` (remove the record first). Two stores holding one key, registered or foreign, refuse the load and admit nothing. A store's `delete` removes the record whose stored schema id the removal's type owns, whatever the arrived record's `type` string. - A store holding a record no registered type claims now loads (it used to throw "No schema is registered..."): the cache does not serve the record, lists it in `foreignRecords()`, keeps its bytes through every write, and refuses a `put` or a `delete` at its key with `SchemaConflictError` (variants:ruling:unheld-removal-refused-at-commit). Registering a type that claims its schema id serves it at the next `pullAllBackingStores`. +- A record whose payload carries a slot past its reader's declared slots, at any depth (the record, a nested record, a list element, a union arm, a unit arm), is now held read-only: it loads and reads, and every `put`, `putEnvelope`, `update` or `delete` of its key is refused with `SchemaConflictError` before a store is touched (cultlib-gaps:ruling:extra-trailing-slots, as the C# reference). A load used to skip the extra slots silently and a later write shed them. The refusal names the schema, the key, the type that was read with extra slots and both slot counts, never a slot value, and carries them as `SchemaConflictError.unreadSlots`. An untouched record keeps its bytes through writes of other keys. Missing trailing slots keep their declared defaults and mark nothing. A `put` never marks (the value is encoded by this runtime); a `putEnvelope` of a payload carrying extra slots lands its bytes and holds the record read-only. A mark at depth 0 is consented to only when the record arrived under a schema id other than the reader's own that the reader lists compatible, and that id's catalog entry lists those slots (a writer that adds a slot and keeps the reader's id is never consented to) (`unreadSlots.explained`); every other mark refuses whatever is declared. New: `CultCache.isWritable(definition, key)` (live, advisory), `CultCache.unreadSlots(definition, key)` and the exported `CultCacheUnreadSlots`. Documents with no slot codec (a plain map, a user formatter) never mark. A `delete` is by key: a type that owns the schema id of the record held at the key as another type removes that one record (one store call, the cache dropped once the store has removed it), and a type that does not own it is refused with `SchemaConflictError`, as a `put` is; `isWritable` answers both. Implied semver: a minor bump (0.16.0) while the package is 0.x; a major one at 1.0 and after. ### Added diff --git a/packages/cultcache-ts/README.md b/packages/cultcache-ts/README.md index 2096d258..6fea306a 100644 --- a/packages/cultcache-ts/README.md +++ b/packages/cultcache-ts/README.md @@ -371,8 +371,11 @@ const sceneDocument = defineDocumentType({ }); ``` -- A slot is `[name, codec, default?]`. A gap is written nil and skipped on read; an extra trailing slot is skipped; a - slot missing from the payload (or nil, when a default is declared) takes its default, and is refused without one. +- A slot is `[name, codec, default?]`. A gap is written nil and skipped on read; an extra trailing slot is read past and + marks the record read-only (every write or removal of its key is refused with `SchemaConflictError`, `cache.isWritable` + asks first). The one writable case is a depth-0 extra slot of a record that arrived under a schema id other than the reader's own, which the + reader lists in `compatibleSchemaIds`, whose catalog entry lists those slots: declaring that id consents to the loss, and a write restamps + the record. A nested slot, or a writer that keeps the reader's own id, is never consented to. A slot missing from the payload (or nil, when a default is declared) takes its default, and is refused without one. - A union refuses an undeclared arm key on read, naming the union, the key and the declared keys. Nothing falls back to a default arm. An arm with no slots is `codecs.unit` and is written `[key, []]`. - A refusal names the union, the field and the slot; it never echoes the value it refused. diff --git a/packages/cultcache-ts/src/browser.ts b/packages/cultcache-ts/src/browser.ts index 668fc0b8..98995acb 100644 --- a/packages/cultcache-ts/src/browser.ts +++ b/packages/cultcache-ts/src/browser.ts @@ -34,6 +34,7 @@ export type { CultCacheIndexScalar, InferCultCacheSchemaValue, CultCacheStoreRegistration, + CultCacheUnreadSlots, PushAllOptions, } from "./types"; export type { diff --git a/packages/cultcache-ts/src/codec.ts b/packages/cultcache-ts/src/codec.ts index 9b885159..a2dd7b15 100644 --- a/packages/cultcache-ts/src/codec.ts +++ b/packages/cultcache-ts/src/codec.ts @@ -198,6 +198,66 @@ type SlotSpec = type SlotsSpec = Readonly }[keyof TValue & string]>>; +// The slots a payload decode read past, noted by `slots` and the unit arm while the one decode entry (decodeNoting) has a scope +// open, and read out by that entry. Module-scoped as the reference's is thread-bound: a decode is synchronous, and a decode +// outside the entry (a codec called directly, a CultMesh read) notes nothing. No codec signature carries it. +export interface NotedSlots { + /** The first type read with extra slots, outermost first. */ + readonly typeName: string; + readonly slotCount: number; + readonly declaredCount: number; + readonly occurrences: number; + /** True when every noted object was the document's own: a union adds no depth, each record or arm inside a field adds one. */ + readonly depth0Only: boolean; +} + +interface SlotScope { + depth: number; + found?: { typeName: string; slotCount: number; declaredCount: number; occurrences: number; depth0Only: boolean }; +} + +let scope: SlotScope | undefined; + +/** The one decode entry: runs the decode with a scope open and returns its value with what it noted. */ +export function decodeNoting(decode: () => T): { value: T; unreadSlots?: NotedSlots } { + const outer = scope; + const mine: SlotScope = { depth: 0 }; + scope = mine; + try { + const value = decode(); + return mine.found ? { value, unreadSlots: mine.found } : { value }; + } finally { + scope = outer; + } +} + +function noteExtraSlots(typeName: string, slotCount: number, declaredCount: number): void { + if (!scope) return; + const found = scope.found; + if (found) { + found.occurrences += 1; + found.depth0Only &&= scope.depth === 0; + } else { + scope.found = { typeName, slotCount, declaredCount, occurrences: 1, depth0Only: scope.depth === 0 }; + } +} + +// What a record inside a field reads at: one deeper than the record holding the field. +function nested(read: () => T): T { + if (!scope) return read(); + scope.depth += 1; + try { + return read(); + } finally { + scope.depth -= 1; + } +} + +// The reference's type name for a mark: the type, not its namespace or the type it is nested in. +function simpleTypeName(typeName: string | undefined): string { + return typeName === undefined ? "record" : typeName.slice(Math.max(typeName.lastIndexOf("."), typeName.lastIndexOf("+")) + 1); +} + const INT32_MIN = -(2 ** 31); const INT32_MAX = 2 ** 31 - 1; @@ -232,7 +292,7 @@ function pascalCase(name: string): string { } // A positional record: slot `key` holds field `name`. Gaps are written nil and skipped on read, extra trailing slots -// are skipped, and a slot missing from the payload takes its declared default or is refused. A default fills only a +// are read past and noted (decodeNoting): the record that carried them is read-only, never rewritten without them, and a slot missing from the payload takes its declared default or is refused. A default fills only a // missing trailing slot, never a nil: as in the reference, a nil in a reference slot (string, bytes, record, union, // list) reads as null and writes back as nil, so a record the reference wrote re-encodes byte-identical, and a nil in // a value slot (int, float, bool) is refused unless the slot is nullable. @@ -281,6 +341,9 @@ export function slots(spec: SlotsSpec, typeName?: decode(raw) { expect(raw, "an array of slots", Array.isArray(raw)); const input = raw as unknown[]; + if (input.length > length) { + noteExtraSlots(simpleTypeName(typeName), input.length, length); + } const out: Record = {}; for (const field of fields) { const present = field.key < input.length; @@ -291,7 +354,7 @@ export function slots(spec: SlotsSpec, typeName?: ? defaultOf(field) : input[field.key] === null && !field.codec.valueType ? null - : at(`${field.name} (slot ${field.key})`, () => field.codec.decode(input[field.key])); + : at(`${field.name} (slot ${field.key})`, () => nested(() => field.codec.decode(input[field.key]))); } return out as TValue; }, @@ -363,6 +426,9 @@ export function union( if (arm === undefined) { throw new Error(`Union ${name}: unknown arm key ${key} (declared keys: ${declared}).`); } + if (arm.body === unit && raw[1].length > 0) { + noteExtraSlots(arm.armName, raw[1].length, 0); + } const fields = arm.body === unit ? {} : at(`arm ${arm.armName}`, () => (arm.body as SlotsCodec).decode(raw[1])); return { kind: arm.armName, ...fields } as unknown as TValue; }, diff --git a/packages/cultcache-ts/src/cult-cache.ts b/packages/cultcache-ts/src/cult-cache.ts index 443c0c18..6cf7a8cb 100644 --- a/packages/cultcache-ts/src/cult-cache.ts +++ b/packages/cultcache-ts/src/cult-cache.ts @@ -1,6 +1,6 @@ import { declaresRole, - decodeDocumentPayload, + decodeDocumentPayloadMarked, documentPayloadBytes, documentPayloadFromBytes, encodeDocumentPayload, @@ -8,6 +8,7 @@ import { schemaIdentityOf, } from "./document"; import { resolveDocumentFormatter } from "./document-formatter"; +import type { NotedSlots } from "./codec"; import type { DocumentPayload } from "./document"; import { SchemaConflictError, schemaIdForEnvelope, writtenOwns } from "./store-format"; import type { @@ -20,6 +21,7 @@ import type { CultCacheEnvelope, CultCacheSchemaCatalogEntry, CultCacheStoreRegistration, + CultCacheUnreadSlots, } from "./types"; type Accessor = (value: unknown) => string | undefined; @@ -27,7 +29,7 @@ type Accessor = (value: unknown) => string | undefined; // A registered definition reads and writes its documents through the owner, on payloads the cache has cloned. type RegisteredFormatter = { encode(value: unknown): DocumentPayload; - decode(payload: DocumentPayload): unknown; + decode(payload: DocumentPayload): { value: unknown; unreadSlots?: NotedSlots }; }; type RegisteredDefinition = { @@ -50,6 +52,10 @@ type HydratedEntry = CultCacheEnvelope & { catalogEntry: NonNullable; value: unknown; lookups: EntryLookups; + // What the payload decode read past: set when the record was admitted, and it makes the record read-only unless explained. + unreadSlots?: CultCacheUnreadSlots; + // The schema id the record arrived under, before it was admitted under its reader's: what an unread-slots refusal names. + storedSchemaId?: string; }; // Everything a pull replaces. Built whole, then swapped in with one assignment. @@ -84,6 +90,20 @@ type DerivedTypeLookups = { readonly indexMaps: Map>; }; +// The consent a compatible-id declaration gives (cultlib-gaps:ruling:slot-mark-consent): a mark is explained only when the record +// arrived under an id other than the reader's own that the reader lists compatible, it is at +// depth 0 and the schema comparison of the catalog entry the record arrived with against the reader's own reported the slots it +// carries as ignored, so its slot count is at most the highest ignored slot plus one. A record written or put through the cache +// arrives under the reader's own entry, which ignores nothing, so no mark of it is explained. +function explainedBy(noted: NotedSlots, reader: CultCacheSchemaCatalogEntry, arrivedId: string, arrived: CultCacheSchemaCatalogEntry | undefined): boolean { + if (arrivedId === reader.schemaId || !(reader.compatibleSchemaIds ?? []).includes(arrivedId)) { + return false; + } + const declared = new Set((reader.members ?? []).map((member) => member.slot)); + const ignored = (arrived?.members ?? []).map((member) => member.slot).filter((slot) => !declared.has(slot)); + return noted.depth0Only && noted.slotCount <= Math.max(-1, ...ignored) + 1; +} + function emptyHydratedState(): HydratedState { return { entries: new Map(), @@ -312,7 +332,7 @@ export class CultCache { } const payload = this.#cloneBytes(entry.payload); - const value = registered.formatter.decode(payload); + const { value, unreadSlots } = registered.formatter.decode(payload); let admitted: CultCacheEnvelope = { ...entry, type, payload }; if (registered.global) { const persistedKey = persistedGlobalKeys.get(type); @@ -327,7 +347,7 @@ export class CultCache { admitted = { ...admitted, key: CultCache.GLOBAL_KEY }; } } - const hydrated = this.#admitEntry(next, registered, admitted, value); + const hydrated = this.#admitEntry(next, registered, admitted, value, unreadSlots); const other = this.#heldUnderOtherType(next, type, hydrated.key); if (other !== undefined) { throw new Error(`Record ${hydrated.key} is held by two stores, as type "${other.type}" and as type "${type}".`); @@ -518,10 +538,9 @@ export class CultCache { if (!envelope.storedAt || envelope.storedAt.trim().length === 0) { throw new Error(`CultCache envelope storedAt for type "${definition.type}" must be non-empty.`); } - this.#requireGlobalKey(registered, envelope.key); const payload = this.#cloneBytes(envelope.payload); - const parsed = registered.formatter.decode(payload) as CultCacheDocumentValue; + const { value: parsed, unreadSlots } = registered.formatter.decode(payload) as { value: CultCacheDocumentValue; unreadSlots?: NotedSlots }; const entry: CultCacheEnvelope = { key: envelope.key, type: envelope.type, @@ -532,7 +551,7 @@ export class CultCache { schemaId: registered.catalogEntry.schemaId, catalogEntry: registered.catalogEntry, }; - await this.#writeNow(registered, entry, parsed); + await this.#writeNow(registered, entry, parsed, unreadSlots); return parsed; }); } @@ -585,6 +604,29 @@ export class CultCache { }); } + /** What the record's payload carried past this build's declared slots, or undefined: the held record is read-only unless `explained`. */ + unreadSlots(definition: AnyCultCacheDocumentDefinition, key: string): CultCacheUnreadSlots | undefined { + return this.#hydrated.entries.get(this.#entryId(this.#requireDefinition(definition).definition.type, key))?.unreadSlots; + } + + /** + * Whether a write of the record's own type at the key would be let through: it asks the judge #writeNow asks, of the record the + * cache holds (a key a store carries foreign, a record of another type the write does not own or cannot replace, a global at + * another key, a record carrying unread slots nothing consents to). It answers the WRITE form. A removal asks the judge's + * removal form and differs in two states: an owner's `delete` of a record held in another store than its own home succeeds + * where this is false (a write would push to the wrong store), and `delete` of a global type at a key other than __global__ is + * the plain false of an unheld key where this is false. Live and advisory: the write or removal still decides. + */ + isWritable(definition: AnyCultCacheDocumentDefinition, key: string): boolean { + const registered = this.#requireDefinition(definition); + try { + this.#judge(registered, key, this.#probe(registered, key)); + return true; + } catch { + return false; + } + } + // What the last pull found that no registered type claims, by key ordinal. A later registration claims a record at the next pull. foreignRecords(): readonly CultCacheForeignRecord[] { return [...this.#hydrated.foreign].sort((left, right) => compareOrdinal(left.key, right.key)); @@ -680,9 +722,8 @@ export class CultCache { value: CultCacheDocumentValue, ): Promise> { const registered = this.#requireDefinition(definition); - this.#requireGlobalKey(registered, key); const payload = registered.formatter.encode(value); - const parsed = registered.formatter.decode(payload) as CultCacheDocumentValue; + const { value: parsed, unreadSlots } = registered.formatter.decode(payload) as { value: CultCacheDocumentValue; unreadSlots?: NotedSlots }; const entry: CultCacheEnvelope = { key, type: definition.type, @@ -691,26 +732,16 @@ export class CultCache { schemaId: registered.catalogEntry.schemaId, catalogEntry: registered.catalogEntry, }; - await this.#writeNow(registered, entry, parsed); + await this.#writeNow(registered, entry, parsed, unreadSlots); return parsed; } // Validates the whole write (home, global singleton, name and index accessors) before the store // is touched; once the store accepts, applying to the cache has no fallible step. - async #writeNow(registered: RegisteredDefinition, entry: CultCacheEnvelope, value: unknown): Promise { + async #writeNow(registered: RegisteredDefinition, entry: CultCacheEnvelope, value: unknown, unreadSlots?: NotedSlots): Promise { const home = this.#homeStore(entry.type); - const retyped = this.#refuseRetype(entry); - if (retyped !== undefined && this.#homeStore(retyped.type) !== home) { - // The write owns the held record's id and would replace it, but the held record is in another store; the push - // reaches only this type's home, so the old record would stay in its own file and load as a second holder. - throw new SchemaConflictError( - schemaIdForEnvelope(entry), - [retyped.type, entry.type], - entry.key, - `CultCache holds record "${entry.key}" as type "${retyped.type}" in another store than the home of type "${entry.type}", so the rename cannot replace it; remove the record first.`, - ); - } - const hydrated = this.#admitEntry(this.#hydrated, registered, entry, value); + const retyped = this.#judge(registered, entry.key, entry); + const hydrated = this.#admitEntry(this.#hydrated, registered, entry, value, unreadSlots); this.#refuseIndexSharing(this.#hydrated, entry.type, registered.catalogEntry.schemaName, entry.key, hydrated.lookups); const legacy = this.#hydrated.legacyGlobals.get(entry.type); await home?.push(entry); @@ -725,6 +756,85 @@ export class CultCache { this.#installEntry(this.#hydrated, registered, hydrated); } + // An envelope of the definition's type at the key, carrying nothing: what the judge asks about a write or a removal by that type. + #probe(registered: RegisteredDefinition, key: string): CultCacheEnvelope { + return { + key, type: registered.definition.type, payload: documentPayloadFromBytes(new Uint8Array(0)), storedAt: "", + schemaId: registered.catalogEntry.schemaId, catalogEntry: registered.catalogEntry, + }; + } + + // The one judge of whether a key may be written (a written envelope) or removed (none), before any store is touched: it throws the + // refusal, and #writeNow, #deleteNow and isWritable all ask it, so the query cannot disagree with the write. For a write it + // returns the held record of another type that the write owns and so restamps, if any. A refusal names keys and schemas, + // never a payload or a slot value. + #judge(registered: RegisteredDefinition, key: string, written?: CultCacheEnvelope): HydratedEntry | undefined { + const type = registered.definition.type; + const held = this.#hydrated.entries.get(this.#entryId(type, key)); + if (written === undefined) { + // A removal is by key: the record it removes is the one held at the key, of this type or of another type whose schema id + // this type owns (the record a write of this type would replace). A held record of a type it does not own is refused. + this.#refuseForeignKey(key, type, "removed"); + this.#refuseUnread(held); + if (held !== undefined) { + return undefined; + } + const other = this.#heldUnderOtherType(this.#hydrated, type, key); + if (other === undefined) { + return undefined; + } + if (!writtenOwns(this.#probe(registered, key), schemaIdForEnvelope(other))) { + throw new SchemaConflictError( + registered.catalogEntry.schemaId, + [other.type, type], + key, + `CultCache holds record "${key}" as type "${other.type}", so type "${type}" cannot remove it.`, + ); + } + this.#refuseUnread(other); + return other; + } + this.#requireGlobalKey(registered, key); + const retyped = this.#refuseRetype(written); + this.#refuseUnread(held); + this.#refuseUnread(retyped); + if (retyped !== undefined && this.#homeStore(retyped.type) !== this.#homeStore(type)) { + // The write owns the held record's id and would replace it, but the held record is in another store; the push + // reaches only this type's home, so the old record would stay in its own file and load as a second holder. + throw new SchemaConflictError( + schemaIdForEnvelope(written), + [retyped.type, type], + key, + `CultCache holds record "${key}" as type "${retyped.type}" in another store than the home of type "${type}", so the rename cannot replace it; remove the record first.`, + ); + } + return retyped; + } + + // The one refusal of a record held read-only for the slots its payload carries past this build's: a write or a removal of + // its key. It names the type and the counts, never a slot value. A mark that is explained (see #admitEntry) is consented to. + #unreadRefusal(held: HydratedEntry | undefined): SchemaConflictError | undefined { + const unread = held?.unreadSlots; + if (held === undefined || unread === undefined || unread.explained) { + return undefined; + } + return new SchemaConflictError( + held.storedSchemaId ?? held.schemaId ?? held.type, + [held.catalogEntry.schemaName], + held.key, + `${held.catalogEntry.schemaName} record ${held.key} is read-only: ${unread.typeName} was read with ${unread.slotCount} slots where this build declares ${unread.declaredCount} ` + + `(${unread.occurrences} object(s) in the record). Writing it would drop data a newer writer stored; read it with types that declare those slots, or write it from that build.`, + unread, + ); + } + + #refuseUnread(held: HydratedEntry | undefined): void { + const refusal = this.#unreadRefusal(held); + if (refusal !== undefined) { + throw refusal; + } + } + // The one key-identity judge of the cache: may this write land at its key, given what the cache holds there under // every registered type and the keys its stores carry foreign. The store judges the file with the same predicate // (writtenOwns); this judges the cache, which a store-less cache has no other judge for. Returns the held record of @@ -782,18 +892,20 @@ export class CultCache { key: string, ): Promise { const registered = this.#requireDefinition(definition); - const entry = this.#hydrated.entries.get(this.#entryId(definition.type, key)); + // variants:ruling:unheld-removal-refused-at-commit: a key a store carries foreign is refused, a key held by nobody is the + // existing false; a key held as another type is removed by an owner of its id and refused to any other type. + // Removal is by key, in one store call: the record held at the key of this type, else the one of another type this type owns. + const owned = this.#judge(registered, key); + const entry = this.#hydrated.entries.get(this.#entryId(definition.type, key)) ?? owned; if (!entry) { - // variants:ruling:unheld-removal-refused-at-commit: a key a store carries foreign is refused, a key held under - // another type or by nobody is the existing false. - this.#refuseForeignKey(key, definition.type, "removed"); return false; } - const home = this.#homeStore(definition.type); - await home?.delete(this.#hydrated.legacyGlobals.get(definition.type) ?? this.#toEnvelope(entry)); - this.#hydrated.legacyGlobals.delete(definition.type); - this.#removeEntry(this.#hydrated, registered, entry); + // The cache drops the record only once the store has removed it. + const home = this.#homeStore(entry.type); + await home?.delete(this.#hydrated.legacyGlobals.get(entry.type) ?? this.#toEnvelope(entry)); + this.#hydrated.legacyGlobals.delete(entry.type); + this.#removeEntry(this.#hydrated, this.#definitions.get(entry.type)!, entry); return true; } @@ -820,7 +932,7 @@ export class CultCache { return { encode: (value: unknown) => this.#cloneBytes(encodeDocumentPayload(definition, value)), - decode: (payload: DocumentPayload) => decodeDocumentPayload(definition, this.#cloneBytes(payload)), + decode: (payload: DocumentPayload) => decodeDocumentPayloadMarked(definition, this.#cloneBytes(payload)), }; } @@ -1000,6 +1112,7 @@ export class CultCache { registered: RegisteredDefinition, entry: CultCacheEnvelope, value: unknown, + noted?: NotedSlots, ): HydratedEntry { if (registered.global) { const currentGlobalKey = state.globalKeys.get(entry.type); @@ -1020,6 +1133,8 @@ export class CultCache { catalogEntry: registered.catalogEntry, value, lookups, + storedSchemaId: entry.schemaId ?? entry.type, + ...(noted ? { unreadSlots: { ...noted, explained: explainedBy(noted, registered.catalogEntry, entry.schemaId ?? entry.type, entry.catalogEntry) } } : {}), }; } diff --git a/packages/cultcache-ts/src/document-formatter.ts b/packages/cultcache-ts/src/document-formatter.ts index 799b2558..92d45df1 100644 --- a/packages/cultcache-ts/src/document-formatter.ts +++ b/packages/cultcache-ts/src/document-formatter.ts @@ -1,7 +1,7 @@ import { decode, encode } from "@msgpack/msgpack"; import { MAX_PAYLOAD_DEPTH } from "./payload-depth"; -import type { SlotsCodec } from "./codec"; +import { decodeNoting, type NotedSlots, type SlotsCodec } from "./codec"; import type { AnyCultCacheDocumentDefinition, CultCacheDocumentFormatter } from "./types"; import { readWire, writeWire } from "./wire"; @@ -16,6 +16,13 @@ export function resolveDocumentFormatter(definition: AnyCultCacheDocumentDefinit return definition.formatter ?? (definition.codec ? codecFormatter(definition.codec) : plainDocumentFormatter); } +// The one decode of a document's bytes, with the slots it read past. A plain formatter or a user formatter has no slot codec +// and carries everything, so it notes nothing. +export function decodeMarked(definition: AnyCultCacheDocumentDefinition, bytes: Uint8Array): { value: unknown; unreadSlots?: NotedSlots } { + const formatter = resolveDocumentFormatter(definition); + return decodeNoting(() => formatter.decode(bytes)); +} + const plainDocumentFormatter: CultCacheDocumentFormatter = { // The encoder's own depth refusal (100 by default) is set above the limit, so the payload walk decides. encode: (value: unknown) => encode(value, { maxDepth: MAX_PAYLOAD_DEPTH * 2 }), diff --git a/packages/cultcache-ts/src/document.ts b/packages/cultcache-ts/src/document.ts index 10af96a6..8d295bd4 100644 --- a/packages/cultcache-ts/src/document.ts +++ b/packages/cultcache-ts/src/document.ts @@ -1,4 +1,5 @@ -import { resolveDocumentFormatter } from "./document-formatter"; +import type { NotedSlots } from "./codec"; +import { decodeMarked, resolveDocumentFormatter } from "./document-formatter"; import { checkPayloadDepth } from "./payload-depth"; import type { AnyCultCacheDocumentDefinition, @@ -118,10 +119,20 @@ export function decodeDocumentPayload { + return decodeDocumentPayloadMarked(definition, payload).value; +} + +// The same read with the slots it read past (cult-cache-ts slot-readonly). The depth check runs first: a payload past the +// limit is refused, never marked. +export function decodeDocumentPayloadMarked( + definition: TDefinition, + payload: DocumentPayload, +): { value: CultCacheDocumentValue; unreadSlots?: NotedSlots } { const bytes = documentPayloadBytes(payload); if (writesLibraryMessagePack(definition)) checkPayloadDepth(bytes, `document type "${definition.type}" payload`); - const decoded = resolveDocumentFormatter(definition).decode(bytes); - return definition.schema.parse(decoded) as CultCacheDocumentValue; + const { value: decoded, unreadSlots } = decodeMarked(definition, bytes); + const value = definition.schema.parse(decoded) as CultCacheDocumentValue; + return unreadSlots ? { value, unreadSlots } : { value }; } // The one writer: the definition's schema first, then its formatter, codec or plain map. diff --git a/packages/cultcache-ts/src/store-format.ts b/packages/cultcache-ts/src/store-format.ts index acd10207..885b1c8e 100644 --- a/packages/cultcache-ts/src/store-format.ts +++ b/packages/cultcache-ts/src/store-format.ts @@ -1,6 +1,6 @@ import { decode } from "@msgpack/msgpack"; -import type { CultCacheEnvelope } from "./types"; +import type { CultCacheEnvelope, CultCacheUnreadSlots } from "./types"; export const STORE_FORMAT_VERSION = "cultcache.store.v1"; @@ -83,13 +83,16 @@ export class SchemaConflictError extends Error { readonly schemaId: string; readonly schemaNames: string[]; readonly recordKey: string; + /** Set when the refusal is for unread slots; undefined for every other refusal. */ + readonly unreadSlots?: CultCacheUnreadSlots; - constructor(schemaId: string, schemaNames: string[], recordKey: string, message?: string) { + constructor(schemaId: string, schemaNames: string[], recordKey: string, message?: string, unreadSlots?: CultCacheUnreadSlots) { super(message ?? `Schema id "${schemaId}" cannot describe record "${recordKey}": ${schemaNames.map((name) => `"${name}"`).join(", ")}.`); this.name = "SchemaConflictError"; this.schemaId = schemaId; this.schemaNames = schemaNames; this.recordKey = recordKey; + this.unreadSlots = unreadSlots; } } diff --git a/packages/cultcache-ts/src/types.ts b/packages/cultcache-ts/src/types.ts index 6f6c0923..10371053 100644 --- a/packages/cultcache-ts/src/types.ts +++ b/packages/cultcache-ts/src/types.ts @@ -99,6 +99,26 @@ export interface CultCacheEnvelope { catalogEntry?: CultCacheSchemaCatalogEntry; } +/** + * What a payload decode found past the slots this build declares, at any depth: the first type read with extra slots, how many + * slots it carried against how many it declares, and how many objects in the record did. It holds no slot value, so a refusal + * can name it. A record carrying it is read-only unless `explained`. + */ +export interface CultCacheUnreadSlots { + /** The first type read with extra slots (the unit arm's arm name for a unit arm). */ + readonly typeName: string; + readonly slotCount: number; + readonly declaredCount: number; + readonly occurrences: number; + /** True when every mark sat in the document's own object. */ + readonly depth0Only: boolean; + /** + * True only when every mark is at depth 0 and the schema comparison of the id the record was stored under reported every slot + * the mark carries (slotCount is at most the highest reported slot plus one): the loss a compatible-id declaration consents to. + */ + readonly explained: boolean; +} + export interface PushAllOptions { soft?: boolean; } diff --git a/packages/cultcache-ts/test/slot-readonly.test.ts b/packages/cultcache-ts/test/slot-readonly.test.ts new file mode 100644 index 00000000..e3285f07 --- /dev/null +++ b/packages/cultcache-ts/test/slot-readonly.test.ts @@ -0,0 +1,461 @@ +import assert from "node:assert/strict"; +import { copyFile, mkdtemp, readFile, writeFile } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join } from "node:path"; +import { test } from "node:test"; + +import { decode, encode } from "@msgpack/msgpack"; +import { z } from "zod"; + +import { codecs } from "../src"; +import { CultCache } from "../src/cult-cache"; +import { defineDocumentType, documentPayloadBytes, documentPayloadFromBytes } from "../src/document"; +import { SingleFileMessagePackBackingStore, readSingleFileStore } from "../src/single-file-messagepack-backing-store"; +import { SchemaConflictError, StoreUnreadableError } from "../src/store-format"; +import type { CacheBackingStore, CultCacheEnvelope, CultCacheUnreadSlots } from "../src/types"; +import { + type Surface, envelopeOf, golden, hex, memoryStore, read, slotUnitDocument, surfaceCodec, surfaceDeclaringDocument, surfaceDocument, surfaceSchema, vectors, +} from "./union-surface-fixture"; + +// cultlib-gaps:cut:slot-readonly-ts. A payload carrying a slot past its reader's declared slots, at any depth, is read, held +// read-only, and every write or removal of its key is refused before a store is touched. The vectors are written by the C# +// reference: tests/vectors/union-parity/README.md, "Extra trailing slot" and "Slot stores". + +const { slots, string } = codecs; +const SLOT_VALUE = "extra-slot"; + +function mark(typeName: string, slotCount: number, declaredCount: number, depth0Only: boolean, explained = false, occurrences = 1): CultCacheUnreadSlots { + return { typeName, slotCount, declaredCount, occurrences, depth0Only, explained }; +} + +// File, then the mark the README table gives for it. +const PAYLOAD_ROWS: [string, CultCacheUnreadSlots][] = [ + ["extra-slot-record.msgpack", mark("UnionSurface", 8, 7, true)], + ["extra-slot-nested-record.msgpack", mark("Composition", 3, 2, false)], + ["extra-slot-arm.msgpack", mark("Box", 3, 2, false)], + ["extra-slot-list-element.msgpack", mark("Text", 3, 2, false)], + ["depth-skipped-at-limit.msgpack", mark("UnionSurface", 8, 7, true)], +]; + +async function readOnlyFiles(name: string): Promise { + const rows = decode(await read(name)) as unknown[][]; + return rows.filter((row) => row[1] === "read-only").map((row) => row[0] as string); +} + +test("the read-only payload vectors are exactly the ones this walker names: four in outcomes, one in bounds", async () => { + const listed = [...(await readOnlyFiles("outcomes.msgpack")), ...(await readOnlyFiles("bounds-outcomes.msgpack"))]; + assert.deepEqual(listed.sort(), PAYLOAD_ROWS.map(([file]) => file).sort()); +}); + +for (const [file, expected] of PAYLOAD_ROWS) { + test(`${file}: read past the extra slot, held read-only with the mark, every write and removal of the key refused`, async () => { + const document = surfaceDocument(); + const entries = [envelopeOf(await read(file))]; + const before = hex(documentPayloadBytes(entries[0]!.payload)); + const cache = CultCache.builder().withDocumentType(document).withGenericStore(memoryStore(entries)).build(); + await cache.pullAllBackingStores(); + + assert.deepEqual(cache.get(document, "surface"), golden, "the extra slot is read past"); + assert.deepEqual(cache.unreadSlots(document, "surface"), expected); + assert.equal(cache.isWritable(document, "surface"), false); + assert.equal(cache.isWritable(document, "another"), true); + + const refused = (error: unknown) => { + assert.ok(error instanceof SchemaConflictError, "a SchemaConflictError"); + assert.deepEqual(error.unreadSlots, expected); + assert.equal(error.recordKey, "surface"); + assert.ok(error.message.includes(expected.typeName), error.message); + assert.ok(error.message.includes(`read with ${expected.slotCount} slots where this build declares ${expected.declaredCount} (${expected.occurrences} object(s)`), error.message); + assert.ok(!error.message.includes(SLOT_VALUE), "the refusal never holds a slot value"); + return true; + }; + await assert.rejects(() => cache.put(document, "surface", golden), refused); + await assert.rejects(() => cache.update(document, "surface", (value) => ({ ...value!, version: 4 })), refused); + const replacement = envelopeOf(await read("union-surface.msgpack")); + await assert.rejects(() => cache.putEnvelope(document, replacement), refused); + await assert.rejects(() => cache.delete(document, "surface"), refused); + assert.equal(entries.length, 1, "nothing reached the store"); + assert.equal(hex(documentPayloadBytes(entries[0]!.payload)), before, "the held bytes are untouched"); + assert.deepEqual(cache.get(document, "surface"), golden, "the record is still held"); + + await cache.put(document, "another", golden); + assert.equal(entries.length, 2, "another key of the same store writes"); + }); +} + +test("a clean payload and a short one are writable: missing trailing slots keep the defaults and mark nothing", async () => { + const document = surfaceDocument(); + const short = encode([...(decode(await read("union-surface.msgpack")) as unknown[]).slice(0, 6)]); + const entries = [envelopeOf(await read("union-surface.msgpack")), { ...envelopeOf(short), key: "short" }]; + const cache = CultCache.builder().withDocumentType(document).withGenericStore(memoryStore(entries)).build(); + await cache.pullAllBackingStores(); + for (const key of ["surface", "short"]) { + assert.equal(cache.unreadSlots(document, key), undefined); + assert.equal(cache.isWritable(document, key), true); + } + await cache.put(document, "short", golden); +}); + +test("a payload past the nesting limit is refused, never marked", async () => { + const entries = [envelopeOf(await read("depth-skipped-past-limit.msgpack"))]; + const cache = CultCache.builder().withDocumentType(surfaceDocument()).withGenericStore(memoryStore(entries)).build(); + await assert.rejects(() => cache.pullAllBackingStores(), /nesting|depth/iu); +}); + +test("putEnvelope of a payload carrying extra slots lands the bytes and holds the record read-only", async () => { + const document = surfaceDocument(); + const entries: ReturnType[] = []; + const cache = CultCache.builder().withDocumentType(document).withGenericStore(memoryStore(entries)).build(); + const marked = envelopeOf(await read("extra-slot-nested-record.msgpack")); + await cache.putEnvelope(document, marked); + assert.equal(hex(documentPayloadBytes(entries[0]!.payload)), hex(await read("extra-slot-nested-record.msgpack")), "the bytes are carried"); + assert.deepEqual(cache.unreadSlots(document, "surface"), mark("Composition", 3, 2, false)); + assert.equal(cache.isWritable(document, "surface"), false); + await assert.rejects(() => cache.put(document, "surface", golden), SchemaConflictError); +}); + +test("a key a store carries foreign is not writable", async () => { + const document = surfaceDocument(); + const foreign = { ...envelopeOf(await read("union-surface.msgpack")), key: "stranger", type: "tests.unknown", schemaId: "tests.unknown.v1" }; + const cache = CultCache.builder().withDocumentType(document).withGenericStore(memoryStore([foreign])).build(); + await cache.pullAllBackingStores(); + assert.equal(cache.isWritable(document, "stranger"), false); + assert.equal(cache.isWritable(document, "other"), true); +}); + +test("a write by a type that owns the held record's id restamps it, and is refused when that record carries unread slots", async () => { + const typed = (type: string, schemaId: string, compatibleSchemaIds: string[] = []) => + defineDocumentType({ type, schemaId, schemaName: type, schemaVersion: `${type}.v1`, compatibleSchemaIds, schema: surfaceSchema, codec: surfaceCodec, name: "id" }); + const older = typed("tests.surface_older", "tests.surface_older.v1"); + const newer = typed("tests.surface_newer", "tests.surface_newer.v1", ["tests.surface_older.v1"]); + for (const [file, lands] of [["union-surface.msgpack", true], ["extra-slot-record.msgpack", false]] as const) { + const held = { ...envelopeOf(await read(file)), type: older.type, schemaId: "tests.surface_older.v1" }; + const cache = CultCache.builder().withDocumentType(older).withDocumentType(newer).withGenericStore(memoryStore([held])).build(); + await cache.pullAllBackingStores(); + assert.equal(cache.isWritable(newer, "surface"), lands, `${file}: isWritable agrees with the write by the type that owns the id`); + const write = cache.put(newer, "surface", golden); + if (lands) { + await write; + } else { + await assert.rejects(() => write, (error: unknown) => error instanceof SchemaConflictError && error.unreadSlots?.typeName === "UnionSurface"); + } + } +}); + +// The store vectors. TS has no variant records: extra-slot-store.cc carries one (marked-variant), so the whole file is refused +// by name, and the walk runs over the file without that record. +async function copyOf(name: string): Promise { + const path = join(await mkdtemp(join(tmpdir(), "slot-readonly-")), name); + await copyFile(join(vectors, name), path); + return path; +} + +async function withoutVariant(path: string): Promise { + const [format, catalog, records] = decode(await readFile(path)) as [unknown, unknown[], unknown[][]]; + await writeFile(path, encode([format, catalog, records.filter((record) => record[0] !== "marked-variant")])); +} + +type SlotRow = [string, string, boolean, boolean | null, string | null, number | null, number | null, number | null, string | null]; + +async function slotRows(): Promise { + return decode(await read("slot-outcomes.msgpack")) as SlotRow[]; +} + +// The catalog entry the file publishes for a schema: its id and the reference's content hash, which a store rewrites with the +// definition's own. +async function catalogOf(file: string, schemaName: string): Promise<[string, string]> { + const [, catalog] = decode(await read(file)) as [unknown, string[][]]; + const entry = catalog.find((candidate) => candidate[1] === schemaName)!; + return [entry[0]!, entry[3]!]; +} + +async function openStore(file: string, consent: boolean) { + const path = await copyOf(file); + if (file === "extra-slot-store.cc") await withoutVariant(path); + const surface = consent ? surfaceDeclaringDocument() : surfaceDocument(...(await catalogOf(file, "tests.union_surface"))); + const slotUnit = consent ? undefined : slotUnitDocument(...(await catalogOf(file, "tests.slot_unit"))); + const builder = CultCache.builder().withDocumentType(surface).withGenericStore(new SingleFileMessagePackBackingStore(path)); + const cache = (slotUnit ? builder.withDocumentType(slotUnit) : builder).build(); + await cache.pullAllBackingStores(); + return { path, cache, surface, slotUnit }; +} + +test("extra-slot-store.cc carries a variant, which this runtime refuses by name; the rest of the file is walked", async () => { + const path = await copyOf("extra-slot-store.cc"); + assert.throws(() => readSingleFileStore(path), StoreUnreadableError); +}); + +test("every slot-outcomes row holds: the mark, IsWritable, and the removal refused exactly where the record is not writable", async () => { + const rows = await slotRows(); + assert.equal(rows.length, 8); + let walked = 0; + for (const [file, key, writable, explained, typeName, slots, declared, occurrences, flattenRefusedBy] of rows) { + if (flattenRefusedBy !== null) continue; // a variant: this runtime holds none + walked += 1; + const { path, cache, surface, slotUnit } = await openStore(file!, file === "extra-slot-consent.cc"); + const document = key === "unit-arm" ? slotUnit! : surface; + const label = `${file} ${key}`; + const found = cache.unreadSlots(document, key!); + if (typeName === null) { + assert.equal(found, undefined, label); + } else { + assert.ok(found, label); + assert.deepEqual([found.explained, found.typeName, found.slotCount, found.declaredCount, found.occurrences], [explained, typeName, slots, declared, occurrences], label); + } + assert.equal(cache.isWritable(document, key!), writable, label); + + const original = await readFile(path); + if (writable) { + assert.equal(await cache.delete(document, key!), true, label); + } else { + await assert.rejects( + () => cache.delete(document, key!), + (error: unknown) => error instanceof SchemaConflictError && error.unreadSlots?.typeName === typeName && !error.message.includes(SLOT_VALUE), + label, + ); + assert.equal(Buffer.compare(original, await readFile(path)), 0, `${label}: the file is unchanged`); + } + } + assert.equal(walked, 7); +}); + +test("consent: a write of the explained record lands; the over-consent write is refused; the nested mark is refused", async () => { + const consented = await openStore("extra-slot-consent.cc", true); + const value = consented.cache.get(consented.surface, "consented") as Surface; + assert.deepEqual(value, golden); + await consented.cache.put(consented.surface, "consented", value); + assert.equal(consented.cache.unreadSlots(consented.surface, "consented"), undefined, "the write dropped the explained slot and the mark with it"); + assert.equal(readSingleFileStore(consented.path).envelopes.find((e) => e.key === "consented")!.schemaId, "tests.union_surface.v2"); + + for (const key of ["over-consent", "nested"]) { + const refused = await openStore("extra-slot-consent.cc", true); + await assert.rejects( + () => refused.cache.put(refused.surface, key, golden), + (error: unknown) => error instanceof SchemaConflictError && error.schemaId === "tests.union_surface.wide", + `${key}: the refusal names the id the record was stored under`, + ); + } +}); + +test("carried: after writing another record, the marked record's bytes, schema id, stored-at and every catalog entry are the originals", async () => { + const { path, cache, surface } = await openStore("extra-slot-store.cc", false); + const originals = readSingleFileStore(path); + await cache.put(surface, "clean", golden); + const after = readSingleFileStore(path); + for (const key of ["marked", "unit-arm"]) { + const was = originals.envelopes.find((e) => e.key === key)!; + const now = after.envelopes.find((e) => e.key === key)!; + assert.equal(hex(documentPayloadBytes(now.payload)), hex(documentPayloadBytes(was.payload)), key); + assert.equal(now.schemaId, was.schemaId, key); + assert.equal(now.storedAt, was.storedAt, key); + assert.deepEqual(now.catalogEntry, was.catalogEntry, key); + } +}); + +// A writer that adds a slot and keeps its schema id (TS ids are hand-written; C# ids hash the members): the reader of that +// same id has declared nothing, so it consents to nothing. +test("a wider writer under the reader's own schema id is read-only, and a real round trip does not shed the slot", async () => { + const doc = (columns: 1 | 2, compatibleSchemaIds: string[] = []) => defineDocumentType({ + type: "tests.own_id", + schemaId: "tests.own_id.v1", + schemaName: "tests.own_id", + compatibleSchemaIds, + schema: z.object({ n: z.string(), extra: z.string().optional() }), + codec: columns === 2 + ? slots<{ n: string; extra: string }>({ 0: ["n", string], 1: ["extra", string] }, "Own") + : slots<{ n: string }>({ 0: ["n", string] }, "Own"), + name: "n", + }); + const path = join(await mkdtemp(join(tmpdir(), "slot-readonly-")), "own.cc"); + const wide = doc(2); + const writer = CultCache.builder().withDocumentType(wide).withGenericStore(new SingleFileMessagePackBackingStore(path)).build(); + await writer.put(wide, "k", { n: "n1", extra: "kept" }); + const bytes = await readFile(path); + + for (const compatible of [[], ["tests.own_id.v1"]]) { + const reader = doc(1, compatible); + const cache = CultCache.builder().withDocumentType(reader).withGenericStore(new SingleFileMessagePackBackingStore(path)).build(); + await cache.pullAllBackingStores(); + assert.deepEqual(cache.unreadSlots(reader, "k"), mark("Own", 2, 1, true), "unexplained: the arrived id is the reader's own"); + assert.equal(cache.isWritable(reader, "k"), false); + await assert.rejects(() => cache.put(reader, "k", { n: "n2" }), SchemaConflictError); + await assert.rejects(() => cache.delete(reader, "k"), SchemaConflictError); + assert.equal(Buffer.compare(bytes, await readFile(path)), 0, "the file still holds the slot"); + } +}); + +test("a record that arrived under an id the reader does not list compatible is never consented to", async () => { + const path = await copyOf("extra-slot-consent.cc"); + const reader = surfaceDocument("tests.union_surface.v2"); // resolves by type name; lists nothing + const cache = CultCache.builder().withDocumentType(reader).withGenericStore(new SingleFileMessagePackBackingStore(path)).build(); + await cache.pullAllBackingStores(); + assert.equal(cache.unreadSlots(reader, "consented")?.explained, false); + assert.equal(cache.isWritable(reader, "consented"), false); +}); + +// isWritable is the write's own judge, not a second one: for every key state it answers what put does. +test("isWritable equals the outcome of put for every key state the write judges", async () => { + const typed = (type: string, schemaId: string, extra: Record = {}) => + defineDocumentType({ type, schemaId, schemaName: type, schemaVersion: `${type}.v1`, schema: surfaceSchema, codec: surfaceCodec, name: "id", ...extra }); + const older = typed("tests.surface_older", "tests.surface_older.v1"); + const newer = typed("tests.surface_newer", "tests.surface_newer.v1", { compatibleSchemaIds: ["tests.surface_older.v1"] }); + const stranger = typed("tests.surface_stranger", "tests.surface_stranger.v1"); + const global = typed("tests.surface_global", "tests.surface_global.v1", { global: true }); + const clean = (key: string, type: string, schemaId: string) => ({ ...envelopeOf(new Uint8Array()), key, type, schemaId }); + const held = async (key: string, file: string, type: string, schemaId: string) => + ({ ...envelopeOf(await read(file)), key, type, schemaId }); + + type Row = { name: string; build: () => Promise<{ cache: CultCache; target: typeof older; key: string }> }; + const rows: Row[] = [ + { name: "owned clean record in the writer's own store", build: async () => { + const cache = CultCache.builder().withDocumentType(older).withDocumentType(newer) + .withGenericStore(memoryStore([await held("k", "union-surface.msgpack", older.type, older.schemaId!)])).build(); + return { cache, target: newer, key: "k" }; + } }, + { name: "owned record held in a store other than the writer's home", build: async () => { + const cache = CultCache.builder().withDocumentType(older).withDocumentType(newer) + .withBackingStore(memoryStore([await held("k", "union-surface.msgpack", older.type, older.schemaId!)]), older) + .withBackingStore(memoryStore([]), newer).build(); + return { cache, target: newer, key: "k" }; + } }, + { name: "a record of a type that does not own the writer's id", build: async () => { + const cache = CultCache.builder().withDocumentType(older).withDocumentType(stranger) + .withGenericStore(memoryStore([await held("k", "union-surface.msgpack", older.type, older.schemaId!)])).build(); + return { cache, target: stranger, key: "k" }; + } }, + { name: "a global type at a key other than the global key", build: async () => { + const cache = CultCache.builder().withDocumentType(global).withGenericStore(memoryStore([])).build(); + return { cache, target: global, key: "x" }; + } }, + { name: "a key a store carries foreign", build: async () => { + const cache = CultCache.builder().withDocumentType(older) + .withGenericStore(memoryStore([{ ...clean("k", "tests.unknown", "tests.unknown.v1"), payload: documentPayloadFromBytes(await read("union-surface.msgpack")) }])).build(); + return { cache, target: older, key: "k" }; + } }, + { name: "a marked record of the writer's own type", build: async () => { + const cache = CultCache.builder().withDocumentType(older) + .withGenericStore(memoryStore([await held("k", "extra-slot-record.msgpack", older.type, older.schemaId!)])).build(); + return { cache, target: older, key: "k" }; + } }, + { name: "an unheld key", build: async () => ({ cache: CultCache.builder().withDocumentType(older).withGenericStore(memoryStore([])).build(), target: older, key: "k" }) }, + ]; + const outcomes: [string, boolean, boolean][] = []; + for (const row of rows) { + const { cache, target, key } = await row.build(); + await cache.pullAllBackingStores(); + const predicted = cache.isWritable(target, key); + const landed = await cache.put(target, key === "x" ? "x" : key, golden).then(() => true, () => false); + assert.equal(predicted, landed, row.name); + outcomes.push([row.name, predicted, landed]); + } + assert.deepEqual(outcomes.map(([, writable]) => writable), [true, false, false, false, false, false, true]); + // The same judge refuses a removal of a key held read-only or carried foreign. + const removals = CultCache.builder().withDocumentType(older) + .withGenericStore(memoryStore([await held("k", "extra-slot-record.msgpack", older.type, older.schemaId!)])).build(); + await removals.pullAllBackingStores(); + await assert.rejects(() => removals.delete(older, "k"), SchemaConflictError); + const stranger2 = CultCache.builder().withDocumentType(older) + .withGenericStore(memoryStore([{ ...envelopeOf(await read("union-surface.msgpack")), key: "f", type: "tests.unknown", schemaId: "tests.unknown.v1" }])).build(); + await stranger2.pullAllBackingStores(); + await assert.rejects(() => stranger2.delete(older, "f"), SchemaConflictError); +}); + +// A removal is by key (C# Remove): the owner of a held record's id removes the one record that exists, in one store call and +// with the cache dropped only once the store has; a type that does not own it is refused like put, and isWritable agrees. +test("delete of a key held as another type: the owner removes it once, a stranger is refused, a failing store leaves cache and store agreeing", async () => { + const typed = (type: string, schemaId: string, extra: Record = {}) => + defineDocumentType({ type, schemaId, schemaName: type, schemaVersion: `${type}.v1`, schema: surfaceSchema, codec: surfaceCodec, name: "id", ...extra }); + const older = typed("tests.surface_older", "tests.surface_older.v1"); + const newer = typed("tests.surface_newer", "tests.surface_newer.v1", { compatibleSchemaIds: ["tests.surface_older.v1"] }); + const stranger = typed("tests.surface_stranger", "tests.surface_stranger.v1"); + const heldRecord = async (): Promise => ({ ...envelopeOf(await read("union-surface.msgpack")), key: "k", type: older.type, schemaId: older.schemaId }); + const open = async (failing = false) => { + const rows = [await heldRecord()]; + const calls: string[] = []; + const store: CacheBackingStore = { + async pullAll() { return rows.map((row) => ({ ...row })); }, + async push(entry) { rows.push(entry); }, + async delete(entry) { + calls.push(entry.key); + if (failing) throw new Error("store refused the removal"); + rows.splice(rows.findIndex((row) => row.key === entry.key), 1); + }, + }; + const cache = CultCache.builder().withDocumentType(older).withDocumentType(newer).withDocumentType(stranger).withGenericStore(store).build(); + await cache.pullAllBackingStores(); + return { cache, rows, calls }; + }; + + const owner = await open(); + assert.equal(owner.cache.isWritable(newer, "k"), true); + assert.equal(await owner.cache.delete(newer, "k"), true, "the owner's removal removes the one record"); + assert.deepEqual([owner.rows.length, owner.calls.length, owner.cache.get(older, "k")], [0, 1, undefined], "one store call, record gone from store and cache"); + + const refused = await open(); + assert.equal(refused.cache.isWritable(stranger, "k"), false); + await assert.rejects(() => refused.cache.delete(stranger, "k"), SchemaConflictError); + assert.deepEqual([refused.rows.length, refused.calls.length, refused.cache.get(older, "k") !== undefined], [1, 0, true]); + + const failing = await open(true); + await assert.rejects(() => failing.cache.delete(newer, "k"), /store refused/u); + assert.deepEqual([failing.rows.length, failing.calls.length, failing.cache.get(older, "k") !== undefined], [1, 1, true], "the store failed: neither side dropped it"); + + const same = await open(); + assert.equal(await same.cache.delete(older, "k"), true); + assert.equal(await same.cache.delete(older, "k"), false, "an unheld key is still the plain false"); +}); + +// The removal form's three finer behaviours (cultlib-gaps:finding:cut-slot-readonly-ts.s4.removal-form-unpinned). +test("an owner's delete: refused for a marked record, taken from the held record's store, and clears the held type's lookups", async () => { + const typed = (type: string, schemaId: string, extra: Record = {}) => + defineDocumentType({ type, schemaId, schemaName: type, schemaVersion: `${type}.v1`, schema: surfaceSchema, codec: surfaceCodec, name: "id", ...extra }); + const older = typed("tests.surface_older", "tests.surface_older.v1", { indexes: { byVersion: (value: Surface) => String(value.version) } }); + const newer = typed("tests.surface_newer", "tests.surface_newer.v1", { compatibleSchemaIds: ["tests.surface_older.v1"] }); + const heldIn = async (file: string): Promise => + ({ ...envelopeOf(await read(file)), key: "k", type: older.type, schemaId: older.schemaId }); + const recording = (rows: CultCacheEnvelope[]) => { + const removed: string[] = []; + const store: CacheBackingStore = { + async pullAll() { return rows.map((row) => ({ ...row })); }, + async push(entry) { rows.push(entry); }, + async delete(entry) { removed.push(entry.key); rows.splice(rows.findIndex((row) => row.key === entry.key), 1); }, + }; + return { store, removed, rows }; + }; + + // 1. The owner may not remove a record carrying unread slots nothing consents to. + const marked = recording([await heldIn("extra-slot-record.msgpack")]); + const markedCache = CultCache.builder().withDocumentType(older).withDocumentType(newer).withGenericStore(marked.store).build(); + await markedCache.pullAllBackingStores(); + await assert.rejects(() => markedCache.delete(newer, "k"), (error: unknown) => error instanceof SchemaConflictError && error.unreadSlots?.typeName === "UnionSurface"); + assert.deepEqual([marked.rows.length, marked.removed.length, markedCache.get(older, "k") !== undefined], [1, 0, true]); + + // 2. The removal goes to the store of the record that exists, not the deleting type's; there the write form is false (it would + // push to the wrong store) while the removal succeeds. + const olderStore = recording([await heldIn("union-surface.msgpack")]); + const newerStore = recording([]); + const split = CultCache.builder().withDocumentType(older).withDocumentType(newer) + .withBackingStore(olderStore.store, older).withBackingStore(newerStore.store, newer).build(); + await split.pullAllBackingStores(); + assert.equal(split.isWritable(newer, "k"), false, "isWritable answers the write form: the rename cannot replace a record in another store"); + assert.equal(await split.delete(newer, "k"), true); + assert.deepEqual([olderStore.removed, newerStore.removed, split.get(older, "k")], [["k"], [], undefined]); + + // 3. The held type's name and index lookups are cleared with the record. + const single = recording([await heldIn("union-surface.msgpack")]); + const cache = CultCache.builder().withDocumentType(older).withDocumentType(newer).withGenericStore(single.store).build(); + await cache.pullAllBackingStores(); + assert.equal(cache.getKeyByName(older, "surface"), "k"); + assert.equal(cache.getKeyByIndex(older, "byVersion", "3"), "k"); + await cache.delete(newer, "k"); + assert.equal(cache.getKeyByName(older, "surface"), undefined); + assert.equal(cache.getKeyByIndex(older, "byVersion", "3"), undefined); + assert.equal(cache.getAll(older).length, 0); +}); + +test("a global type at a key other than the global key: isWritable is false and delete is the plain false of an unheld key", async () => { + const global = defineDocumentType({ type: "tests.surface_global", schemaId: "tests.surface_global.v1", schemaName: "tests.surface_global", schemaVersion: "tests.surface_global.v1", schema: surfaceSchema, codec: surfaceCodec, name: "id", global: true }); + const cache = CultCache.builder().withDocumentType(global).withGenericStore(memoryStore([])).build(); + assert.equal(cache.isWritable(global, "x"), false); + assert.equal(await cache.delete(global, "x"), false); +}); diff --git a/packages/cultcache-ts/test/union-parity.test.ts b/packages/cultcache-ts/test/union-parity.test.ts index 46dcb42f..d688cb74 100644 --- a/packages/cultcache-ts/test/union-parity.test.ts +++ b/packages/cultcache-ts/test/union-parity.test.ts @@ -1,8 +1,7 @@ import assert from "node:assert/strict"; -import { existsSync } from "node:fs"; import { mkdtemp, readFile, writeFile } from "node:fs/promises"; import { tmpdir } from "node:os"; -import { join, resolve } from "node:path"; +import { join } from "node:path"; import { test } from "node:test"; import { decode, encode } from "@msgpack/msgpack"; @@ -22,190 +21,14 @@ import { encodeDocumentPayload, } from "../src/document"; import { SingleFileMessagePackBackingStore } from "../src/single-file-messagepack-backing-store"; -import type { AnyCultCacheDocumentDefinition, CacheBackingStore, CultCacheEnvelope } from "../src/types"; +import type { AnyCultCacheDocumentDefinition, CultCacheEnvelope } from "../src/types"; + +import { + CANARY, type Node, type Surface, golden, hex, node, outcome, read, surfaceCodec, surfaceDocument, surfaceSchema, vectors, envelopeOf, memoryStore, +} from "./union-surface-fixture"; const { array, boolean, bytes, float32, float64, int32, int64, lazy, list, nullable, slots, string, union, unit } = codecs; -// The names the C# reference gives the fixture's types (CultSchemaTypeNames.FromType), as union-surface.cc carries them. -const CS = "GameCult.Caching.Tests.UnionParityVectorTests+"; - -// The golden is tests/vectors/union-parity, written by the C# reference. Everything below that is bytes or an expected -// outcome is read from those committed files; only the value the payload must decode to is spelled here. -function findVectors(from: string): string { - for (let dir = from; ; dir = resolve(dir, "..")) { - if (existsSync(join(dir, "tests", "vectors", "union-parity"))) return join(dir, "tests", "vectors", "union-parity"); - if (resolve(dir, "..") === dir) throw new Error("tests/vectors/union-parity not found above the test."); - } -} -const vectors = findVectors(__dirname); -const read = async (name: string) => new Uint8Array(await readFile(join(vectors, name))); - -// The string the C# writer put inside every refusal payload, to prove a refusal never echoes the payload. -const CANARY = "CANARY-union-golden-4c2a"; - -type Node = - | { kind: "Box"; id: string; children: (Node | null)[] } - | { kind: "Text"; id: string; content: string } - | { kind: "Use"; id: string; composition: Composition; slots: (Node | null)[] } - | { kind: "Image"; id: string; uri: string; alt: string } - | { kind: "Wrap"; inner: Node | null; id: string }; -interface Composition { name: string; body: Node | null } -type Outcome = { kind: "Pending"; message: string } | { kind: "Completed"; sourceVersion: number } | { kind: "Denied" } | { kind: "Floor" } | { kind: "Ceiling" }; -// The golden's Outcome: the arms of a list carry an element id, so Denied, Floor and Ceiling are one-slot records. -type GoldenOutcome = - | { kind: "Pending"; message: string; id: string } - | { kind: "Completed"; sourceVersion: number; id: string } - | { kind: "Denied"; id: string } - | { kind: "Floor"; id: string } - | { kind: "Ceiling"; id: string }; -type Item = { kind: "Gem"; name: string } | { kind: "Tool"; name: string; uses: number } | { kind: "Part"; name: string; count: number }; -type CraftedItem = { kind: "Tool"; name: string; uses: number } | { kind: "Part"; name: string; count: number }; -interface Surface { - id: string; - version: number; - root: Node | null; - outcomes: (GoldenOutcome | null)[]; - part: Item | null; - crafted: CraftedItem | null; - missing: Node | null; -} - -const composition = slots({ 0: ["name", string], 1: ["body", lazy(() => node)] }, `${CS}Composition`); -const node: Codec = union("Node", { - 0: ["Box", slots<{ id: string; children: (Node | null)[] }>({ 0: ["id", string], 1: ["children", list(lazy(() => node))] })], - 1: ["Text", slots<{ id: string; content: string }>({ 0: ["id", string], 1: ["content", string] })], - 2: ["Use", slots<{ id: string; composition: Composition; slots: (Node | null)[] }>({ - 0: ["id", string], 1: ["composition", composition], 2: ["slots", list(lazy(() => node))], - })], - 5: ["Image", slots<{ id: string; uri: string; alt: string }>({ 0: ["id", string], 1: ["uri", string], 3: ["alt", string] })], - 6: ["Wrap", slots<{ inner: Node | null; id: string }>({ 0: ["inner", lazy(() => node)], 1: ["id", string] })], -}, `${CS}Node`); -const outcome = union("Outcome", { - 0: ["Pending", slots<{ message: string }>({ 0: ["message", string] })], - 1: ["Completed", slots<{ sourceVersion: number }>({ 0: ["sourceVersion", int32] })], - 2: ["Denied", unit], - [-2147483648]: ["Floor", unit], - 2147483647: ["Ceiling", unit], -}, `${CS}Outcome`); -const idOnly = slots<{ id: string }>({ 0: ["id", string] }); -const goldenOutcome = union("Outcome", { - 0: ["Pending", slots<{ message: string; id: string }>({ 0: ["message", string], 1: ["id", string] })], - 1: ["Completed", slots<{ sourceVersion: number; id: string }>({ 0: ["sourceVersion", int32], 1: ["id", string] })], - 2: ["Denied", idOnly], - [-2147483648]: ["Floor", idOnly], - 2147483647: ["Ceiling", idOnly], -}, `${CS}Outcome`); -const toolArm = slots<{ name: string; uses: number }>({ 0: ["name", string], 1: ["uses", int32] }); -const partArm = slots<{ name: string; count: number }>({ 0: ["name", string], 1: ["count", int32] }); -const item = union("Item", { - 0: ["Gem", slots<{ name: string }>({ 0: ["name", string] })], - 1: ["Tool", toolArm], - 2: ["Part", partArm], -}, `${CS}Item`); -const craftedItem = union("CraftedItem", { 0: ["Tool", toolArm], 1: ["Part", partArm] }, `${CS}CraftedItem`); -const surfaceCodec = slots({ - 0: ["id", string], - 1: ["version", int32], - 2: ["root", node], - 3: ["outcomes", list(goldenOutcome)], - 4: ["part", item], - 5: ["crafted", craftedItem], - 6: ["missing", node, null], -}, `${CS}UnionSurface`); - -const nodeSchema: z.ZodType = z.lazy(() => - z.union([ - z.null(), - z.discriminatedUnion("kind", [ - z.object({ kind: z.literal("Box"), id: z.string(), children: z.array(nodeSchema) }), - z.object({ kind: z.literal("Text"), id: z.string(), content: z.string() }), - z.object({ - kind: z.literal("Use"), - id: z.string(), - composition: z.object({ name: z.string(), body: nodeSchema }), - slots: z.array(nodeSchema), - }), - z.object({ kind: z.literal("Image"), id: z.string(), uri: z.string(), alt: z.string() }), - z.object({ kind: z.literal("Wrap"), inner: nodeSchema, id: z.string() }), - ]), - ])); -const surfaceSchema = z.object({ - id: z.string(), - version: z.number(), - root: nodeSchema, - outcomes: z.array(z.union([ - z.null(), - z.discriminatedUnion("kind", [ - z.object({ kind: z.literal("Pending"), message: z.string(), id: z.string() }), - z.object({ kind: z.literal("Completed"), sourceVersion: z.number(), id: z.string() }), - z.object({ kind: z.literal("Denied"), id: z.string() }), - z.object({ kind: z.literal("Floor"), id: z.string() }), - z.object({ kind: z.literal("Ceiling"), id: z.string() }), - ]), - ])), - part: z.union([z.null(), z.discriminatedUnion("kind", [ - z.object({ kind: z.literal("Gem"), name: z.string() }), - z.object({ kind: z.literal("Tool"), name: z.string(), uses: z.number() }), - z.object({ kind: z.literal("Part"), name: z.string(), count: z.number() }), - ])]), - crafted: z.union([z.null(), z.discriminatedUnion("kind", [ - z.object({ kind: z.literal("Tool"), name: z.string(), uses: z.number() }), - z.object({ kind: z.literal("Part"), name: z.string(), count: z.number() }), - ])]), - missing: nodeSchema, -}) as unknown as z.ZodType; - -function surfaceDocument(schemaId?: string) { - return defineDocumentType({ - type: "tests.union_surface", - schemaId: schemaId ?? "tests.union_surface.v1", - schemaName: "tests.union_surface", - schemaVersion: "tests.union_surface.v1", - schema: surfaceSchema, - codec: surfaceCodec, - name: "id", - }); -} - -const golden: Surface = { - id: "surface", - version: 3, - root: { - kind: "Box", - id: "0000000000c0", - children: [ - { kind: "Text", id: "0000000000c1", content: "hello" }, - { - kind: "Use", - id: "0000000000c2", - composition: { - name: "comp", - body: { - kind: "Box", - id: "0000000000d0", - children: [ - { kind: "Text", id: "0000000000d1", content: "x" }, - { kind: "Image", id: "0000000000d2", uri: "uri://a", alt: "alt" }, - ], - }, - }, - slots: [{ kind: "Text", id: "0000000000e1", content: "slot" }, null, { kind: "Box", id: "0000000000e2", children: [] }], - }, - { kind: "Image", id: "0000000000c3", uri: "uri://b", alt: "" }, - ], - }, - outcomes: [ - { kind: "Pending", message: "wait", id: "0000000000a1" }, - { kind: "Completed", sourceVersion: 7, id: "0000000000a2" }, - { kind: "Denied", id: "0000000000a3" }, - ], - part: { kind: "Tool", name: "hammer", uses: 12 }, - crafted: { kind: "Part", name: "gear", count: 4 }, - missing: null, -}; - -const hex = (data: Uint8Array) => Buffer.from(data).toString("hex"); - async function outcomeEntries() { const entries = decode(await read("outcomes.msgpack")) as [string, string, string, number | null, string][]; return entries.map(([file, outcome, unionName, key, detail]) => ({ file, outcome, unionName, key, detail })); @@ -729,19 +552,6 @@ test("the Int32 edges of an arm key: the largest is an arm, one past either end assert.throws(() => edge.decode(readWire(fromHex("92d3ffffffff8000000090"))), { message: "Union U: unknown arm key -2147483648 (declared keys: 2147483647)." }); }); -function envelopeOf(payload: Uint8Array): CultCacheEnvelope { - return { key: "surface", type: "tests.union_surface", payload: documentPayloadFromBytes(payload), storedAt: new Date().toISOString() }; -} - -function memoryStore(entries: CultCacheEnvelope[]): CacheBackingStore & { entries: CultCacheEnvelope[] } { - return { - entries, - async pullAll() { return entries.map((entry) => ({ ...entry })); }, - async push(entry) { entries.push(entry); }, - async delete() {}, - }; -} - test("arm keys span Int32 and are listed ascending; slot keys are non-negative Int32", () => { const keys = union<{ kind: "High" | "Zero" | "Neg" | "Low" }>("U", { 2147483647: ["High", unit], diff --git a/packages/cultcache-ts/test/union-surface-fixture.ts b/packages/cultcache-ts/test/union-surface-fixture.ts new file mode 100644 index 00000000..1b556eb1 --- /dev/null +++ b/packages/cultcache-ts/test/union-surface-fixture.ts @@ -0,0 +1,240 @@ +import { existsSync } from "node:fs"; +import { readFile } from "node:fs/promises"; +import { join, resolve } from "node:path"; + +import { z } from "zod"; + +import { codecs } from "../src"; +import type { Codec } from "../src/codec"; +import { defineDocumentType, documentPayloadFromBytes } from "../src/document"; +import type { CacheBackingStore, CultCacheEnvelope } from "../src/types"; + +// The union-surface fixture: the codecs, schema and value tests/vectors/union-parity was written from, shared by the +// union-parity walker and the slot-readonly walker so the shape is spelled once. +const { list, int32, lazy, slots, string, union, unit } = codecs; + + +// The names the C# reference gives the fixture's types (CultSchemaTypeNames.FromType), as union-surface.cc carries them. +export const CS = "GameCult.Caching.Tests.UnionParityVectorTests+"; + +// The golden is tests/vectors/union-parity, written by the C# reference. Everything below that is bytes or an expected +// outcome is read from those committed files; only the value the payload must decode to is spelled here. +export function findVectors(from: string): string { + for (let dir = from; ; dir = resolve(dir, "..")) { + if (existsSync(join(dir, "tests", "vectors", "union-parity"))) return join(dir, "tests", "vectors", "union-parity"); + if (resolve(dir, "..") === dir) throw new Error("tests/vectors/union-parity not found above the test."); + } +} +export const vectors = findVectors(__dirname); +export const read = async (name: string) => new Uint8Array(await readFile(join(vectors, name))); + +// The string the C# writer put inside every refusal payload, to prove a refusal never echoes the payload. +export const CANARY = "CANARY-union-golden-4c2a"; + +export type Node = + | { kind: "Box"; id: string; children: (Node | null)[] } + | { kind: "Text"; id: string; content: string } + | { kind: "Use"; id: string; composition: Composition; slots: (Node | null)[] } + | { kind: "Image"; id: string; uri: string; alt: string } + | { kind: "Wrap"; inner: Node | null; id: string }; +export interface Composition { name: string; body: Node | null } +export type Outcome = { kind: "Pending"; message: string } | { kind: "Completed"; sourceVersion: number } | { kind: "Denied" } | { kind: "Floor" } | { kind: "Ceiling" }; +// The golden's Outcome: the arms of a list carry an element id, so Denied, Floor and Ceiling are one-slot records. +export type GoldenOutcome = + | { kind: "Pending"; message: string; id: string } + | { kind: "Completed"; sourceVersion: number; id: string } + | { kind: "Denied"; id: string } + | { kind: "Floor"; id: string } + | { kind: "Ceiling"; id: string }; +export type Item = { kind: "Gem"; name: string } | { kind: "Tool"; name: string; uses: number } | { kind: "Part"; name: string; count: number }; +export type CraftedItem = { kind: "Tool"; name: string; uses: number } | { kind: "Part"; name: string; count: number }; +export interface Surface { + id: string; + version: number; + root: Node | null; + outcomes: (GoldenOutcome | null)[]; + part: Item | null; + crafted: CraftedItem | null; + missing: Node | null; +} + +export const composition = slots({ 0: ["name", string], 1: ["body", lazy(() => node)] }, `${CS}Composition`); +export const node: Codec = union("Node", { + 0: ["Box", slots<{ id: string; children: (Node | null)[] }>({ 0: ["id", string], 1: ["children", list(lazy(() => node))] }, `${CS}Box`)], + 1: ["Text", slots<{ id: string; content: string }>({ 0: ["id", string], 1: ["content", string] }, `${CS}Text`)], + 2: ["Use", slots<{ id: string; composition: Composition; slots: (Node | null)[] }>({ + 0: ["id", string], 1: ["composition", composition], 2: ["slots", list(lazy(() => node))], + })], + 5: ["Image", slots<{ id: string; uri: string; alt: string }>({ 0: ["id", string], 1: ["uri", string], 3: ["alt", string] })], + 6: ["Wrap", slots<{ inner: Node | null; id: string }>({ 0: ["inner", lazy(() => node)], 1: ["id", string] })], +}, `${CS}Node`); +export const outcome = union("Outcome", { + 0: ["Pending", slots<{ message: string }>({ 0: ["message", string] })], + 1: ["Completed", slots<{ sourceVersion: number }>({ 0: ["sourceVersion", int32] })], + 2: ["Denied", unit], + [-2147483648]: ["Floor", unit], + 2147483647: ["Ceiling", unit], +}, `${CS}Outcome`); +export const idOnly = slots<{ id: string }>({ 0: ["id", string] }); +export const goldenOutcome = union("Outcome", { + 0: ["Pending", slots<{ message: string; id: string }>({ 0: ["message", string], 1: ["id", string] })], + 1: ["Completed", slots<{ sourceVersion: number; id: string }>({ 0: ["sourceVersion", int32], 1: ["id", string] })], + 2: ["Denied", idOnly], + [-2147483648]: ["Floor", idOnly], + 2147483647: ["Ceiling", idOnly], +}, `${CS}Outcome`); +export const toolArm = slots<{ name: string; uses: number }>({ 0: ["name", string], 1: ["uses", int32] }); +export const partArm = slots<{ name: string; count: number }>({ 0: ["name", string], 1: ["count", int32] }); +export const item = union("Item", { + 0: ["Gem", slots<{ name: string }>({ 0: ["name", string] })], + 1: ["Tool", toolArm], + 2: ["Part", partArm], +}, `${CS}Item`); +export const craftedItem = union("CraftedItem", { 0: ["Tool", toolArm], 1: ["Part", partArm] }, `${CS}CraftedItem`); +export const surfaceCodecOf = (typeName: string) => slots({ + 0: ["id", string], + 1: ["version", int32], + 2: ["root", node], + 3: ["outcomes", list(goldenOutcome)], + 4: ["part", item], + 5: ["crafted", craftedItem], + 6: ["missing", node, null], +}, typeName); +export const surfaceCodec = surfaceCodecOf(`${CS}UnionSurface`); + +export const nodeSchema: z.ZodType = z.lazy(() => + z.union([ + z.null(), + z.discriminatedUnion("kind", [ + z.object({ kind: z.literal("Box"), id: z.string(), children: z.array(nodeSchema) }), + z.object({ kind: z.literal("Text"), id: z.string(), content: z.string() }), + z.object({ + kind: z.literal("Use"), + id: z.string(), + composition: z.object({ name: z.string(), body: nodeSchema }), + slots: z.array(nodeSchema), + }), + z.object({ kind: z.literal("Image"), id: z.string(), uri: z.string(), alt: z.string() }), + z.object({ kind: z.literal("Wrap"), inner: nodeSchema, id: z.string() }), + ]), + ])); +export const surfaceSchema = z.object({ + id: z.string(), + version: z.number(), + root: nodeSchema, + outcomes: z.array(z.union([ + z.null(), + z.discriminatedUnion("kind", [ + z.object({ kind: z.literal("Pending"), message: z.string(), id: z.string() }), + z.object({ kind: z.literal("Completed"), sourceVersion: z.number(), id: z.string() }), + z.object({ kind: z.literal("Denied"), id: z.string() }), + z.object({ kind: z.literal("Floor"), id: z.string() }), + z.object({ kind: z.literal("Ceiling"), id: z.string() }), + ]), + ])), + part: z.union([z.null(), z.discriminatedUnion("kind", [ + z.object({ kind: z.literal("Gem"), name: z.string() }), + z.object({ kind: z.literal("Tool"), name: z.string(), uses: z.number() }), + z.object({ kind: z.literal("Part"), name: z.string(), count: z.number() }), + ])]), + crafted: z.union([z.null(), z.discriminatedUnion("kind", [ + z.object({ kind: z.literal("Tool"), name: z.string(), uses: z.number() }), + z.object({ kind: z.literal("Part"), name: z.string(), count: z.number() }), + ])]), + missing: nodeSchema, +}) as unknown as z.ZodType; + +// The reference's SurfaceDeclaring: the same layout, schema tests.union_surface.v2, listing the older wide id compatible. +export function surfaceDeclaringDocument() { + return defineDocumentType({ + type: "tests.union_surface", + schemaId: "tests.union_surface.v2", + schemaName: "tests.union_surface", + schemaVersion: "tests.union_surface.v2", + compatibleSchemaIds: ["tests.union_surface.wide"], + schema: surfaceSchema, + codec: surfaceCodecOf(`${CS}SurfaceDeclaring`), + name: "id", + }); +} + +export function surfaceDocument(schemaId?: string, contentHash?: string) { + return defineDocumentType({ + ...(contentHash === undefined ? {} : { contentHash }), + type: "tests.union_surface", + schemaId: schemaId ?? "tests.union_surface.v1", + schemaName: "tests.union_surface", + schemaVersion: "tests.union_surface.v1", + schema: surfaceSchema, + codec: surfaceCodec, + name: "id", + }); +} + +export const golden: Surface = { + id: "surface", + version: 3, + root: { + kind: "Box", + id: "0000000000c0", + children: [ + { kind: "Text", id: "0000000000c1", content: "hello" }, + { + kind: "Use", + id: "0000000000c2", + composition: { + name: "comp", + body: { + kind: "Box", + id: "0000000000d0", + children: [ + { kind: "Text", id: "0000000000d1", content: "x" }, + { kind: "Image", id: "0000000000d2", uri: "uri://a", alt: "alt" }, + ], + }, + }, + slots: [{ kind: "Text", id: "0000000000e1", content: "slot" }, null, { kind: "Box", id: "0000000000e2", children: [] }], + }, + { kind: "Image", id: "0000000000c3", uri: "uri://b", alt: "" }, + ], + }, + outcomes: [ + { kind: "Pending", message: "wait", id: "0000000000a1" }, + { kind: "Completed", sourceVersion: 7, id: "0000000000a2" }, + { kind: "Denied", id: "0000000000a3" }, + ], + part: { kind: "Tool", name: "hammer", uses: 12 }, + crafted: { kind: "Part", name: "gear", count: 4 }, + missing: null, +}; +export const hex = (data: Uint8Array) => Buffer.from(data).toString("hex"); + +// tests.slot_unit: a document [Name, Outcome: UnitOutcome], whose one arm Denied2 is a unit arm (extra-slot-store.cc, record +// unit-arm). The reader's names are the reference's: a mark names Denied2. +export const unitOutcome = union<{ kind: "Denied2" }>("UnitOutcome", { 2: ["Denied2", unit] }, `${CS}UnitOutcome`); +export function slotUnitDocument(schemaId: string, contentHash?: string) { + return defineDocumentType({ + ...(contentHash === undefined ? {} : { contentHash }), + type: "tests.slot_unit", + schemaId, + schemaName: "tests.slot_unit", + schemaVersion: "tests.slot_unit.v1", + schema: z.object({ name: z.string(), outcome: z.union([z.null(), z.object({ kind: z.literal("Denied2") })]) }), + codec: slots<{ name: string; outcome: { kind: "Denied2" } | null }>({ 0: ["name", string], 1: ["outcome", unitOutcome] }, `${CS}SlotUnit`), + name: "name", + }); +} + +export function envelopeOf(payload: Uint8Array): CultCacheEnvelope { + return { key: "surface", type: "tests.union_surface", payload: documentPayloadFromBytes(payload), storedAt: new Date().toISOString() }; +} + +export function memoryStore(entries: CultCacheEnvelope[]): CacheBackingStore & { entries: CultCacheEnvelope[] } { + return { + entries, + async pullAll() { return entries.map((entry) => ({ ...entry })); }, + async push(entry) { entries.push(entry); }, + async delete() {}, + }; +} + From 13ce5ea50a0e231819eb6f33ae76db1296ac06d2 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 05:23:25 +0000 Subject: [PATCH 05/12] Python: a payload is one MessagePack value (bytes-after-py) (#129) --- packages/cultcache-py/CHANGELOG.md | 6 +- packages/cultcache-py/README.md | 1 + packages/cultcache-py/pyproject.toml | 2 +- .../cultcache-py/src/cultcache_py/__init__.py | 2 + .../src/cultcache_py/documents.py | 26 ++++- .../cultcache-py/tests/test_union_parity.py | 72 ++++++++++++++ packages/cultmesh-py/pyproject.toml | 2 +- packages/cultnet-py/CHANGELOG.md | 1 + packages/cultnet-py/pyproject.toml | 2 +- .../cultnet-py/src/cultnet_py/messages.py | 3 +- .../cultnet-py/src/cultnet_py/replication.py | 6 +- packages/cultnet-py/tests/test_cultnet.py | 45 +++++++++ .../tests/test_declared_dependency_ranges.py | 94 +++++++++++++++++++ 13 files changed, 251 insertions(+), 11 deletions(-) create mode 100644 packages/cultnet-py/tests/test_declared_dependency_ranges.py diff --git a/packages/cultcache-py/CHANGELOG.md b/packages/cultcache-py/CHANGELOG.md index 08fb4d1f..ea3f72fd 100644 --- a/packages/cultcache-py/CHANGELOG.md +++ b/packages/cultcache-py/CHANGELOG.md @@ -2,6 +2,10 @@ ## Unreleased +### Added + +- `read_payload_tree(type_name, payload, decoder=None)`: the exported owner of the one-value, depth-first payload read that `decode_payload` uses, for payloads read outside a definition (`cultnet-py`'s witness bundle decode needs it, so `cultnet-py` now requires `cultcache-py>=0.4.0`). The manifest version is 0.4.0, the minor these Unreleased breaking changes imply, set in this change so the workspace install resolves the new bound; the `## Unreleased` heading becomes `## [0.4.0]` when it is tagged. + ### Changed - `msgpack` is a hard dependency, as `pyproject.toml` declared it. The optional-import fallbacks and the `cultcache-py[msgpack]` extra they named (which never existed) are gone: a missing msgpack is an `ImportError` at import. @@ -10,7 +14,7 @@ ### Breaking - A record whose payload carries a slot past its reader's declared slots, at any depth (the document, a nested record, a list element, a union arm, a unit arm), is read, held read-only, and every write or removal of its key (`put`, `put_global`, `update`, `put_envelope`, `put_envelopes`, `delete`, `delete_global`) is refused with `SchemaConflictError` before any store is touched (`cultlib-gaps:ruling:extra-trailing-slots`). Reading dropped those slots silently and the next write shed them. `SchemaConflictError.unread_slots` is an `UnreadSlots(type_name, slot_count, declared_count, occurrences, depth0_only)` and the message names the type and both counts, never a slot value. A depth-0 mark is consented, and the record writable, only when the record is stored under an id the document lists in `compatible_schema_ids` and that id's catalog entry has slots past the document's, covering the mark's slot count (`cultlib-gaps:ruling:slot-mark-consent`); a nested mark, or a mark under the document's own id, keeps the record read-only whatever is declared. Missing trailing slots keep their declared defaults. An untouched marked record keeps its stored bytes through every other write. Added `CultCache.is_writable(document, key)` (advisory; the write still decides: it also answers False for a foreign-carried key, for a key held as a type the document does not own, and for a global document at a non-global key; `delete` asks the same judges, so removing a key held as another type is refused with `SchemaConflictError` unless the document owns that record's schema id (its own or one it lists compatible), in which case the removal removes that record, its file row and its lookups, as a write of the key would replace it; it returned silently and removed nothing), `DocumentDefinition.decode_payload_marked`, and `UnreadSlots`. A document with a custom `decode` (no slot codec) never marks. A store holding a variant record is still unreadable here. Implies a minor bump while the version is 0.x (0.3.0 to 0.4.0): code that wrote back a record from a newer writer now gets `SchemaConflictError`. - +- A payload is one MessagePack value: `DocumentDefinition.decode_payload` refuses bytes after it, by count, with `ValueError` ` payload has N bytes after its first MessagePack value (refused on read).` (never the bytes, never the depth refusal; a payload that ends early stays the decoder's), as the C# reference does (`tests/vectors/union-parity/bytes-after-outcomes.msgpack`). It raised `msgpack.ExtraData` (a `ValueError`, so only a caller naming `msgpack.ExtraData` is affected); this is a minor-version bump while pre-1.0. The refused set is unchanged. Cache loads, `get` and every definition built on the MessagePack decoder go through it; a user `payload_decoder` that raises `ExtraData` is refused the same way. Depth is judged first, so a value nested past 256 with a tail is the depth refusal, as in C#. - A cache holds one record per key whatever its type (`variants:ruling:cache-refuses-retype`). A `put`, `put_envelope`, `put_envelopes` or `put_global` of a second type under a key the cache holds, or under a key a store carries foreign, is refused with `SchemaConflictError` naming the key, even with no store, and changes nothing; retype a key by deleting it, then putting the new type. A write may replace another registered type's record only when the written type owns its schema id (declares it or lists it compatible) and both types share one home store. A load that reads one key from two stores, whatever the types, is refused with `CultCacheError` naming the key and both types, and admits nothing. A store that holds a record no registered type claims now loads and is carried: the cache does not serve it, the store keeps its bytes through every write, `delete` or `put` at its key is refused with `SchemaConflictError`, and registering a type that declares its schema id claims it at the next load. It raised `CultCacheError` "Unknown persisted document type". Two global types share the key `__global__`, so a second global type in one cache is now refused, and a store that already holds two global types at `__global__` (a JSONL store written by caches that never pulled could produce one) no longer loads: the pull raises `CultCacheError` and admits nothing. C# refuses a second persisted global on pull the same way (`cultcache-store-composition.md`, "Globals never invented"); repair such a file by deleting one of the two records with the runtime that wrote it. - Added `foreign_records()` and the frozen `ForeignRecord(key, schema_id, schema_name, stored_at)`, sorted by key in UTF-16 code-unit order. `stores.retyped` and `stores.written_owns` lose their underscore: the cache judge and the single-file store's commit guard ask one predicate. diff --git a/packages/cultcache-py/README.md b/packages/cultcache-py/README.md index 72d51562..4a7236fd 100644 --- a/packages/cultcache-py/README.md +++ b/packages/cultcache-py/README.md @@ -19,6 +19,7 @@ discovery, sessions, local server) build on it, mirroring `cultcache-rs` / `[format, schemaCatalog, records]` - records store raw MessagePack payload bytes under schema-catalog identity - payloads decode only through the registered document definition for that type +- a payload is exactly one MessagePack value: bytes after it are refused on read, by count - a cache holds one record per key whatever its type: a write of a second type at a held key is refused with `SchemaConflictError` (in memory too), and two stores holding one key refuse the load - a store record no registered type claims loads and is carried as foreign: `cache.foreign_records()` lists it, the store diff --git a/packages/cultcache-py/pyproject.toml b/packages/cultcache-py/pyproject.toml index 6c46d102..15b724ba 100644 --- a/packages/cultcache-py/pyproject.toml +++ b/packages/cultcache-py/pyproject.toml @@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta" [project] name = "cultcache-py" -version = "0.3.0" +version = "0.4.0" description = "Python CultCache runtime for GameCult typed state." readme = "README.md" requires-python = ">=3.10" diff --git a/packages/cultcache-py/src/cultcache_py/__init__.py b/packages/cultcache-py/src/cultcache_py/__init__.py index 4873b6ed..69828bf3 100644 --- a/packages/cultcache-py/src/cultcache_py/__init__.py +++ b/packages/cultcache-py/src/cultcache_py/__init__.py @@ -16,6 +16,7 @@ from .documents import ( define_document_type, define_record_codec, define_union, + read_payload_tree, ) from .leaves import SlotCodec from .stores import JsonLinesBackingStore, SchemaConflictError, SingleFileMessagePackBackingStore, StoreUnreadableError @@ -40,6 +41,7 @@ __all__ = [ "StoreUnreadableError", "UnionArm", "UnreadSlots", + "read_payload_tree", "leaves", "database_entry_field", "define_database_entry_type", diff --git a/packages/cultcache-py/src/cultcache_py/documents.py b/packages/cultcache-py/src/cultcache_py/documents.py index 4f7dfe7d..ff9039dc 100644 --- a/packages/cultcache-py/src/cultcache_py/documents.py +++ b/packages/cultcache-py/src/cultcache_py/documents.py @@ -158,6 +158,26 @@ def _messagepack_payload_decoder(payload: bytes) -> Any: return msgpack.unpackb(payload, raw=False) +def read_payload_tree(type_name: str, payload: bytes, decoder: Callable[[bytes], Any] | None = None) -> Any: + """The one read of a payload's tree: exactly one MessagePack value, within the nesting limit. Depth is judged first, + on the value read so far, so a tail never pre-empts it (C# throws at container 257 before it reaches the tail); then a + tail is refused by count and never echoed. Raised outside the handler so no `__context__` keeps the tail.""" + after = -1 + unpacked: Any = None + try: + tree = (decoder or _messagepack_payload_decoder)(payload) + except _STACK_FAULTS: + raise CultCachePayloadDepthError(type_name, "read") from None + except msgpack.ExtraData as error: + after = len(error.extra) + unpacked = error.unpacked + if after >= 0: + _refuse_past_depth_limit(unpacked, type_name, "read") + raise ValueError(f"{type_name} payload has {after} bytes after its first MessagePack value (refused on read).") + _refuse_past_depth_limit(tree, type_name, "read") + return tree + + @dataclass(frozen=True) class DatabaseEntryField: name: str @@ -234,11 +254,7 @@ class DocumentDefinition(Generic[T]): def decode_payload_marked(self, payload: bytes) -> tuple[T, UnreadSlots | None]: """The value and what the decode found past the declared slots, at any depth. The payload depth check runs first: a payload past the limit is refused, never marked.""" - try: - tree = self.payload_decoder(payload) - except _STACK_FAULTS: - raise CultCachePayloadDepthError(self.type, "read") from None - _refuse_past_depth_limit(tree, self.type, "read") + tree = read_payload_tree(self.type, payload, self.payload_decoder) outer, opened = _current_scope(), _SlotScope() _scope.open = opened try: diff --git a/packages/cultcache-py/tests/test_union_parity.py b/packages/cultcache-py/tests/test_union_parity.py index 1e19ce8d..793933ff 100644 --- a/packages/cultcache-py/tests/test_union_parity.py +++ b/packages/cultcache-py/tests/test_union_parity.py @@ -661,6 +661,78 @@ def _as_tuple(tree: object) -> object: return tuple(_as_tuple(child) for child in tree) if isinstance(tree, list) else tree + +class BytesAfterVectorTests(unittest.TestCase): + CLAUSE = "bytes after its first MessagePack value (refused on read)" + + def _assert_chain_clean(self, error: BaseException, file: str) -> None: + link: BaseException | None = error + while link is not None: + self.assertNotIn(CANARY, str(link), file) + self.assertNotIsInstance(link, msgpack.ExtraData, file) + link = link.__cause__ or link.__context__ + + def test_every_bytes_after_vector_is_refused(self) -> None: + base = _vector("union-surface.msgpack") + entries = msgpack.unpackb(_vector("bytes-after-outcomes.msgpack"), raw=False) + self.assertEqual(len(entries), 4) + for file, outcome_kind, union, key, detail in entries: + self.assertEqual((outcome_kind, union, key, detail), ("refused", "UnionSurface", None, "bytes-after"), file) + payload = _vector(file) + self.assertTrue(payload.startswith(base), file) + count = len(payload) - len(base) + with self.assertRaises(ValueError, msg=file) as caught: + surface_document.decode_payload(payload) + error = caught.exception + self.assertIs(type(error), ValueError, file) + self.assertNotIsInstance(error, CultCachePayloadDepthError, file) + self.assertEqual( + str(error), f"{surface_document.type} payload has {count} {self.CLAUSE}.", file + ) + self.assertNotIn("the limit is", str(error), file) + self._assert_chain_clean(error, file) + + def test_a_single_file_store_holding_a_tail_as_a_record_payload_refuses_on_load(self) -> None: + with tempfile.TemporaryDirectory() as tmp: + path = Path(tmp) / "tail.cc" + shutil.copyfile(VECTORS / "union-surface.cc", path) + store = SingleFileMessagePackBackingStore(path) + [envelope] = store.pull_all() + store.push(dataclasses.replace(envelope, payload=_vector("bytes-after-nil.msgpack"))) + cache = ( + CultCache.builder() + .register_document_type(surface_document) + .add_generic_store(SingleFileMessagePackBackingStore(path)) + .build() + ) + with self.assertRaises(ValueError) as caught: + cache.pull_all_backing_stores() + self.assertIn(f"1 {self.CLAUSE}", str(caught.exception)) + self.assertIn(surface_document.type, str(caught.exception)) + self.assertIsNone(cache.get(surface_document, "surface")) + + def test_depth_wins_over_a_tail_whatever_the_unpackers_own_stack_limit(self) -> None: + # Vector-independent: a value nested past 256 is the depth refusal even with a tail, as C# throws at container 257 + # before it reaches the tail; 2000 is past msgpack's own stack, so the winner is not a library constant. + for depth in (257, 300, 900, 2000, 100000): + for tail in (b"\xc0", b"\xc1\xc1"): + with self.subTest(depth=depth, tail=len(tail)): + with self.assertRaises(CultCachePayloadDepthError) as caught: + surface_document.decode_payload(b"\x91" * depth + b"\xc0" + tail) + self.assertIn("refused on read", str(caught.exception)) + for depth in (1, 255, 256): + with self.subTest(at=depth): + with self.assertRaises(ValueError) as caught: + surface_document.decode_payload(b"\x91" * depth + b"\xc0\xc0") + self.assertNotIsInstance(caught.exception, CultCachePayloadDepthError) + self.assertIn(f"1 {self.CLAUSE}", str(caught.exception)) + + def test_a_payload_that_ends_early_is_the_decoders_not_bytes_after(self) -> None: + truncated = _vector("union-surface.msgpack")[:-1] + with self.assertRaises(Exception) as caught: + surface_document.decode_payload(truncated) + self.assertNotIn("bytes after", str(caught.exception)) + class PayloadDepthRuleTests(unittest.TestCase): """The nesting rule at its seam, on definitions that are not unions: the same routine serves any payload codec.""" diff --git a/packages/cultmesh-py/pyproject.toml b/packages/cultmesh-py/pyproject.toml index 5f9bda4d..18f23060 100644 --- a/packages/cultmesh-py/pyproject.toml +++ b/packages/cultmesh-py/pyproject.toml @@ -16,7 +16,7 @@ classifiers = [ "Programming Language :: Python :: 3.11", "Programming Language :: Python :: 3.12", ] -dependencies = ["cultcache-py>=0.2.0", "cultnet-py>=0.1.0", "cryptography>=41"] +dependencies = ["cultcache-py>=0.4.0", "cultnet-py>=0.1.0", "cryptography>=41"] [project.scripts] cultmesh-py-daemon = "cultmesh_py.daemon:main" cultmesh-py-verify = "cultmesh_py.verify:main" diff --git a/packages/cultnet-py/CHANGELOG.md b/packages/cultnet-py/CHANGELOG.md index b8a3773e..525bd4bd 100644 --- a/packages/cultnet-py/CHANGELOG.md +++ b/packages/cultnet-py/CHANGELOG.md @@ -2,6 +2,7 @@ ## Unreleased (breaking; version 0.x, so the next minor) +- `decode_witness_artifact_bundle_payload` (and `CultNetWitnessArtifactBundle.from_payload`) read the payload through `cultcache_py.read_payload_tree`: bytes after its first MessagePack value raise `ValueError` `cultnet.witness_artifact_bundle payload has N bytes after its first MessagePack value (refused on read).` instead of `msgpack.ExtraData`, and a payload nested past 256 raises `CultCachePayloadDepthError`. Needs `cultcache-py` 0.4.0, the release that exports `read_payload_tree`: the dependency is now `cultcache-py>=0.4.0` (it was `>=0.2.0`, which admits releases whose `import cultnet_py` fails), and `tests/test_declared_dependency_ranges.py` fails when a declared bound admits a tagged release lacking a name the package imports. The same check found `cultmesh-py` importing `cache.index_text`, which `cultcache-py` 0.2.0 lacks; its bound is `cultcache-py>=0.4.0` too (it has no changelog of its own). A raw record with bytes after its first value gives no schema hint to `resolve_document_for_raw_record`, so an unknown schema id stays `KeyError` rather than being routed; a registered one is refused by count when applied. - `apply_raw_document_record` and `apply_raw_snapshot` raise `SchemaConflictError` when a remote record is of another type than the one the cache holds at its key (`cultcache-py` holds one record per key whatever its type); the cache is unchanged, and the caller decides whether to drop the record or the peer. `cultmesh-py`'s server answers such a put with `message_invalid`; its client's `read_next_change` lets the error out. - `CultNetDatabaseChange` carries `commit_id` and `commit_count` (`commitId`/`commitCount`; None outside a multi-record commit, as C# writes nil): `from_wire` refuses an empty or non-string id and a count that is not an integer of at least 2, `to_wire` writes them when set, and the published `database_change_raw` schema lists both. Python applies no batch. The `simulation_fact` document's `shard`, `frame`, `subject` and `claim_kind` aliases are `roles` (many facts share each); the catalog bytes do not move. - Decodes the write receipt frames of `contracts/cultnet/interop/write-receipt.md`: `parse_message` validates `cultnet.write_admitted.v0` and `cultnet.error.v1` against their closed schemas (unknown keys, empty or non-string ids, an unknown operation or code are refused; `error.v1` requires `messageId`, string or null; `limit_reached` is `error.v1` only). Added `CULTNET_ERROR_V1_CODES`, `CULTNET_WRITE_RECEIPT_MESSAGE_VERSIONS`, `write_content_hash` and `hello_advertises_write_receipts`; `hello` takes `supported_message_versions`. `tests/test_write_receipt.py` replays the vectors' `contentHash`, `wire` and `schemaInvalid` sections. Not implemented here: the client write tracker (`scenarios`) and the server rules (`dedupe`, `blankFields`); this package has no receipt server and no write client. diff --git a/packages/cultnet-py/pyproject.toml b/packages/cultnet-py/pyproject.toml index 41c4e8f6..09db2733 100644 --- a/packages/cultnet-py/pyproject.toml +++ b/packages/cultnet-py/pyproject.toml @@ -16,7 +16,7 @@ classifiers = [ "Programming Language :: Python :: 3.11", "Programming Language :: Python :: 3.12", ] -dependencies = ["cultcache-py>=0.2.0", "msgpack>=1.0"] +dependencies = ["cultcache-py>=0.4.0", "msgpack>=1.0"] [project.optional-dependencies] crypto = ["cryptography>=42"] diff --git a/packages/cultnet-py/src/cultnet_py/messages.py b/packages/cultnet-py/src/cultnet_py/messages.py index d5e26b7a..c9b7d4a6 100644 --- a/packages/cultnet-py/src/cultnet_py/messages.py +++ b/packages/cultnet-py/src/cultnet_py/messages.py @@ -5,6 +5,7 @@ from dataclasses import dataclass from typing import Any import msgpack +from cultcache_py import read_payload_tree from .unicode_messageid_tables import DEFAULT_IGNORABLE, LETTER_OR_NUMBER @@ -544,7 +545,7 @@ def encode_witness_artifact_bundle_payload(bundle: dict[str, Any] | CultNetWitne def decode_witness_artifact_bundle_payload(payload: bytes) -> dict[str, Any]: - slots = msgpack.unpackb(payload, raw=False) + slots = read_payload_tree("cultnet.witness_artifact_bundle", payload) if not isinstance(slots, list) or len(slots) < 8: raise ValueError("CultNet witness artifact bundle payload must be an 8-slot MessagePack array") return { diff --git a/packages/cultnet-py/src/cultnet_py/replication.py b/packages/cultnet-py/src/cultnet_py/replication.py index 531c0012..28f5da33 100644 --- a/packages/cultnet-py/src/cultnet_py/replication.py +++ b/packages/cultnet-py/src/cultnet_py/replication.py @@ -171,9 +171,13 @@ def resolve_document_for_raw_record( def _infer_schema_version_from_payload(payload: bytes) -> str | None: + import msgpack # type: ignore try: - import msgpack # type: ignore decoded = msgpack.unpackb(payload, raw=False) + except msgpack.ExtraData: + # A payload with bytes after its value gives no routing hint: the record is unroutable by inference, and a + # registered schema id routes it to a definition whose decode refuses the tail by count. + return None except Exception: return None if isinstance(decoded, list) and decoded and isinstance(decoded[0], str): diff --git a/packages/cultnet-py/tests/test_cultnet.py b/packages/cultnet-py/tests/test_cultnet.py index 48972221..cdfaee5b 100644 --- a/packages/cultnet-py/tests/test_cultnet.py +++ b/packages/cultnet-py/tests/test_cultnet.py @@ -1971,6 +1971,34 @@ class CultNetTests(unittest.TestCase): apply_raw_document_record(cache, schema_document_map([first, second]), raw(first)) # the same type still replaces self.assertEqual(cache.get_required(first, "shared").name, "remote") + def test_a_raw_record_with_a_tail_is_unroutable_by_inference_and_refused_by_count_when_routed(self) -> None: + import msgpack # type: ignore + from cultcache_py import define_document_type + from cultnet_py import resolve_document_for_raw_record + + document = define_document_type("tests.tail_note", schema_id="tests.tail_note", schema_version="tests.tail_note.v1", + payload_encoder=msgpack.packb, payload_decoder=lambda payload: msgpack.unpackb(payload, raw=False)) + known = document.catalog_entry().schema_id + head = msgpack.packb(["tests.tail_note.v1", "x"], use_bin_type=True) + + def raw(schema_id: str, payload: bytes) -> dict: + return {"schemaId": schema_id, "recordKey": "k", "storedAt": "2026-06-13T00:00:00Z", + "payloadEncoding": "messagepack", "payload": payload} + + documents = schema_document_map([document]) + # Routing by inference reads a hint, not a payload: a payload with a tail gives no hint, so an unknown schema id + # stays the unknown-schema error (KeyError) even when the first value names a registered schema. + self.assertIs(resolve_document_for_raw_record(documents, "unknown", raw("unknown", head)), document) + with self.assertRaises(KeyError): + resolve_document_for_raw_record(documents, "unknown", raw("unknown", head + b"\xc0")) + # A registered schema id routes, and the tail is refused by count when the record is applied. + cache = CultCache() + cache.register_document_type(document) + with self.assertRaises(ValueError) as caught: + apply_raw_document_record(cache, documents, raw(known, head + b"\xc0")) + self.assertIn("1 bytes after its first MessagePack value", str(caught.exception)) + self.assertIsNone(cache.get(document, "k")) + def test_cultnet_replication_helpers_apply_raw_snapshot_and_shard_log(self) -> None: document = define_database_entry_type( "replica.item", @@ -2661,6 +2689,23 @@ class CultNetTests(unittest.TestCase): self.assertEqual(responses[0]["details"]["value"], schema_version) self.assertNotIn("messageId", responses[0]) + def test_witness_artifact_bundle_payload_is_one_value_and_depth_wins_over_a_tail(self) -> None: + import msgpack # type: ignore + from cultcache_py import CultCachePayloadDepthError + + payload = msgpack.packb(["b", "k", "t", {}, [], [], [], {}], use_bin_type=True) + decode_witness_artifact_bundle_payload(payload) + for tail, count in ((b"\xc0", 1), (b"\xc1\xc1", 2)): + with self.assertRaises(ValueError) as caught: + decode_witness_artifact_bundle_payload(payload + tail) + self.assertIs(type(caught.exception), ValueError) + self.assertEqual( + str(caught.exception), + f"cultnet.witness_artifact_bundle payload has {count} bytes after its first MessagePack value (refused on read).", + ) + with self.assertRaises(CultCachePayloadDepthError): + decode_witness_artifact_bundle_payload(b"\x91" * 300 + b"\xc0\xc0") + def test_cultnet_simulation_consensus_dedupes_witnesses_and_requires_quorum(self) -> None: from cultnet_py import CultNetSimulationConsensus diff --git a/packages/cultnet-py/tests/test_declared_dependency_ranges.py b/packages/cultnet-py/tests/test_declared_dependency_ranges.py new file mode 100644 index 00000000..9baf15fc --- /dev/null +++ b/packages/cultnet-py/tests/test_declared_dependency_ranges.py @@ -0,0 +1,94 @@ +"""A sibling package resolves through the workspace's editable install, so a package that starts importing a new name from +cultcache-py passes whatever range it declares. A consumer outside the workspace installs the newest release the range +admits. This is the Python form of scripts/test-declared-dependency-ranges.mjs: for every tagged cultcache-py release +(`cultcache-py-vX.Y.Z`) that the `cultcache-py>=` bound of cultnet-py or cultmesh-py admits, every name those packages +import from cultcache_py must be defined by that release's source. It needs git with the release tags and skips without +them. Only `>=` bounds are read; star imports and names reached by attribute access are not checked.""" + +from __future__ import annotations + +import ast +import re +import subprocess +import unittest +from pathlib import Path + +ROOT = Path(__file__).resolve().parents[3] +PACKAGES = ROOT / "packages" +CONSUMERS = ("cultnet-py", "cultmesh-py") + + +def _version(text: str) -> tuple[int, ...]: + return tuple(int(part) for part in text.split(".")) + + +def _git(*args: str) -> str | None: + try: + done = subprocess.run(["git", *args], cwd=ROOT, capture_output=True, text=True, check=True) + except (OSError, subprocess.CalledProcessError): + return None + return done.stdout + + +def _lower_bound(package: str) -> tuple[int, ...]: + manifest = (PACKAGES / package / "pyproject.toml").read_text(encoding="utf-8") + match = re.search(r'"cultcache-py>=(\d+\.\d+\.\d+)"', manifest) + assert match, f"{package} declares no cultcache-py>= bound" + return _version(match.group(1)) + + +def _imports(package: str) -> dict[str, set[str]]: + """module -> names imported from it, for every `from cultcache_py[.x] import names` under the package's src.""" + found: dict[str, set[str]] = {} + for source in (PACKAGES / package / "src").rglob("*.py"): + for node in ast.walk(ast.parse(source.read_text(encoding="utf-8-sig"))): + if isinstance(node, ast.ImportFrom) and node.level == 0 and node.module: + if node.module == "cultcache_py" or node.module.startswith("cultcache_py."): + found.setdefault(node.module, set()).update(alias.name for alias in node.names if alias.name != "*") + return found + + +def _defined(tag: str, module: str) -> set[str] | None: + base = "packages/cultcache-py/src/" + module.replace(".", "/") + for path in (base + ".py", base + "/__init__.py"): + text = _git("show", f"{tag}:{path}") + if text is None: + continue + names: set[str] = set() + for node in ast.parse(text.lstrip("\ufeff")).body: + if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)): + names.add(node.name) + elif isinstance(node, (ast.Import, ast.ImportFrom)): + names.update((alias.asname or alias.name).split(".")[0] for alias in node.names) + elif isinstance(node, (ast.Assign, ast.AnnAssign)): + for target in node.targets if isinstance(node, ast.Assign) else [node.target]: + if isinstance(target, ast.Name): + names.add(target.id) + return names + return None + + +class DeclaredDependencyRangeTests(unittest.TestCase): + def test_every_admitted_cultcache_release_defines_what_the_consumers_import(self) -> None: + listed = _git("tag", "--list", "cultcache-py-v*") + tags = sorted((t for t in (listed or "").split() if re.fullmatch(r"cultcache-py-v\d+\.\d+\.\d+", t)), + key=lambda t: _version(t.rsplit("v", 1)[1])) + if not tags: + self.skipTest("git with the cultcache-py release tags is not available") + current = re.search(r'^version = "(\d+\.\d+\.\d+)"', (PACKAGES / "cultcache-py" / "pyproject.toml").read_text(encoding="utf-8"), re.M) + assert current + for package in CONSUMERS: + bound = _lower_bound(package) + self.assertLessEqual(bound, _version(current.group(1)), f"{package}'s bound is above the workspace cultcache-py") + imports = _imports(package) + for tag in tags: + if _version(tag.rsplit("v", 1)[1]) < bound: + continue + for module, names in imports.items(): + defined = _defined(tag, module) + self.assertIsNotNone(defined, f"{package} imports {module}, which {tag} does not have") + self.assertEqual(sorted(names - defined), [], f"{package} imports names {tag} lacks from {module}") + + +if __name__ == "__main__": + unittest.main() From 363a8d6fe5b0e09330fe5a2f26bcdb80238e5dd5 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 05:43:55 +0000 Subject: [PATCH 06/12] Rust: a payload is one MessagePack value (bytes-after-rs) (#131) --- docs/cultcache-union-parity-map.md | 2 +- packages/cultcache-rs/CHANGELOG.md | 2 + packages/cultcache-rs/src/lib.rs | 137 +++++++++++++--- packages/cultcache-rs/tests/slot_readonly.rs | 52 +++++- packages/cultcache-rs/tests/union_parity.rs | 162 ++++++++++++++++++- packages/cultnet-rs/CHANGELOG.md | 1 + packages/cultnet-rs/tests/cultnet.rs | 10 +- packages/cultnet-rs/tests/decode_door.rs | 10 ++ 8 files changed, 346 insertions(+), 30 deletions(-) diff --git a/docs/cultcache-union-parity-map.md b/docs/cultcache-union-parity-map.md index 368be16a..1442abdc 100644 --- a/docs/cultcache-union-parity-map.md +++ b/docs/cultcache-union-parity-map.md @@ -259,7 +259,7 @@ landed in `ead5260c`, Python in `5618d8dd`, Rust in `c028c511`. TypeScript is no | Runtime | Owner of the decision (constant and walker) | Doors routed through it | |---|---|---| | C# | `CultPayloadDepth.Check` (`src/GameCult.Caching.MessagePack/CultPayloadDepth.cs`), limit `CultMessagePackSecurity.MaximumPayloadDepth`. MessagePack's own graph counter is set to the same value as a backstop only. | Write: `CultDocumentMessagePackSerialization.SerializeUntyped`. Read: `DeserializeUntyped`. The cache store (`SerializePayload`/`DeserializePayload`), the directory store, `CultMesh` and single-file documents all call those two; C# has no separate raw-envelope or prepared-batch door. | -| Rust | `cultcache_rs::check_payload_depth` and `MAX_PAYLOAD_DEPTH` (`packages/cultcache-rs/src/lib.rs`). | Read: `decode_payload` (get, list, `validate_envelope`/`put_envelope`, `finish_prepared_entry`). Write: `put`/`prepare_entry`/`prepare_entry_named` reach it through `finish_prepared_entry`, which decodes what it just encoded. Raw envelope: `put_raw_envelope`, the door CultMesh network puts use (`cultmesh-rs` `lib.rs`, two call sites). Prepared batch: `put_prepared_batch` via `require_payload_depth`. cultnet-rs: every free-form or skippable wire payload decodes through `cultcache_rs::decode_payload` (message decode in `contracts.rs`, `provider_session`, `media_stream_wire`, `replication`, `service_identity`). | +| Rust | `check_record_payload` (private; depth, then no bytes after the first value) over `first_value_end`, and `MAX_PAYLOAD_DEPTH`; `cultcache_rs::check_payload_depth` is the depth-only probe for outside callers (`packages/cultcache-rs/src/lib.rs`). | Read: `decode_payload` (get, list, `validate_envelope`/`put_envelope`, `finish_prepared_entry`). Write: `put`/`prepare_entry`/`prepare_entry_named` reach it through `finish_prepared_entry`, which decodes what it just encoded. Raw envelope: `put_raw_envelope`, the door CultMesh network puts use (`cultmesh-rs` `lib.rs`, two call sites). Prepared batch: `put_prepared_batch` via `require_record_payload`. Load: `pull_all_backing_stores` and `load_soa` run `require_record_payload` on every record, so a store holding a payload past the bound (or with bytes after it) is refused at load, and the cache then refuses every write and removal (`CultCacheLoadRefused`) until a load succeeds. cultnet-rs: every free-form or skippable wire payload decodes through `cultcache_rs::decode_payload` (message decode in `contracts.rs`, `provider_session`, `media_stream_wire`, `replication`, `service_identity`). | | Python | `_refuse_past_depth_limit` and `MAX_PAYLOAD_DEPTH` (`packages/cultcache-py/src/cultcache_py/documents.py`); a msgpack `StackError` on read maps to the same error. | `DocumentDefinition.encode_payload` (write) and `decode_payload` (read). `CultCache.put`, `put_envelope`, `put_envelopes` and load-time hydration go through them; there is no raw-envelope door. cultmesh-py round trips call the same pair. | - **Doors deliberately not routed.** cultnet-rs `idunn_lifecycle_brake`, `provider_health_contracts` and diff --git a/packages/cultcache-rs/CHANGELOG.md b/packages/cultcache-rs/CHANGELOG.md index 353facea..4e97588c 100644 --- a/packages/cultcache-rs/CHANGELOG.md +++ b/packages/cultcache-rs/CHANGELOG.md @@ -15,6 +15,8 @@ Semver: this is a breaking change in the 0.x line (0.3.x to 0.4.0). - `SchemaConflictError` gains the public field `unread_slots: Option`; code that builds it with a struct literal must set it (`None`). - A unit arm with slots (`[N, [extra...]]`) decodes as the unit arm and marks the record, where the derive refused it. - The registration map's value changes from the schema name to a private struct holding the schema name and the unread-slots check. +- A cache whose last `pull_all_backing_stores` was refused (a tail, a payload past the nesting bound, a duplicate unique-index value, a retype conflict, a record from the wrong store) refuses every `put`, `put_envelope`, `put_raw_envelope`, `put_prepared_batch`, `update` and `delete` with `CultCacheLoadRefused` until a pull succeeds. Before, it stayed writable over an empty image, and the next batch write replaced the store with only the new records. +- `decode_payload`, `put_raw_envelope` and prepared batches refuse a payload with bytes after its first MessagePack value (`payload has {n} bytes after its first MessagePack value`), by count; the tail is never walked or quoted. They used to read the first value and keep or drop the tail, so a record with a stray tail was served and rewritten without it. A store holding such a record (single-file, redb, `load_soa`) is refused at load, naming the type, the key and the count, and stays unreadable until the record is repaired; the same check now runs on every record at load, as the C# and Python runtimes do at open. The nesting bound is part of that check, so a stored payload nested past it is refused at load too, where it used to load and refuse on read. - A single-file store written as `cultcache.store.v3` is read and written: a store the C# runtime marked v3 (it holds a record with an element id: an owned list element with a `[CultElementId]` member) keeps its header here, and is carried through every rewrite with its ids; this runtime never drops the mark. A runtime from before this change refuses a v3 store by name. Rolling back to an older release after a store has been written as v3 is one-way: there is no carry-back, so restore from a copy taken before upgrading. - A write over a file the reader refuses (a zero-byte file, a truncated store, a stamp this runtime does not read) refuses with `CultCacheStoreUnreadable` and leaves the file as it is. A whole-store write used to replace such a file. This reverses 0.3.1, where `push_all` replaced a file with no readable stamp (healing it); the stamp rule now has one owner, the reader every write asks under the lock, and it accepts v1 and v3 (0.3.1 refused v3). - `#[cultcache(index)]` means unique, as in C# `RefuseIndexSharing`: within one type and one alias no two records hold one non-blank value. A put, batch, push or load that would make two records hold a value is refused, naming both keys and the alias, and a store that already holds a duplicate does not open. A grouping index is no longer expressible; a consumer that relied on shared values must drop the `index` and look up by another way. diff --git a/packages/cultcache-rs/src/lib.rs b/packages/cultcache-rs/src/lib.rs index 59ffa1af..522e226e 100644 --- a/packages/cultcache-rs/src/lib.rs +++ b/packages/cultcache-rs/src/lib.rs @@ -600,14 +600,22 @@ impl std::error::Error for PayloadTooDeep {} /// Walks the MessagePack headers of one payload without recursion, counting every array and map and /// refusing the first container past [`MAX_PAYLOAD_DEPTH`]. Scalars are skipped by length and nothing -/// is allocated per element. A payload that ends early is the decoder's to refuse. +/// is allocated per element. Only the first value is walked; a payload that ends early is the +/// decoder's to refuse. A depth-only probe: [`decode_payload`] also refuses bytes after the value. pub fn check_payload_depth(bytes: &[u8]) -> std::result::Result<(), PayloadTooDeep> { + first_value_end(bytes).map(|_| ()) +} + +/// Where the first MessagePack value of `bytes` ends, walking only that value (its tail is never +/// read). `None` when the payload ends inside a header; a final scalar cut short ends past the +/// payload's length, which is not bytes after the value either. Both are the decoder's to refuse. +fn first_value_end(bytes: &[u8]) -> std::result::Result, PayloadTooDeep> { let mut remaining = [0usize; MAX_PAYLOAD_DEPTH]; let mut depth = 0usize; let mut position = 0usize; while position < bytes.len() { let Some((advance, children)) = element_header(bytes, position) else { - return Ok(()); + return Ok(None); }; position = position.saturating_add(advance); if let Some(children) = children { @@ -629,7 +637,18 @@ pub fn check_payload_depth(bytes: &[u8]) -> std::result::Result<(), PayloadTooDe depth -= 1; } if depth == 0 { - return Ok(()); + return Ok(Some(position)); + } + } + Ok(None) +} + +/// The one rule for "is this payload one value": the nesting bound, then no bytes after the first +/// value. The count is all it reports; the tail is never walked or quoted. +fn check_record_payload(bytes: &[u8]) -> Result<()> { + if let Some(end) = first_value_end(bytes)? { + if end < bytes.len() { + bail!("payload has {} bytes after its first MessagePack value", bytes.len() - end); } } Ok(()) @@ -784,7 +803,7 @@ fn decode_grown(bytes: &[u8]) -> std::result::Re grown(&mut reader, |grown| T::deserialize(grown)) } -/// The one decode of a document payload: the nesting bound first, then [`decode_grown`], whose +/// The one decode of a document payload: the payload check first (nesting bound, one value), then [`decode_grown`], whose /// rmp-serde depth counter (1024) is only a backstop. A failure keeps its kind and never quotes the /// bytes. Readers of stored envelope payloads outside this crate decode through it too. /// @@ -796,7 +815,7 @@ fn decode_grown(bytes: &[u8]) -> std::result::Re /// verified on x86_64-pc-windows-msvc by hand (`decode_stack`, revision 10a39dac, 13 tests, debug and release); a /// `stacker` bump or a change to the stack-size tests repeats it. pub fn decode_payload(payload: &[u8]) -> Result { - check_payload_depth(payload)?; + check_record_payload(payload)?; decode_grown(payload).map_err(opaque_decode_error) } @@ -918,10 +937,11 @@ impl Drop for UnreadScope { } } -/// [`decode_payload`] with the mark: the same depth check and grown decode, inside a scope the derives' slot -/// visitors note into. A payload past the nesting limit is refused before the decode, never marked. +/// [`decode_payload`] with the mark: the same payload check and grown decode, inside a scope the derives' slot +/// visitors note into. A payload past the nesting limit or with bytes after its value is refused before the +/// decode, never marked. pub fn decode_payload_marked(payload: &[u8]) -> Result<(T, Option)> { - check_payload_depth(payload)?; + check_record_payload(payload)?; let scope = UnreadScope::open(); let value = decode_grown(payload).map_err(opaque_decode_error)?; Ok((value, scope.mark())) @@ -948,10 +968,10 @@ struct Definition { unread_slots: fn(&[u8]) -> Result>, } -/// The nesting bound for an envelope admitted without a typed decode, from the same walk that -/// [`decode_payload`] runs first. Names the entry's type and key, never the payload. -fn require_payload_depth(entry: &CultCacheEnvelope) -> Result<()> { - check_payload_depth(&entry.payload).with_context(|| { +/// The payload check for an envelope admitted without a typed decode, the one [`decode_payload`] +/// runs first. Names the entry's type and key, never the payload. +fn require_record_payload(entry: &CultCacheEnvelope) -> Result<()> { + check_record_payload(&entry.payload).with_context(|| { format!( "refused CultCache envelope {:?} at key {:?}", entry.r#type, entry.key @@ -2917,8 +2937,30 @@ pub struct CultCache { /// Records a store holds that no registered type claims: (store index, the envelope as the store read it). Derived from the last /// pull and replaced with it; written nowhere else. It decides only refusals at its keys. foreign: Vec<(usize, CultCacheEnvelope)>, + /// Why the last pull of the backing stores was refused; `None` once a pull succeeds. While set, every write and removal is refused. + load_refused: Option, } +/// A write or removal refused because the cache's last load of its backing stores was refused. The cache holds no +/// image of those stores, so any write would persist an image that lacks the records it never read. Repair the store +/// named by `reason` (which names a type, a key and a clause, never a payload) and load again. +#[derive(Clone, Debug, PartialEq, Eq)] +pub struct CultCacheLoadRefused { + pub reason: String, +} + +impl std::fmt::Display for CultCacheLoadRefused { + fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!( + formatter, + "CultCache refuses writes and removals: its load of the backing stores failed ({}); repair the store and load again", + self.reason + ) + } +} + +impl std::error::Error for CultCacheLoadRefused {} + /// A record a store holds that no registered type claims: carried, never served. #[derive(Clone, Debug, PartialEq, Eq)] pub struct ForeignRecord { @@ -2980,6 +3022,7 @@ impl CultCache { projectors: BTreeMap::new(), stores: Vec::new(), foreign: Vec::new(), + load_refused: None, } } @@ -3116,12 +3159,39 @@ impl CultCache { /// Reads every store into a staged image and replaces the cache's view /// only when all of them loaded, so a refused load admits nothing. pub fn pull_all_backing_stores(&mut self) -> Result<()> { + match self.stage_pull() { + Ok((loaded, foreign, holders)) => { + self.entries = loaded; + self.foreign = foreign; + self.index_holders = holders; + self.load_refused = None; + Ok(()) + } + Err(error) => { + self.load_refused = Some(format!("{error:#}")); + Err(error) + } + } + } + + /// The one gate of every write and removal: a cache whose load was refused holds no image of its stores. + fn require_loaded(&self) -> Result<()> { + match &self.load_refused { + Some(reason) => Err(CultCacheLoadRefused { reason: reason.clone() }.into()), + None => Ok(()), + } + } + + /// The image every store holds, judged whole, installed nowhere. + #[allow(clippy::type_complexity)] + fn stage_pull(&self) -> Result<(BTreeMap<(String, String), CultCacheEnvelope>, Vec<(usize, CultCacheEnvelope)>, IndexHolders)> { let mut loaded = BTreeMap::new(); let mut foreign = Vec::new(); let known_types: BTreeSet = self.definitions.keys().cloned().collect(); let schema_name_definitions = self.schema_name_definitions.clone(); for (index, registration) in self.stores.iter().enumerate() { for mut entry in registration.store.pull_all()? { + require_record_payload(&entry)?; let Some(canonical_type) = resolve_registered_type( &known_types, &schema_name_definitions, @@ -3154,10 +3224,7 @@ impl CultCache { } self.refuse_second_globals(&loaded)?; let holders = self.judge_image(&loaded)?; - self.entries = loaded; - self.foreign = foreign; - self.index_holders = holders; - Ok(()) + Ok((loaded, foreign, holders)) } /// The records the last pull found in a store that no registered type claims, sorted by key (UTF-16 order). The cache does not @@ -3383,6 +3450,7 @@ impl CultCache { } pub fn put(&mut self, key: impl Into, value: &T) -> Result { + self.require_loaded()?; let (entry, parsed) = self.prepare_entry(key, value)?; self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_second_global(&self.entries, &entry.r#type, &entry.key)?; @@ -3460,6 +3528,7 @@ impl CultCache { } pub fn put_prepared_batch(&mut self, staged: Vec) -> Result<()> { + self.require_loaded()?; if staged.is_empty() { return Err(anyhow!("CultCache batch must contain at least one entry")); } @@ -3484,7 +3553,7 @@ impl CultCache { entry.r#type )); } - require_payload_depth(entry)?; + require_record_payload(entry)?; self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_claimed_id(entry)?; let Some(home) = self.home_index(&entry.r#type)? else { @@ -3527,6 +3596,7 @@ impl CultCache { } pub fn put_envelope(&mut self, entry: CultCacheEnvelope) -> Result { + self.require_loaded()?; let parsed = self.validate_envelope::(&entry)?; self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_claimed_id(&entry)?; @@ -3547,6 +3617,7 @@ impl CultCache { /// admission; CultCache still enforces registered type, identity, timestamp, /// the payload nesting bound ([`MAX_PAYLOAD_DEPTH`]), routing, and persistence. pub fn put_raw_envelope(&mut self, entry: CultCacheEnvelope) -> Result<()> { + self.require_loaded()?; if !self.definitions.contains_key(&entry.r#type) { return Err(anyhow!( "No entry type registered for CultCache envelope type {:?}", @@ -3565,7 +3636,7 @@ impl CultCache { entry.r#type )); } - require_payload_depth(&entry)?; + require_record_payload(&entry)?; self.refuse_unread_slots(&entry.r#type, &entry.key)?; self.refuse_claimed_id(&entry)?; self.refuse_second_global(&self.entries, &entry.r#type, &entry.key)?; @@ -3621,6 +3692,7 @@ impl CultCache { } pub fn delete(&mut self, key: &str) -> Result { + self.require_loaded()?; self.require_entry_type::()?; self.refuse_unread_slots(T::TYPE, key)?; let id = entry_id_parts(T::TYPE, key); @@ -3662,6 +3734,7 @@ impl CultCache { entry.r#type )); } + require_record_payload(&entry)?; self.home_index(&entry.r#type)?; self.refuse_claimed_id(&entry)?; self.refuse_retype(&loaded, &self.foreign, &entry)?; @@ -6296,6 +6369,34 @@ mod tests { Ok(()) } + #[test] + fn put_raw_envelope_refuses_bytes_after_the_value() -> Result<()> { + let temp = tempfile::tempdir()?; + let store_path = temp.path().join("tail.cc"); + let mut cache = CultCache::new(); + cache.register_entry_type::()?; + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + let tailed = CultCacheEnvelope { + key: "app".to_string(), + r#type: Settings::TYPE.to_string(), + payload: vec![0x91, 0x01, 0xc0], + stored_at: now_utc_second(), + schema_id: None, + }; + let error = format!("{:#}", cache.put_raw_envelope(tailed.clone()).unwrap_err()); + assert!(error.contains("payload has 1 bytes after its first MessagePack value"), "{error}"); + assert!(error.contains("app") && error.contains(Settings::TYPE), "{error}"); + assert!(cache.snapshot().is_empty()); + assert!(!store_path.exists() || SingleFileMessagePackBackingStore::new(&store_path).pull_all()?.is_empty()); + let error = format!("{:#}", cache.put_prepared_batch(vec![tailed.clone()]).unwrap_err()); + assert!(error.contains("payload has 1 bytes after its first MessagePack value"), "{error}"); + assert!(cache.snapshot().is_empty()); + // The same bytes without the tail are admitted: the refusal is the count, not the shape. + cache.put_raw_envelope(CultCacheEnvelope { payload: vec![0x91, 0x01], ..tailed })?; + assert_eq!(cache.snapshot().len(), 1); + Ok(()) + } + #[test] fn soa_snapshot_round_trips_registered_entries() -> Result<()> { let temp = tempfile::tempdir()?; diff --git a/packages/cultcache-rs/tests/slot_readonly.rs b/packages/cultcache-rs/tests/slot_readonly.rs index 5bffcd9a..0031e372 100644 --- a/packages/cultcache-rs/tests/slot_readonly.rs +++ b/packages/cultcache-rs/tests/slot_readonly.rs @@ -481,9 +481,11 @@ fn load_soa_refuses_a_same_length_replacement_of_a_marked_record() { let (mut cache, _dir, _) = held_marked_and_clean(); let all = cache.snapshot_envelopes(); let marked = all.iter().find(|held| held.key == "marked").unwrap(); - // The clean record's bytes padded to the marked record's length: same length, different bytes. - let mut same_length = vec![0xc0; marked.payload.len()]; - same_length[..2].copy_from_slice(&[0x91, 0xa1]); + // One value (a bin16) of the marked record's length: same length, different bytes. A value followed by padding + // would be refused at admission as bytes after the value, before the read-only check is reached. + let length = marked.payload.len(); + let mut same_length = vec![0u8; length]; + same_length[..3].copy_from_slice(&[0xc5, ((length - 3) >> 8) as u8, (length - 3) as u8]); assert_eq!(same_length.len(), marked.payload.len()); let mut replaced = all.clone(); replaced.iter_mut().find(|held| held.key == "marked").unwrap().payload = same_length; @@ -491,3 +493,47 @@ fn load_soa_refuses_a_same_length_replacement_of_a_marked_record() { assert_refused_with(&error, "marked", &mark("Composition", 3, 2, 1)); assert_eq!(cache.snapshot_envelopes(), all); } + +// The payload check runs before the mark scope opens: a tail on a slot-bearing payload is refused by count, never marked. +#[test] +fn a_tail_on_a_slot_payload_is_refused_before_any_mark() { + let mut tailed = vector("extra-slot-record.msgpack"); + tailed.push(0xc0); + let error = decode_payload_marked::(&tailed).unwrap_err(); + assert!(format!("{error:#}").contains("payload has 1 bytes after its first MessagePack value"), "{error:#}"); + // Nothing leaked: the next marked decode finds exactly its own mark. + let found = decode_payload_marked::(&vector("extra-slot-record.msgpack")).unwrap().1.unwrap(); + assert_eq!(found.occurrences, 1); +} + +// A store holding marked records and one with a tail is refused at load; the cache then refuses every write with the load +// refusal (not a per-key mark conflict), and the store file is untouched. +#[test] +fn a_store_with_marked_records_and_a_tail_is_refused_and_its_cache_writes_nothing() { + use cultcache_rs::CacheBackingStore; + use cultcache_rs::CultCacheLoadRefused; + let (mut held, _dir, path) = held_marked_and_clean(); + let mut tailed = vector("union-surface.msgpack"); + tailed.push(0xc0); + SingleFileMessagePackBackingStore::new(&path).push(&envelope("tailed", tailed)).unwrap(); + let before = std::fs::read(&path).unwrap(); + // The cache that held the store re-reads it and is refused. + assert!(format!("{:#}", held.pull_all_backing_stores().unwrap_err()).contains("bytes after")); + let fresh_value = constructed(); + let mut fresh = registered(); + fresh.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + assert!(fresh.pull_all_backing_stores().is_err()); + for cache in [&mut held, &mut fresh] { + let staged = cache.prepare_entry("new", &fresh_value).unwrap().0; + let errors = [ + cache.put("new", &fresh_value).unwrap_err(), + cache.put_prepared_batch(vec![staged.clone()]).unwrap_err(), + cache.put_raw_envelope(staged).unwrap_err(), + cache.delete::("new").unwrap_err(), + ]; + for error in errors { + assert!(error.downcast_ref::().is_some(), "{error:#}"); + } + } + assert_eq!(std::fs::read(&path).unwrap(), before); +} diff --git a/packages/cultcache-rs/tests/union_parity.rs b/packages/cultcache-rs/tests/union_parity.rs index c51bd3d8..67f2b389 100644 --- a/packages/cultcache-rs/tests/union_parity.rs +++ b/packages/cultcache-rs/tests/union_parity.rs @@ -4,9 +4,13 @@ use cultcache_rs::CultCache; use cultcache_rs::CacheBackingStore; use cultcache_rs::CultCacheEnvelope; +use cultcache_rs::CultCacheLoadRefused; +use cultcache_rs::CultCacheSoaTable; use cultcache_rs::MAX_PAYLOAD_DEPTH; use cultcache_rs::PayloadTooDeep; use cultcache_rs::check_payload_depth; +use cultcache_rs::decode_payload; +use cultcache_rs::RedbMessagePackBackingStore; use cultcache_rs::DatabaseEntry; use cultcache_rs::SingleFileMessagePackBackingStore; @@ -464,7 +468,7 @@ fn envelope(key: &str, payload: &[u8]) -> CultCacheEnvelope { } // A payload past the bound cannot be admitted through the cache, so a store file written behind its -// back stands in for one a peer or an older runtime left. +// back stands in for one a peer or an older runtime left; load or read refuses it. fn through_cache(payload: &[u8]) -> Result { let temp = tempfile::tempdir().unwrap(); let path = temp.path().join("foreign.cc"); @@ -472,7 +476,10 @@ fn through_cache(payload: &[u8]) -> Result { let mut cache = CultCache::new(); cache.register_entry_type::().unwrap(); cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); - cache.pull_all_backing_stores().unwrap(); + // The payload check runs at load, so a payload past the bound is refused there, before any read. + if let Err(error) = cache.pull_all_backing_stores() { + return Err(format!("{error:#}")); + } match cache.get::("k") { Ok(value) => Ok(value.expect("the envelope was stored")), Err(error) => Err(format!("{error:#}")), @@ -529,6 +536,157 @@ fn bounds_vectors() { assert_eq!((accepted, refused, read_only), (4, 7, 1)); } +// A payload is one MessagePack value (bytes-after-outcomes.msgpack): every door refuses a tail by +// its count, names the clause, never the depth limit (the tail is not walked) and never a value. +#[test] +fn bytes_after_vectors() { + let rows: Vec = rmp_serde::from_slice(&vector("bytes-after-outcomes.msgpack")).unwrap(); + assert_eq!(rows.len(), 4); + let surface = vector("union-surface.msgpack"); + for row in &rows { + let fields = row.as_array().unwrap(); + let (file, outcome, union, detail) = (fields[0].as_str().unwrap(), fields[1].as_str().unwrap(), fields[2].as_str().unwrap(), fields[4].as_str().unwrap()); + assert_eq!((outcome, union, detail), ("refused", "UnionSurface", "bytes-after"), "{file}"); + let bytes = vector(file); + assert_eq!(&bytes[..surface.len()], &surface[..], "{file}"); + let clause = format!("payload has {} bytes after its first MessagePack value", bytes.len() - surface.len()); + let check = |door: &str, message: &str| { + assert!(message.contains(&clause), "{file} {door}: {message}"); + assert!(!message.contains("the limit is"), "{file} {door} walked the tail: {message}"); + assert!(!message.contains("CANARY-bytes-after-7d1e"), "{file} {door} echoed the tail: {message}"); + assert!(!message.contains(CANARY), "{file} {door} echoed a value: {message}"); + }; + let error = decode_payload::(&bytes).unwrap_err(); + assert!(error.downcast_ref::().is_none(), "{file}"); + check("decode_payload", &format!("{error:#}")); + // A store written behind the cache's back is refused at load, naming the type, the key and the count. + let temp = tempfile::tempdir().unwrap(); + let path = temp.path().join("foreign.cc"); + SingleFileMessagePackBackingStore::new(&path).push(&envelope("k", &bytes)).unwrap(); + let mut single = CultCache::new(); + single.register_entry_type::().unwrap(); + single.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + let message = format!("{:#}", single.pull_all_backing_stores().unwrap_err()); + check("single-file load", &message); + assert!(message.contains("\"k\"") && message.contains(UnionSurface::TYPE), "{file}: {message}"); + let redb_path = temp.path().join("foreign.redb"); + RedbMessagePackBackingStore::new(&redb_path).unwrap().push(&envelope("k", &bytes)).unwrap(); + let mut redb = CultCache::new(); + redb.register_entry_type::().unwrap(); + redb.add_generic_backing_store(RedbMessagePackBackingStore::new(&redb_path).unwrap()).unwrap(); + let message = format!("{:#}", redb.pull_all_backing_stores().unwrap_err()); + check("redb load", &message); + assert!(message.contains("\"k\"") && message.contains(UnionSurface::TYPE), "{file}: {message}"); + let mut soa = CultCache::new(); + soa.register_entry_type::().unwrap(); + let message = format!("{:#}", soa.load_soa(CultCacheSoaTable::from_envelopes(vec![envelope("k", &bytes)])).unwrap_err()); + check("load_soa", &message); + assert!(message.contains("\"k\"") && message.contains(UnionSurface::TYPE), "{file}: {message}"); + assert!(soa.snapshot().is_empty(), "{file}"); + // Raw admission refuses and stages nothing. + let mut raw = CultCache::new(); + raw.register_entry_type::().unwrap(); + let message = format!("{:#}", raw.put_raw_envelope(envelope("k", &bytes)).unwrap_err()); + check("put_raw_envelope", &message); + assert!(raw.snapshot().is_empty(), "{file}"); + } +} + +// Depth is found inside the first value, so a tail never hides it: a payload nested past the bound +// with a tail after it is the depth refusal, not bytes after. +#[test] +fn depth_beats_a_tail() { + let mut deep = vec![0x91u8; MAX_PAYLOAD_DEPTH + 1]; + deep.push(0xc0); + deep.push(0xc0); + let error = decode_payload::(&deep).unwrap_err(); + assert_eq!(error.downcast_ref::(), Some(&TOO_DEEP), "{error:#}"); +} + +// Depth is refused where the payload is loaded: a store holding one nested past the bound is not read. +#[test] +fn depth_is_refused_at_load() { + let mut deep = vec![0x91u8; MAX_PAYLOAD_DEPTH + 1]; + deep.push(0xc0); + let temp = tempfile::tempdir().unwrap(); + let path = temp.path().join("deep.cc"); + SingleFileMessagePackBackingStore::new(&path).push(&envelope("k", &deep)).unwrap(); + let mut cache = CultCache::new(); + cache.register_entry_type::().unwrap(); + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + let error = cache.pull_all_backing_stores().unwrap_err(); + assert_eq!(error.downcast_ref::(), Some(&TOO_DEEP), "{error:#}"); + assert!(cache.snapshot().is_empty()); +} + +// A cache whose load was refused refuses every write and removal and leaves the store as it was: a +// later write would otherwise persist the empty image over the records it never read. +#[test] +fn a_refused_load_refuses_every_write() { + let tail = { + let mut bytes = vector("union-surface.msgpack"); + bytes.push(0xc0); + bytes + }; + let good = envelope("good", &vector("union-surface.msgpack")); + let bad = envelope("bad", &tail); + for redb in [false, true] { + let temp = tempfile::tempdir().unwrap(); + let path = temp.path().join(if redb { "seed.redb" } else { "seed.cc" }); + let mut cache = CultCache::new(); + cache.register_entry_type::().unwrap(); + if redb { + let mut store = RedbMessagePackBackingStore::new(&path).unwrap(); + store.push(&good).unwrap(); + store.push(&bad).unwrap(); + drop(store); + cache.add_generic_backing_store(RedbMessagePackBackingStore::new(&path).unwrap()).unwrap(); + } else { + let mut store = SingleFileMessagePackBackingStore::new(&path); + store.push(&good).unwrap(); + store.push(&bad).unwrap(); + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + } + let load = cache.pull_all_backing_stores().unwrap_err(); + assert!(format!("{load:#}").contains("bytes after"), "{load:#}"); + let before = std::fs::read(&path).unwrap(); + let value = constructed(); + let staged = cache.prepare_entry("new", &value).unwrap().0; + let results: Vec<(&str, anyhow::Error)> = vec![ + ("put", cache.put("new", &value).unwrap_err()), + ("put_prepared_batch", cache.put_prepared_batch(vec![staged.clone()]).unwrap_err()), + ("put_envelope", cache.put_envelope::(staged.clone()).unwrap_err()), + ("put_raw_envelope", cache.put_raw_envelope(staged.clone()).unwrap_err()), + ("update", cache.update::("good", |current| current.unwrap_or_else(constructed)).unwrap_err()), + ("delete", cache.delete::("good").unwrap_err()), + ("delete a key the image lacks", cache.delete::("absent").unwrap_err()), + ]; + for (door, error) in results { + let refusal = error.downcast_ref::().unwrap_or_else(|| panic!("{door} redb={redb}: {error:#}")); + assert!(refusal.to_string().contains("bytes after"), "{door}: {refusal}"); + } + assert_eq!(std::fs::read(&path).unwrap(), before, "redb={redb}: the store changed"); + assert!(cache.snapshot().is_empty()); + // Repairing the store and loading again lifts it. + if redb { + RedbMessagePackBackingStore::new(&path).unwrap().delete(&bad).unwrap(); + } else { + SingleFileMessagePackBackingStore::new(&path).delete(&bad).unwrap(); + } + cache.pull_all_backing_stores().unwrap(); + cache.put("new", &value).unwrap(); + } +} + +// A payload that ends early is not bytes after the value: the decoder refuses it, not the check. +#[test] +fn a_truncated_payload_is_the_decoders_not_bytes_after() { + let mut truncated = vec![0x91, 0xd9, 0x40]; + truncated.extend_from_slice(b"short"); + let message = format!("{:#}", decode_payload::(&truncated).unwrap_err()); + assert!(!message.contains("bytes after"), "{message}"); +} + fn wrap_chain(wraps: usize, leaf: Node) -> UnionSurface { let mut node = leaf; for _ in 0..wraps { diff --git a/packages/cultnet-rs/CHANGELOG.md b/packages/cultnet-rs/CHANGELOG.md index 384395ec..1a01dbe1 100644 --- a/packages/cultnet-rs/CHANGELOG.md +++ b/packages/cultnet-rs/CHANGELOG.md @@ -15,5 +15,6 @@ only in git history. ### Breaking +- A CultNet frame with bytes after its message is refused, by count (through `cultcache_rs::decode_payload`); the decoder used to read the message and drop the tail. No producer emits one. Behaviour change: the next release is a minor bump (0.x breaking). - `CultNetDocumentRegistry::register` returns `Result<&mut Self>` where it returned `&mut Self`. A binding that claims a document type or a schema id another binding already holds is refused with `cultcache_rs::SchemaConflictError` (naming the new binding's schema id and both document types) and nothing is bound; binding a type to the schema id it already holds still replaces that binding. Callers chain `register(...)?`. - `CultMeshCdnChunkRef` gains a public `id` field, the decimal text of `offset` (the reference's element id, slot 4). A manifest written before ids reads `""`; encoding writes the slot. diff --git a/packages/cultnet-rs/tests/cultnet.rs b/packages/cultnet-rs/tests/cultnet.rs index 64c57aaf..7bc84957 100644 --- a/packages/cultnet-rs/tests/cultnet.rs +++ b/packages/cultnet-rs/tests/cultnet.rs @@ -1810,12 +1810,10 @@ fn a_replicated_payload_nested_past_the_bound_is_refused_typed_on_apply() -> Res }; assert_eq!(documents.len(), 1); - backing.push(&envelope("past", limit + 1))?; - let mut cache = CultCache::new(); - cache.register_entry_type::()?; - cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store))?; - cache.pull_all_backing_stores()?; - let past = registry.create_raw_snapshot_response(&cache, "past", None, Some(&["past".to_string()]))?; + // The sender cannot load a payload past the bound, so the frame is the at-limit one with its bytes swapped. + let mut past = at_limit.clone(); + let CultNetMessage::SnapshotResponseRaw { documents, .. } = &mut past else { unreachable!() }; + documents[0].payload = envelope("past", limit + 1).payload; let mut target = CultCache::new(); target.register_entry_type::()?; let error = registry diff --git a/packages/cultnet-rs/tests/decode_door.rs b/packages/cultnet-rs/tests/decode_door.rs index fd5bab6a..c55a6404 100644 --- a/packages/cultnet-rs/tests/decode_door.rs +++ b/packages/cultnet-rs/tests/decode_door.rs @@ -190,6 +190,16 @@ fn message_past_the_bound_is_refused_typed() { }); } +// A frame is one message: cultcache's payload rule refuses bytes after it, by count. +#[test] +fn frame_with_bytes_after_is_refused() { + let mut bytes = hello_with(&nested(0)); + assert!(decode_message(bytes.clone(), BIG).is_ok()); + bytes.push(0xc0); + let error = format!("{:#}", decode_message(bytes, BIG).unwrap_err()); + assert!(error.contains("1 bytes after its first MessagePack value"), "{error}"); +} + fn raw_v1_bytes(header_unknown: Option<&[u8]>) -> Vec { let mut header: Vec<(rmpv::Value, rmpv::Value)> = vec![ ("schemaId".into(), "s".into()), From 66bea107d5ce61e2871b3ca3dedd4d9c5cc6ec51 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 06:02:51 +0000 Subject: [PATCH 07/12] TypeScript: a payload is one MessagePack value (bytes-after-ts) (#130) --- packages/cultcache-ts/CHANGELOG.md | 2 + packages/cultcache-ts/README.md | 6 +- .../cultcache-ts/src/cult-cache-inspector.ts | 18 ++- packages/cultcache-ts/src/document.ts | 8 +- packages/cultcache-ts/src/payload-depth.ts | 39 ++++-- packages/cultcache-ts/src/wire.ts | 5 +- .../cultcache-ts/test/payload-depth.test.ts | 20 ++++ .../cultcache-ts/test/union-parity.test.ts | 113 +++++++++++++++++- .../test/provider-session-wire.test.ts | 13 ++ packages/cultnet-ts/test/content.test.ts | 7 ++ packages/cultnet-ts/test/cultnet.test.ts | 23 ++++ 11 files changed, 229 insertions(+), 25 deletions(-) diff --git a/packages/cultcache-ts/CHANGELOG.md b/packages/cultcache-ts/CHANGELOG.md index b91a0535..2da49fe9 100644 --- a/packages/cultcache-ts/CHANGELOG.md +++ b/packages/cultcache-ts/CHANGELOG.md @@ -24,10 +24,12 @@ All notable changes to `@gamecult/cultcache-ts` are recorded here, following ### Added +- `checkPayloadDepth` returns the offset where the first value ends, or `undefined` when the payload ends early (a header, a container or a str/bin/scalar body cut short; it was `void`); it still refuses only depth. The bytes-after refusal is internal (`checkPayloadIsOneValue`, not exported); no new public name. - `CultCache.foreignRecords()`: the `{key, schemaId, schemaName, storedAt}` of every record the last pull carried and no registered type claims, sorted by key ordinal. ### Changed +- A document payload is exactly one MessagePack value (cultlib-gaps:ruling:trailing-bytes-refuse). `decodeDocumentPayload`, the one reader (store loads, `put`/`putEnvelope`, CultNet replication, the provider command), and the inspector refuse bytes after the first value with `document type "T" payload has N bytes after its first MessagePack value.`, counted from the walk's end and never walked or echoed; a payload that ends early stays the decoder's to refuse. The refused set is unchanged; only the text changed (it was `@msgpack`'s `RangeError` "Extra N of M byte(s)" for a plain document and the codec reader's "bytes after its value" for a codec document, and the inspector's preview no longer shows a tail), so code matching either text must match this one. `readWire` reads one value and no longer checks what follows. Stated limit: a definition with its own `formatter` is not walked, so whether its payload is one value is the formatter's. Implied bump: 0.16.0 (pre-1.0 minor, with the Breaking entries above). - Opening a legacy (pre-v1) store is a pure read: payloads that were not bytes are normalized in memory and the file is left as it was, so a codec that refuses the content cannot lose the old bytes. The next push, delete or flush rewrites the file as v1, merging what the open read. diff --git a/packages/cultcache-ts/README.md b/packages/cultcache-ts/README.md index 6fea306a..a081cd92 100644 --- a/packages/cultcache-ts/README.md +++ b/packages/cultcache-ts/README.md @@ -49,7 +49,9 @@ through `decodeDocumentPayload`, and when you need the bytes themselves (hashing write) take them with `documentPayloadBytes`, the sanctioned escape. - `decodeDocumentPayload(definition, payload)` reads a payload through the definition's formatter, - codec or plain map and then its schema. The definition is required. + codec or plain map and then its schema. The definition is required. A payload is exactly one + MessagePack value: bytes after the first are refused by count, naming the type, and never walked or + echoed (a payload that ends early is the decoder's to refuse). - `encodeDocumentPayload(definition, value)` is the writer: the schema first, then the formatter, codec or plain map. - `documentPayloadFromBytes(bytes)` is where bytes enter (a record you validated, a store read, a @@ -327,7 +329,7 @@ If you do not supply one, `CultCacheTS` uses MessagePack plus Zod validation by The 256-container nesting bound (`MAX_PAYLOAD_DEPTH`, refused with `CultCachePayloadDepthError`) covers the MessagePack the library itself writes: the default format and a codec. A payload your formatter writes is opaque bytes in any format, so the -library does not walk it as MessagePack; its format and its bounds are your formatter's. +library does not walk it as MessagePack; its format, its bounds and whether it is one value are your formatter's. Zod is no longer mandatory. Any schema object with a `parse(...)` function will do, which means generated Ajv-backed JSON-schema contracts can feed the cache diff --git a/packages/cultcache-ts/src/cult-cache-inspector.ts b/packages/cultcache-ts/src/cult-cache-inspector.ts index 1c359157..71643913 100644 --- a/packages/cultcache-ts/src/cult-cache-inspector.ts +++ b/packages/cultcache-ts/src/cult-cache-inspector.ts @@ -1,6 +1,6 @@ import { decode } from "@msgpack/msgpack"; -import { checkPayloadDepth } from "./payload-depth"; +import { checkPayloadDepth, checkPayloadIsOneValue } from "./payload-depth"; import { type StoreFormat, isStoreSnapshot, refuseUnreadStoreBytes, requireStoreSlots, requireV1RecordSlots } from "./store-format"; export interface CultCacheInspection { @@ -235,12 +235,22 @@ function decodeLegacyEnvelope(value: unknown): InspectedRecord { }; } -// The inspector's one payload read: the depth walk first, so a payload past the limit is refused before any decoder sees it. +// The inspector's one payload read: the depth walk first, so a payload past the limit is refused before any decoder sees it, +// and bytes after the first value are refused with the shared clause (the tail is never decoded or shown). function decodeBoundedPayload(payload: Uint8Array): unknown { - checkPayloadDepth(payload); + checkPayloadIsOneValue(payload, "document payload"); return decode(payload); } +// Where the first value ends (the whole payload when it ends early or cannot be walked): a preview never shows a tail. +function valueEnd(payload: Uint8Array): number { + try { + return checkPayloadDepth(payload) ?? payload.length; + } catch { + return payload.length; + } +} + function previewPayload(payload: Uint8Array): Pick { try { return { payloadPreview: toJsonSafe(decodeBoundedPayload(payload)) }; @@ -248,7 +258,7 @@ function previewPayload(payload: Uint8Array): Pick value.toString(16).padStart(2, "0")) .join(" "), }, diff --git a/packages/cultcache-ts/src/document.ts b/packages/cultcache-ts/src/document.ts index 8d295bd4..cb1ef589 100644 --- a/packages/cultcache-ts/src/document.ts +++ b/packages/cultcache-ts/src/document.ts @@ -1,6 +1,6 @@ import type { NotedSlots } from "./codec"; import { decodeMarked, resolveDocumentFormatter } from "./document-formatter"; -import { checkPayloadDepth } from "./payload-depth"; +import { checkPayloadDepth, checkPayloadIsOneValue } from "./payload-depth"; import type { AnyCultCacheDocumentDefinition, CultCacheSchemaCatalogMember, @@ -129,7 +129,11 @@ export function decodeDocumentPayloadMarked; unreadSlots?: NotedSlots } { const bytes = documentPayloadBytes(payload); - if (writesLibraryMessagePack(definition)) checkPayloadDepth(bytes, `document type "${definition.type}" payload`); + if (writesLibraryMessagePack(definition)) { + // A payload is exactly one MessagePack value: bytes after the first are refused by count before any formatter reads, + // so a tail is refused before any slot is marked. + checkPayloadIsOneValue(bytes, `document type "${definition.type}" payload`); + } const { value: decoded, unreadSlots } = decodeMarked(definition, bytes); const value = definition.schema.parse(decoded) as CultCacheDocumentValue; return unreadSlots ? { value, unreadSlots } : { value }; diff --git a/packages/cultcache-ts/src/payload-depth.ts b/packages/cultcache-ts/src/payload-depth.ts index 40ca7afe..45b8d5ca 100644 --- a/packages/cultcache-ts/src/payload-depth.ts +++ b/packages/cultcache-ts/src/payload-depth.ts @@ -12,8 +12,9 @@ export class CultCachePayloadDepthError extends Error { // Walks the headers of one MessagePack value iteratively (a count stack bounded by the limit, scalars skipped by length, // nothing read) and throws at the first container past the limit. The message names the subject, the depth and the limit, -// never payload bytes. A payload that ends early is the reader's to refuse. -export function checkPayloadDepth(bytes: Uint8Array, subject = "document payload"): void { +// never payload bytes. Returns the offset where the first value ends, or undefined when the payload ends early (the +// reader's to refuse); what follows that offset is never walked, and whether it may exist is the caller's rule. +export function checkPayloadDepth(bytes: Uint8Array, subject = "document payload"): number | undefined { const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength); const remaining = new Float64Array(MAX_PAYLOAD_DEPTH); let depth = 0; @@ -33,12 +34,12 @@ export function checkPayloadDepth(bytes: Uint8Array, subject = "document payload else if (code <= 0x8f) { container = true; header = 1; children = 2 * (code & 0x0f); } else if (code <= 0x9f) { container = true; header = 1; children = code & 0x0f; } else if (code <= 0xbf) skip = 1 + (code & 0x1f); - else if (code === 0xc4 || code === 0xd9) { const n = width(1); if (n === undefined) return; skip = 2 + n; } - else if (code === 0xc5 || code === 0xda) { const n = width(2); if (n === undefined) return; skip = 3 + n; } - else if (code === 0xc6 || code === 0xdb) { const n = width(4); if (n === undefined) return; skip = 5 + n; } - else if (code === 0xc7) { const n = width(1); if (n === undefined) return; skip = 3 + n; } - else if (code === 0xc8) { const n = width(2); if (n === undefined) return; skip = 4 + n; } - else if (code === 0xc9) { const n = width(4); if (n === undefined) return; skip = 6 + n; } + else if (code === 0xc4 || code === 0xd9) { const n = width(1); if (n === undefined) return undefined; skip = 2 + n; } + else if (code === 0xc5 || code === 0xda) { const n = width(2); if (n === undefined) return undefined; skip = 3 + n; } + else if (code === 0xc6 || code === 0xdb) { const n = width(4); if (n === undefined) return undefined; skip = 5 + n; } + else if (code === 0xc7) { const n = width(1); if (n === undefined) return undefined; skip = 3 + n; } + else if (code === 0xc8) { const n = width(2); if (n === undefined) return undefined; skip = 4 + n; } + else if (code === 0xc9) { const n = width(4); if (n === undefined) return undefined; skip = 6 + n; } else if (code === 0xca || code === 0xce || code === 0xd2) skip = 5; else if (code === 0xcb || code === 0xcf || code === 0xd3) skip = 9; else if (code === 0xcc || code === 0xd0) skip = 2; @@ -47,10 +48,10 @@ export function checkPayloadDepth(bytes: Uint8Array, subject = "document payload else if (code === 0xd6) skip = 6; else if (code === 0xd7) skip = 10; else if (code === 0xd8) skip = 18; - else if (code === 0xdc) { const n = width(2); if (n === undefined) return; container = true; header = 3; children = n; } - else if (code === 0xdd) { const n = width(4); if (n === undefined) return; container = true; header = 5; children = n; } - else if (code === 0xde) { const n = width(2); if (n === undefined) return; container = true; header = 3; children = 2 * n; } - else { const n = width(4); if (n === undefined) return; container = true; header = 5; children = 2 * n; } + else if (code === 0xdc) { const n = width(2); if (n === undefined) return undefined; container = true; header = 3; children = n; } + else if (code === 0xdd) { const n = width(4); if (n === undefined) return undefined; container = true; header = 5; children = n; } + else if (code === 0xde) { const n = width(2); if (n === undefined) return undefined; container = true; header = 3; children = 2 * n; } + else { const n = width(4); if (n === undefined) return undefined; container = true; header = 5; children = 2 * n; } if (container) { if (depth === MAX_PAYLOAD_DEPTH) throw new CultCachePayloadDepthError(MAX_PAYLOAD_DEPTH + 1, MAX_PAYLOAD_DEPTH, subject); @@ -63,8 +64,20 @@ export function checkPayloadDepth(bytes: Uint8Array, subject = "document payload position += skip; } + // A scalar whose declared length runs past the payload is an early end, not a value. + if (position > bytes.length) return undefined; // One element is complete; close every container it was the last element of. while (depth > 0 && --remaining[depth - 1] === 0) depth--; - if (depth === 0) return; + if (depth === 0) return position; } + return undefined; +} + +// The one rule that a payload is exactly one MessagePack value: the depth walk, then a refusal of bytes after the first +// value, by count, naming the subject, from the walk's end. The tail is never walked or echoed; a payload that ends early +// stays the decoder's to refuse. decodeDocumentPayload and the inspector speak this clause and no other. +export function checkPayloadIsOneValue(bytes: Uint8Array, subject: string): number | undefined { + const end = checkPayloadDepth(bytes, subject); + if (end !== undefined && end < bytes.length) throw new Error(`${subject} has ${bytes.length - end} bytes after its first MessagePack value.`); + return end; } diff --git a/packages/cultcache-ts/src/wire.ts b/packages/cultcache-ts/src/wire.ts index 12bf572a..d351c446 100644 --- a/packages/cultcache-ts/src/wire.ts +++ b/packages/cultcache-ts/src/wire.ts @@ -128,7 +128,6 @@ export function readWire(payload: Uint8Array): unknown { } } - const tree = next(); - if (offset !== payload.length) throw new Error("the MessagePack payload has bytes after its value."); - return tree; + // One value is read; the caller owns what follows it (decodeDocumentPayload refuses a payload with a tail). + return next(); } diff --git a/packages/cultcache-ts/test/payload-depth.test.ts b/packages/cultcache-ts/test/payload-depth.test.ts index 382f0672..6fb81265 100644 --- a/packages/cultcache-ts/test/payload-depth.test.ts +++ b/packages/cultcache-ts/test/payload-depth.test.ts @@ -119,6 +119,19 @@ test("a payload that ends early, is empty or carries trailing bytes is the reade assert.doesNotThrow(() => checkPayloadDepth(bytes(0xc0))); }); +test("the walk returns where the first value ends, and undefined when the payload ends early", () => { + assert.equal(checkPayloadDepth(bytes(0xc0)), 1); + assert.equal(checkPayloadDepth(bytes(0xc0, 0xc1, 0xc1)), 1, "a top-level scalar ends at its own length; the tail is not judged"); + assert.equal(checkPayloadDepth(bytes(0x92, 0x01, 0xa1, 0x61, 0xc0)), 5 - 1, "a container ends when its last child completes"); + assert.equal(checkPayloadDepth(cat(bytes(0x90), arrays(L + 50))), 1, "the tail is never walked"); + assert.equal(checkPayloadDepth(bytes()), undefined); + assert.equal(checkPayloadDepth(bytes(0x92, 0x01)), undefined); + assert.equal(checkPayloadDepth(bytes(0xd9)), undefined); + for (const cut of [[0xcc], [0xcb, 0x3f, 0xf8], [0xa5, 0x61, 0x62, 0x63], [0xc4, 0x02, 0x01], [0xd7, 0x01]]) { + assert.equal(checkPayloadDepth(bytes(...cut)), undefined, `a scalar cut short: ${cut.map((value) => value.toString(16)).join(" ")}`); + } +}); + test("the walk is iterative: 100000 nested arrays are refused with the depth error, not a RangeError", () => { refused(arrays(100_000), "100000 nested arrays"); refused(chain(bytes(0x81, 0x00), 100_000), "100000 nested maps"); @@ -198,6 +211,13 @@ test("the inspector previews a payload at the limit and reports one past it, or } }); +test("the inspector refuses bytes after a payload's first value with the shared clause and previews no tail", async () => { + const file = await storeWith(bytes(0xc0, 0xc1, 0xc1)); + const record = inspectCultCacheBytes(file, await readFile(file)).records[0]!; + assert.equal(record.payloadDecodeError, "document payload has 2 bytes after its first MessagePack value."); + assert.deepEqual(record.payloadPreview, { bytes: 3, hexPrefix: "c0" }); +}); + // A header whose length or count needs all four bytes: the low sixteen bits of the length are small, the whole is not. const be32 = (n: number) => bytes((n >>> 24) & 0xff, (n >>> 16) & 0xff, (n >>> 8) & 0xff, n & 0xff); diff --git a/packages/cultcache-ts/test/union-parity.test.ts b/packages/cultcache-ts/test/union-parity.test.ts index d688cb74..d917ea7a 100644 --- a/packages/cultcache-ts/test/union-parity.test.ts +++ b/packages/cultcache-ts/test/union-parity.test.ts @@ -338,7 +338,7 @@ test("the wire reader keeps 64-bit integers whole, tells a float from an integer assert.throws(() => readWire(fromHex("dd")), /truncated/u); assert.throws(() => readWire(fromHex("ddffffffff")), /truncated/u, "a huge array header reserves nothing"); assert.throws(() => readWire(fromHex("a5616263")), /truncated/u); - assert.throws(() => readWire(fromHex("0102")), /bytes after its value/u); + assert.equal(readWire(fromHex("0102")), 1, "the reader reads one value; the document owner refuses what follows"); assert.throws(() => readWire(fromHex("81a161a162")), /map, extension or reserved/u); assert.throws(() => readWire(fromHex("c1")), /map, extension or reserved/u); assert.throws(() => readWire(fromHex("d4010a")), /map, extension or reserved/u); @@ -725,3 +725,114 @@ test("a cache load refuses a store whose record is past the limit, naming the ty const refusing = CultCache.builder().withDocumentType(document).withGenericStore(memoryStore([envelopeOf(await read("depth-past-limit.msgpack"))])).build(); await assert.rejects(() => refusing.pullAllBackingStores(), (error: Error) => error instanceof CultCachePayloadDepthError && !error.message.includes(CANARY)); }); + +// ---- bytes-after vectors: a payload is exactly one MessagePack value (tests/vectors/union-parity/README.md) ---- + +// The C# reference writes the vectors: union-surface.msgpack followed by a tail the first value does not own. The one +// owner refuses by count, naming the type, from the walk's end; it never walks or echoes the tail. +async function bytesAfterEntries() { + const raw = decode(await read("bytes-after-outcomes.msgpack")) as unknown[][]; + return raw.map(([file, outcome, unionName, key, detail]) => ({ file: file as string, outcome: outcome as string, unionName: unionName as string, key: key as number | null, detail: detail as string })); +} + +// The C# vector writer's canary (BytesAfterVectorTests): it sits in the deep tail, so an echo of the tail would carry it. +const TAIL_CANARY = "CANARY-bytes-after-7d1e"; +const valueLength = async () => (await read("union-surface.msgpack")).length; +const plainDocument = () => defineDocumentType({ type: "tests.plain_surface", schema: z.unknown() }); +const codecDocument = () => defineDocumentType({ type: "tests.codec_surface", schema: surfaceSchema, codec: surfaceCodec }); + +async function assertBytesAfterRefused(file: string, tail: number, run: () => unknown | Promise, type: string) { + const clause = `${type}`; + await assert.rejects( + async () => run(), + (error: Error) => { + assert.ok(!(error instanceof CultCachePayloadDepthError), `${file}: the tail was walked`); + assert.ok(!error.message.includes(CANARY) && !error.message.includes(TAIL_CANARY), `${file}: the refusal echoed the tail`); + assert.ok(!error.message.includes("the limit is"), `${file}: the depth refusal`); + assert.ok(error.message.includes(clause), `${file}: names the type: ${error.message}`); + assert.ok(error.message.includes(`${tail} bytes after its first MessagePack value`), `${file}: names the count: ${error.message}`); + return true; + }, + file, + ); +} + +test("every bytes-after vector is refused by the one owner, by count, on the codec, plain and store paths", async () => { + const entries = await bytesAfterEntries(); + assert.equal(entries.length, 4); + const length = await valueLength(); + const surface = surfaceDocument(); + const plain = plainDocument(); + const codecOnly = codecDocument(); + const dir = await mkdtemp(join(tmpdir(), "bytes-after-")); + for (const { file, outcome, unionName, detail } of entries) { + assert.equal(outcome, "refused"); + assert.equal(unionName, "UnionSurface"); + assert.equal(detail, "bytes-after"); + const bytes = await read(file); + const tail = bytes.length - length; + assert.ok(tail > 0, file); + if (file === "bytes-after-deep-tail.msgpack") assert.ok(Buffer.from(bytes).includes(TAIL_CANARY), "the deep tail carries the canary the assertions look for"); + const payload = documentPayloadFromBytes(bytes); + await assertBytesAfterRefused(file, tail, () => decodeDocumentPayload(surface, payload), surface.type); + await assertBytesAfterRefused(file, tail, () => decodeDocumentPayload(plain, payload), plain.type); + await assertBytesAfterRefused(file, tail, () => decodeDocumentPayload(codecOnly, payload), codecOnly.type); + + // The store path: a single-file store holding the vector as a record payload refuses on load. + const path = join(dir, `${file}.cc`); + await new SingleFileMessagePackBackingStore(path).push(envelopeOf(bytes)); + const cache = CultCache.builder().withDocumentType(surface).withGenericStore(new SingleFileMessagePackBackingStore(path)).build(); + await assertBytesAfterRefused(file, tail, () => cache.pullAllBackingStores(), surface.type); + + // The put path: an envelope arriving with the tail is refused and held nowhere. + const written: CultCacheEnvelope[] = []; + const held = CultCache.builder().withDocumentType(surface).withGenericStore(memoryStore(written)).build(); + await assertBytesAfterRefused(file, tail, () => held.putEnvelope(surface, envelopeOf(bytes)), surface.type); + assert.equal(written.length, 0, `${file}: nothing reached the store`); + assert.equal(held.get(surface, "surface"), undefined, `${file}: nothing is held in memory`); + } +}); + +test("a payload with an unread slot and a tail is refused by count before any slot is marked or the record held", async () => { + const withSlot = await read("extra-slot-record.msgpack"); + const surface = surfaceDocument(); + const tailed = Uint8Array.from([...withSlot, 0xc0]); + const whole = CultCache.builder().withDocumentType(surface).withGenericStore(memoryStore([envelopeOf(withSlot)])).build(); + await whole.pullAllBackingStores(); + assert.ok(whole.unreadSlots(surface, "surface"), "without a tail the record is held with its unread slots"); + + const loading = CultCache.builder().withDocumentType(surface).withGenericStore(memoryStore([envelopeOf(tailed)])).build(); + await assertBytesAfterRefused("extra-slot-record+tail", 1, () => loading.pullAllBackingStores(), surface.type); + assert.equal(loading.get(surface, "surface"), undefined); + assert.equal(loading.unreadSlots(surface, "surface"), undefined); + + const written: CultCacheEnvelope[] = []; + const putting = CultCache.builder().withDocumentType(surface).withGenericStore(memoryStore(written)).build(); + await assertBytesAfterRefused("extra-slot-record+tail", 1, () => putting.putEnvelope(surface, envelopeOf(tailed)), surface.type); + assert.equal(written.length, 0); + assert.equal(putting.get(surface, "surface"), undefined); + assert.equal(putting.unreadSlots(surface, "surface"), undefined); +}); + +test("a payload that ends early is not bytes after its value: the decoder refuses it, and a user formatter's bytes are not walked", async () => { + const whole = await read("union-surface.msgpack"); + const surface = surfaceDocument(); + for (const cut of [0, 1, 10, whole.length - 1]) { + assert.throws( + () => decodeDocumentPayload(surface, documentPayloadFromBytes(whole.slice(0, cut))), + (error: Error) => !error.message.includes("bytes after") && !(error instanceof CultCachePayloadDepthError), + `truncated at ${cut}`, + ); + } + // A header cut short of its length is an early end too, whatever width it declares. + for (const early of ["dc00", "dd000000", "de00", "c4", "d9", "a5616263", "9291"]) { + assert.throws( + () => decodeDocumentPayload(surface, documentPayloadFromBytes(fromHex(early))), + (error: Error) => !error.message.includes("bytes after") && !(error instanceof CultCachePayloadDepthError), + `early end ${early}`, + ); + } + const opaque = defineDocumentType({ type: "tests.opaque", schema: z.unknown(), formatter: { encode: (value: Uint8Array) => value, decode: (bytes: Uint8Array) => bytes } }); + const tailed = await read("bytes-after-nil.msgpack"); + assert.equal((decodeDocumentPayload(opaque, documentPayloadFromBytes(tailed)) as Uint8Array).length, tailed.length); +}); diff --git a/packages/cultmesh-ts/test/provider-session-wire.test.ts b/packages/cultmesh-ts/test/provider-session-wire.test.ts index 4905f161..69314fbe 100644 --- a/packages/cultmesh-ts/test/provider-session-wire.test.ts +++ b/packages/cultmesh-ts/test/provider-session-wire.test.ts @@ -83,3 +83,16 @@ test("a provider command that is not a map, or lacks a required field, is refuse (error: Error) => error.message === "CultMesh provider command must be a MessagePack map.", ); }); + +test("a provider command payload with bytes after its first value is refused by count, without echoing the tail", () => { + const command = { commandId: "c1", commandKind: "k", providerId: "p", serviceInstanceId: "s", payload: { n: 1 } }; + const whole = documentPayloadBytes(encodeDocumentPayload(providerCommandDocument, command)); + const tailed = documentPayloadFromBytes(Uint8Array.from([...whole, ...Buffer.from("canary-tail-8c1f")])); + assert.throws( + () => decodeProviderCommandDocument({ + schemaId: cultMeshProviderSessionSchemas.command, recordKey: "p/s/c1", storedAt: "2026-10-11T00:00:00Z", + payloadEncoding: "messagepack" as const, payload: tailed, + }), + (error: Error) => error.message.includes("16 bytes after its first MessagePack value") && !error.message.includes("canary-tail-8c1f"), + ); +}); diff --git a/packages/cultnet-ts/test/content.test.ts b/packages/cultnet-ts/test/content.test.ts index ec23b64e..e2ae5ed9 100644 --- a/packages/cultnet-ts/test/content.test.ts +++ b/packages/cultnet-ts/test/content.test.ts @@ -392,6 +392,13 @@ test("decodeManifest accepts and refuses exactly what the reference does, and re } }); +test("a manifest with bytes after its first value is refused by count, whatever the tail is", () => { + const whole = encodeManifest(manifestOf()); + for (const tail of [[0xc0], [0xc1, 0xc1], [0x90]]) { + assert.throws(() => decodeManifest(Uint8Array.from([...whole, ...tail])), new RegExp(`${tail.length} bytes after its first MessagePack value`)); + } +}); + test("a manifest written as a map names the manifest, not a field inside it", () => { assert.throws(() => decodeManifest(encode({ 0: "a", 9: { k: "v" } })), /field 'manifest' has the wrong type/); }); diff --git a/packages/cultnet-ts/test/cultnet.test.ts b/packages/cultnet-ts/test/cultnet.test.ts index ac4f539d..581fd66a 100644 --- a/packages/cultnet-ts/test/cultnet.test.ts +++ b/packages/cultnet-ts/test/cultnet.test.ts @@ -14,6 +14,8 @@ import { SchemaConflictError, SingleFileMessagePackBackingStore, defineDocumentType, + documentPayloadBytes, + documentPayloadFromBytes, schemaIdentityOf, } from "@gamecult/cultcache-ts"; @@ -3140,3 +3142,24 @@ test("CultNet registry names the canonical schema id of a compatible alias, and assert.equal(registry.canonicalSchemaId("p3.nobody"), undefined); assert.equal(registry.canonicalSchemaId("p3.shared"), undefined, "an id several bindings claim resolves to none"); }); + +test("a replicated record with bytes after its first MessagePack value is refused by count and held nowhere", async () => { + const definition = defineDocumentType({ type: "tails.note", schemaId: "tails.note.v1", schema: z.object({ body: z.string() }) }); + const registry = new CultNetDocumentRegistry([defineCultNetDocumentBinding({ definition })]); + const origin = CultCache.builder().withDocumentType(definition).build(); + await origin.put(definition, "k", { body: "x" }); + const document = registry.createRawSnapshotResponse(origin, "tails").documents[0]!; + const whole = documentPayloadBytes(document.payload); + const target = CultCache.builder().withDocumentType(definition).build(); + await assert.rejects( + registry.applyRawDocumentPutMessage(target, { + schemaVersion: "cultnet.document_put_raw.v0", + messageId: "tails", + document: { ...document, payload: documentPayloadFromBytes(Uint8Array.from([...whole, 0xc0, 0xc1])) }, + }), + (error: Error) => error.message.includes('document type "tails.note" payload has 2 bytes after its first MessagePack value'), + ); + assert.deepEqual(target.snapshot(), []); + await registry.applyRawDocumentPutMessage(target, { schemaVersion: "cultnet.document_put_raw.v0", messageId: "whole", document }); + assert.deepEqual(target.getRequired(definition, "k"), { body: "x" }); +}); From 12f3801893956272a26d61413ea98301251e7171 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 08:01:32 +0000 Subject: [PATCH 08/12] Rust: a cache must pull its stores before it writes (cache-must-pull-before-write) (#133) --- packages/cultcache-rs/CHANGELOG.md | 1 + packages/cultcache-rs/src/lib.rs | 170 +++++++++++++++++-- packages/cultcache-rs/tests/slot_readonly.rs | 2 + packages/cultcache-rs/tests/union_parity.rs | 2 + 4 files changed, 160 insertions(+), 15 deletions(-) diff --git a/packages/cultcache-rs/CHANGELOG.md b/packages/cultcache-rs/CHANGELOG.md index 4e97588c..c6d72492 100644 --- a/packages/cultcache-rs/CHANGELOG.md +++ b/packages/cultcache-rs/CHANGELOG.md @@ -16,6 +16,7 @@ Semver: this is a breaking change in the 0.x line (0.3.x to 0.4.0). - A unit arm with slots (`[N, [extra...]]`) decodes as the unit arm and marks the record, where the derive refused it. - The registration map's value changes from the schema name to a private struct holding the schema name and the unread-slots check. - A cache whose last `pull_all_backing_stores` was refused (a tail, a payload past the nesting bound, a duplicate unique-index value, a retype conflict, a record from the wrong store) refuses every `put`, `put_envelope`, `put_raw_envelope`, `put_prepared_batch`, `update` and `delete` with `CultCacheLoadRefused` until a pull succeeds. Before, it stayed writable over an empty image, and the next batch write replaced the store with only the new records. +- A cache with a backing store also refuses every write and removal, with `CultCacheLoadRefused`, until it has pulled: a cache that never pulled, or that gained a store since its last pull, holds no image of its stores, and a batch from it replaced a populated store with only the new records. `is_writable` answers false meanwhile. A cache with no store needs no pull. Callers that attach a store and write (cultmesh nodes created with `pull_on_start: false`) must call `pull_all_backing_stores` first. - `decode_payload`, `put_raw_envelope` and prepared batches refuse a payload with bytes after its first MessagePack value (`payload has {n} bytes after its first MessagePack value`), by count; the tail is never walked or quoted. They used to read the first value and keep or drop the tail, so a record with a stray tail was served and rewritten without it. A store holding such a record (single-file, redb, `load_soa`) is refused at load, naming the type, the key and the count, and stays unreadable until the record is repaired; the same check now runs on every record at load, as the C# and Python runtimes do at open. The nesting bound is part of that check, so a stored payload nested past it is refused at load too, where it used to load and refuse on read. - A single-file store written as `cultcache.store.v3` is read and written: a store the C# runtime marked v3 (it holds a record with an element id: an owned list element with a `[CultElementId]` member) keeps its header here, and is carried through every rewrite with its ids; this runtime never drops the mark. A runtime from before this change refuses a v3 store by name. Rolling back to an older release after a store has been written as v3 is one-way: there is no carry-back, so restore from a copy taken before upgrading. - A write over a file the reader refuses (a zero-byte file, a truncated store, a stamp this runtime does not read) refuses with `CultCacheStoreUnreadable` and leaves the file as it is. A whole-store write used to replace such a file. This reverses 0.3.1, where `push_all` replaced a file with no readable stamp (healing it); the stamp rule now has one owner, the reader every write asks under the lock, and it accepts v1 and v3 (0.3.1 refused v3). diff --git a/packages/cultcache-rs/src/lib.rs b/packages/cultcache-rs/src/lib.rs index 522e226e..47a3dd18 100644 --- a/packages/cultcache-rs/src/lib.rs +++ b/packages/cultcache-rs/src/lib.rs @@ -2937,13 +2937,24 @@ pub struct CultCache { /// Records a store holds that no registered type claims: (store index, the envelope as the store read it). Derived from the last /// pull and replaced with it; written nowhere else. It decides only refusals at its keys. foreign: Vec<(usize, CultCacheEnvelope)>, - /// Why the last pull of the backing stores was refused; `None` once a pull succeeds. While set, every write and removal is refused. - load_refused: Option, + /// Whether the cache holds an image of its stores. While it is not `Pulled`, every write and removal is refused. + load: LoadState, } -/// A write or removal refused because the cache's last load of its backing stores was refused. The cache holds no -/// image of those stores, so any write would persist an image that lacks the records it never read. Repair the store -/// named by `reason` (which names a type, a key and a clause, never a payload) and load again. +/// What the cache knows of its stores. `Unpulled` is the state of a cache that has never pulled, or that gained a store +/// since: its image lacks whatever the stores hold, and a batch would replace them with it. +#[derive(Clone, Debug, PartialEq, Eq)] +enum LoadState { + Unpulled, + Pulled, + /// Why the last pull was refused. + Refused(String), +} + +/// A write or removal refused because the cache holds no image of its backing stores: its last load was refused, or it has +/// not pulled since its stores were attached. Any write would persist an image that lacks the records it never read. +/// Pull; for a refused load, first repair the store named by `reason` (which names a type, a key and a clause, never a +/// payload). #[derive(Clone, Debug, PartialEq, Eq)] pub struct CultCacheLoadRefused { pub reason: String, @@ -2953,7 +2964,7 @@ impl std::fmt::Display for CultCacheLoadRefused { fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { write!( formatter, - "CultCache refuses writes and removals: its load of the backing stores failed ({}); repair the store and load again", + "CultCache refuses writes and removals: it holds no image of its backing stores ({}); repair the store if it was refused and pull again", self.reason ) } @@ -3022,7 +3033,7 @@ impl CultCache { projectors: BTreeMap::new(), stores: Vec::new(), foreign: Vec::new(), - load_refused: None, + load: LoadState::Unpulled, } } @@ -3128,6 +3139,10 @@ impl CultCache { store: Box::new(store), types, }); + let loaded_before = self.load.clone(); + if self.load == LoadState::Pulled { + self.load = LoadState::Unpulled; + } // A record's home cannot change after it is admitted. let candidate = self.stores.len() - 1; let held: BTreeSet = self.entries.keys().map(|(type_id, _)| type_id.clone()).collect(); @@ -3146,6 +3161,7 @@ impl CultCache { self.describe_store(after) ); self.stores.pop(); + self.load = loaded_before; return Err(error); } } @@ -3164,21 +3180,27 @@ impl CultCache { self.entries = loaded; self.foreign = foreign; self.index_holders = holders; - self.load_refused = None; + self.load = LoadState::Pulled; Ok(()) } Err(error) => { - self.load_refused = Some(format!("{error:#}")); + self.load = LoadState::Refused(format!("{error:#}")); Err(error) } } } /// The one gate of every write and removal: a cache whose load was refused holds no image of its stores. + /// A cache without stores has nothing to lose and needs no pull. fn require_loaded(&self) -> Result<()> { - match &self.load_refused { - Some(reason) => Err(CultCacheLoadRefused { reason: reason.clone() }.into()), - None => Ok(()), + match &self.load { + _ if self.stores.is_empty() => Ok(()), + LoadState::Pulled => Ok(()), + LoadState::Refused(reason) => Err(CultCacheLoadRefused { reason: reason.clone() }.into()), + LoadState::Unpulled => Err(CultCacheLoadRefused { + reason: "pull_all_backing_stores has not run since the stores were attached".to_string(), + } + .into()), } } @@ -3269,9 +3291,12 @@ impl CultCache { /// Whether a write or removal of `(type_id, key)` would be admitted: false when the held payload carries slots past /// its registered type's declared slots, at any depth, or cannot be read by this build at all (the record is - /// read-only), when a store carries a foreign record at the key, or when another registered type holds it. True for a + /// read-only), when the cache holds no image of its stores (see [`CultCacheLoadRefused`]), when a store carries a foreign record at the key, or when another registered type holds it. True for a /// key nothing holds. An unregistered type is an error. Advisory: the write entries ask the same checks themselves. pub fn is_writable(&self, type_id: &str, key: &str) -> Result { + if self.require_loaded().is_err() { + return Ok(false); + } if !matches!(self.held_finding(type_id, key)?, HeldFinding::Free) { return Ok(false); } @@ -4724,6 +4749,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(path))?; + cache.pull_all_backing_stores()?; records(&mut cache) } @@ -4965,6 +4991,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_backing_store(SingleFileMessagePackBackingStore::new(&routed), [Settings::TYPE])?; + cache.pull_all_backing_stores()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic))?; cache.pull_all_backing_stores()?; let (staged, _) = cache.prepare_entry("other", &settings_one())?; @@ -5066,6 +5093,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put("shared", &settings_one())?; let before = std::fs::read(&store_path)?; @@ -5098,6 +5126,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put("kept", &settings_one())?; let before = std::fs::read(&store_path)?; @@ -5206,6 +5235,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; let (settings, _) = cache.prepare_entry( "app", @@ -5912,6 +5942,7 @@ mod tests { inner: SingleFileMessagePackBackingStore::new(&store_path), refuse_batch: refusal.clone(), })?; + cache.pull_all_backing_stores()?; cache.put( "app", &Settings { @@ -5952,6 +5983,7 @@ mod tests { ["settings"], )?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; + cache.pull_all_backing_stores()?; cache.put( "app", @@ -5980,6 +6012,7 @@ mod tests { let temp = tempfile::tempdir()?; let mut cache = CultCache::new(); cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("a.cc")))?; + cache.pull_all_backing_stores()?; let error = cache .add_generic_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("b.cc"))) .unwrap_err(); @@ -5993,6 +6026,7 @@ mod tests { let temp = tempfile::tempdir()?; let mut cache = CultCache::new(); cache.add_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("a.cc")), ["settings"])?; + cache.pull_all_backing_stores()?; let error = cache .add_backing_store( SingleFileMessagePackBackingStore::new(temp.path().join("b.cc")), @@ -6002,6 +6036,7 @@ mod tests { assert!(error.to_string().contains("already routed"), "{error}"); // A refused registration claims nothing, so "note" is still free. cache.add_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("c.cc")), ["note"])?; + cache.pull_all_backing_stores()?; assert_eq!(cache.stores.len(), 2); Ok(()) } @@ -6014,6 +6049,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; + cache.pull_all_backing_stores()?; cache.put("app", &Settings { theme: "v1-generic".to_string(), retries: 1 })?; let error = cache @@ -6043,15 +6079,17 @@ mod tests { let mut writer = CultCache::new(); writer.register_entry_type::()?; writer.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; + writer.pull_all_backing_stores()?; writer.put("app", &Settings { theme: "stray".to_string(), retries: 0 })?; let generic_bytes = fs::read(&generic_path)?; let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_backing_store(SingleFileMessagePackBackingStore::new(&settings_path), ["settings"])?; - cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; + cache.pull_all_backing_stores()?; // The cache already holds a record, in its home store, before the refused load. cache.put("held", &Settings { theme: "home".to_string(), retries: 0 })?; + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; let before = cache.snapshot(); let settings_bytes = fs::read(&settings_path)?; let error = cache.pull_all_backing_stores().unwrap_err(); @@ -6075,7 +6113,9 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_backing_store(SingleFileMessagePackBackingStore::new(&settings_path), [Settings::TYPE])?; + cache.pull_all_backing_stores()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; + cache.pull_all_backing_stores()?; cache.put("app", &Settings { theme: "before".to_string(), retries: 1 })?; let before = cache.snapshot(); @@ -6117,6 +6157,7 @@ mod tests { let mut homeless = CultCache::new(); homeless.register_registry(TestEntries)?; homeless.add_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("settings.cc")), ["settings"])?; + homeless.pull_all_backing_stores()?; homeless.put("held", &Settings { theme: "held".to_string(), retries: 0 })?; let before = homeless.snapshot(); let error = homeless.load_soa(table).unwrap_err(); @@ -6133,7 +6174,9 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_backing_store(SingleFileMessagePackBackingStore::new(&settings_path), ["settings"])?; + cache.pull_all_backing_stores()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&generic_path))?; + cache.pull_all_backing_stores()?; cache.put("memo", &Note { title: "generic".to_string(), body: "home".to_string() })?; cache.put("old", &Settings { theme: "kept".to_string(), retries: 0 })?; @@ -6164,6 +6207,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put("app", &Settings { theme: "before".to_string(), retries: 1 })?; let before = cache.snapshot(); let store_bytes = fs::read(&store_path)?; @@ -6203,6 +6247,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; let updated = cache.update::("app", |current| { let mut current = current.unwrap_or(Settings { @@ -6225,6 +6270,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put( "app", &Settings { @@ -6288,6 +6334,7 @@ mod tests { let mut origin = CultCache::new(); origin.register_entry_type::()?; origin.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&origin_store))?; + origin.pull_all_backing_stores()?; origin.put( "app", &Settings { @@ -6301,6 +6348,7 @@ mod tests { let mut target = CultCache::new(); target.register_entry_type::()?; target.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&target_store))?; + target.pull_all_backing_stores()?; let applied = target.put_envelope::(envelope.clone())?; assert_eq!( @@ -6334,6 +6382,7 @@ mod tests { cache.register_entry_type::()?; cache.register_entry_type::()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put_raw_envelope(envelope.clone())?; assert_eq!(cache.snapshot(), vec![envelope.clone()]); @@ -6376,6 +6425,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; let tailed = CultCacheEnvelope { key: "app".to_string(), r#type: Settings::TYPE.to_string(), @@ -6404,6 +6454,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put( "app", &Settings { @@ -6441,6 +6492,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_registry(TestEntries)?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put( "settings:a", &Settings { @@ -6523,6 +6575,7 @@ mod tests { let mut cache = CultCache::new(); cache.register_entry_type::()?; cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&store_path))?; + cache.pull_all_backing_stores()?; cache.put( "primary", &Settings { @@ -8051,4 +8104,91 @@ mod tests { assert_eq!(fs::read(&path)?, before); Ok(()) } -} \ No newline at end of file + + /// A store already holding `Settings:"held"`, then a fresh cache attached to it that has not pulled. + fn unpulled_cache_over_populated_stores() -> Result<(tempfile::TempDir, Vec, Vec)> { + let temp = tempfile::tempdir()?; + let single = temp.path().join("single.cc"); + let redb = temp.path().join("redb.db"); + let held = Settings { theme: "held".to_string(), retries: 1 }; + let mut seeder = CultCache::new(); + seeder.register_registry(TestEntries)?; + seeder.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&single))?; + seeder.pull_all_backing_stores()?; + seeder.put("held", &held)?; + let mut seeder = CultCache::new(); + seeder.register_registry(TestEntries)?; + seeder.add_generic_backing_store(RedbMessagePackBackingStore::new(&redb)?)?; + seeder.pull_all_backing_stores()?; + seeder.put("held", &held)?; + + let mut over_single = CultCache::new(); + over_single.register_registry(TestEntries)?; + over_single.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&single))?; + let mut over_redb = CultCache::new(); + over_redb.register_registry(TestEntries)?; + over_redb.add_generic_backing_store(RedbMessagePackBackingStore::new(&redb)?)?; + Ok((temp, vec![over_single, over_redb], vec![single, redb])) + } + + #[test] + fn a_cache_that_never_pulled_refuses_every_write_and_removal() -> Result<()> { + let (_temp, caches, _paths) = unpulled_cache_over_populated_stores()?; + for mut cache in caches { + let fresh = Note { title: "t".to_string(), body: "b".to_string() }; + let (staged, _) = cache.prepare_entry("fresh", &fresh)?; + for error in [ + cache.put_prepared_batch(vec![staged.clone()]).unwrap_err(), + cache.put("fresh", &fresh).unwrap_err(), + cache.put_raw_envelope(staged.clone()).unwrap_err(), + cache.delete::("held").unwrap_err(), + ] { + assert!(error.downcast_ref::().is_some(), "{error:#}"); + } + assert!(cache.snapshot().is_empty()); + } + Ok(()) + } + + #[test] + fn a_batch_from_a_cache_that_never_pulled_cannot_wipe_the_store() -> Result<()> { + let (_temp, caches, paths) = unpulled_cache_over_populated_stores()?; + for (mut cache, path) in caches.into_iter().zip(paths) { + let (staged, _) = cache.prepare_entry("fresh", &Note { title: "t".to_string(), body: "b".to_string() })?; + let _ = cache.put_prepared_batch(vec![staged]); + let mut reader = CultCache::new(); + reader.register_registry(TestEntries)?; + if path.extension().is_some_and(|extension| extension == "db") { + reader.add_generic_backing_store(RedbMessagePackBackingStore::new(&path)?)?; + reader.pull_all_backing_stores()?; + } else { + reader.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path))?; + reader.pull_all_backing_stores()?; + } + reader.pull_all_backing_stores()?; + assert_eq!(reader.get_required::("held")?.theme, "held", "{path:?}"); + } + Ok(()) + } + + #[test] + fn pulling_lifts_the_gate_and_a_store_attached_later_closes_it_again() -> Result<()> { + let (temp, mut caches, _paths) = unpulled_cache_over_populated_stores()?; + let mut cache = caches.remove(0); + cache.pull_all_backing_stores()?; + cache.put("fresh", &Note { title: "t".to_string(), body: "b".to_string() })?; + assert_eq!(cache.get_required::("held")?.theme, "held"); + + let mut routed = CultCache::new(); + routed.register_registry(TestEntries)?; + let settings = Settings { theme: "t".to_string(), retries: 0 }; + routed.add_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("routed.cc")), [Settings::TYPE])?; + routed.pull_all_backing_stores()?; + routed.put("fresh", &settings)?; + routed.add_generic_backing_store(SingleFileMessagePackBackingStore::new(temp.path().join("other.cc")))?; + assert!(routed.put("again", &settings).unwrap_err().downcast_ref::().is_some()); + routed.pull_all_backing_stores()?; + routed.put("again", &settings)?; + Ok(()) + } +} diff --git a/packages/cultcache-rs/tests/slot_readonly.rs b/packages/cultcache-rs/tests/slot_readonly.rs index 0031e372..cea0853a 100644 --- a/packages/cultcache-rs/tests/slot_readonly.rs +++ b/packages/cultcache-rs/tests/slot_readonly.rs @@ -149,6 +149,7 @@ fn every_read_only_vector_decodes_marked_and_is_not_writable() { let mut cache = registered(); let dir = tempfile::tempdir().unwrap(); cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(dir.path().join("s.cc"))).unwrap(); + cache.pull_all_backing_stores().unwrap(); cache.put_raw_envelope(envelope("k", payload.clone())).unwrap(); assert_eq!(cache.get::("k").unwrap(), Some(constructed()), "{file}"); assert!(!cache.is_writable(UnionSurface::TYPE, "k").unwrap(), "{file}"); @@ -470,6 +471,7 @@ fn is_writable_is_false_for_a_type_with_no_home_store() { let mut cache = registered(); // One store, routed to SlotUnit only: UnionSurface has no home. cache.add_backing_store(SingleFileMessagePackBackingStore::new(dir.path().join("s.cc")), [SlotUnit::TYPE]).unwrap(); + cache.pull_all_backing_stores().unwrap(); assert!(cache.put("k", &constructed()).is_err()); assert!(!cache.is_writable(UnionSurface::TYPE, "k").unwrap()); assert!(cache.is_writable(SlotUnit::TYPE, "k").unwrap()); diff --git a/packages/cultcache-rs/tests/union_parity.rs b/packages/cultcache-rs/tests/union_parity.rs index 67f2b389..bc3aba6d 100644 --- a/packages/cultcache-rs/tests/union_parity.rs +++ b/packages/cultcache-rs/tests/union_parity.rs @@ -718,6 +718,7 @@ fn write_refused_past_limit() { let mut cache = CultCache::new(); cache.register_entry_type::().unwrap(); cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + cache.pull_all_backing_stores().unwrap(); let at_limit = wrap_chain(126, boxed("0000000000f0", vec![])); assert_eq!(cache.put("at", &at_limit).unwrap(), at_limit); @@ -801,6 +802,7 @@ fn the_raw_doors_hold_the_bound_for_arms_maps_and_arrays() { cache.register_entry_type::().unwrap(); cache.register_entry_type::().unwrap(); cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path)).unwrap(); + cache.pull_all_backing_stores().unwrap(); cache.put_raw_envelope(raw(type_name, "at-raw", at.clone())).unwrap(); cache.put_prepared_batch(vec![raw(type_name, "at-batch", at.clone())]).unwrap(); let stored = cache.snapshot_envelopes(); From fbeb773d35412fb02ed29ca2fdb260a9c0c52686 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 08:45:52 +0000 Subject: [PATCH 09/12] Rust: batch writes merge onto the store (rs-batch-merges-onto-store) (#135) --- packages/cultcache-rs/CHANGELOG.md | 1 + packages/cultcache-rs/README.md | 7 +- packages/cultcache-rs/src/lib.rs | 167 ++++++++++++++++++++--------- 3 files changed, 121 insertions(+), 54 deletions(-) diff --git a/packages/cultcache-rs/CHANGELOG.md b/packages/cultcache-rs/CHANGELOG.md index c6d72492..b43c13c6 100644 --- a/packages/cultcache-rs/CHANGELOG.md +++ b/packages/cultcache-rs/CHANGELOG.md @@ -16,6 +16,7 @@ Semver: this is a breaking change in the 0.x line (0.3.x to 0.4.0). - A unit arm with slots (`[N, [extra...]]`) decodes as the unit arm and marks the record, where the derive refused it. - The registration map's value changes from the schema name to a private struct holding the schema name and the unread-slots check. - A cache whose last `pull_all_backing_stores` was refused (a tail, a payload past the nesting bound, a duplicate unique-index value, a retype conflict, a record from the wrong store) refuses every `put`, `put_envelope`, `put_raw_envelope`, `put_prepared_batch`, `update` and `delete` with `CultCacheLoadRefused` until a pull succeeds. Before, it stayed writable over an empty image, and the next batch write replaced the store with only the new records. +- `CacheBackingStore::push_all` writes the given entries onto the store as it is when the write lands, under the store's lock or transaction: a record under the same identity is replaced and no other record is removed. It used to replace everything the store held, so a cache whose image was stale (another writer had added a record since its pull) wiped that record with `put_prepared_batch`. `put_prepared_batch` now hands the store its staged records alone, as C# `CommitBatch` and Python `push_all` do; the lay-back of foreign records is gone because they are never rewritten. A custom `CacheBackingStore` must give `push_all` merge semantics. `push_all` reads the store, so through a dangling link it is refused like `push` and no longer replaces the link; a soft flush is unchanged. - A cache with a backing store also refuses every write and removal, with `CultCacheLoadRefused`, until it has pulled: a cache that never pulled, or that gained a store since its last pull, holds no image of its stores, and a batch from it replaced a populated store with only the new records. `is_writable` answers false meanwhile. A cache with no store needs no pull. Callers that attach a store and write (cultmesh nodes created with `pull_on_start: false`) must call `pull_all_backing_stores` first. - `decode_payload`, `put_raw_envelope` and prepared batches refuse a payload with bytes after its first MessagePack value (`payload has {n} bytes after its first MessagePack value`), by count; the tail is never walked or quoted. They used to read the first value and keep or drop the tail, so a record with a stray tail was served and rewritten without it. A store holding such a record (single-file, redb, `load_soa`) is refused at load, naming the type, the key and the count, and stays unreadable until the record is repaired; the same check now runs on every record at load, as the C# and Python runtimes do at open. The nesting bound is part of that check, so a stored payload nested past it is refused at load too, where it used to load and refuse on read. - A single-file store written as `cultcache.store.v3` is read and written: a store the C# runtime marked v3 (it holds a record with an element id: an owned list element with a `[CultElementId]` member) keeps its header here, and is carried through every rewrite with its ids; this runtime never drops the mark. A runtime from before this change refuses a v3 store by name. Rolling back to an older release after a store has been written as v3 is one-way: there is no carry-back, so restore from a copy taken before upgrading. diff --git a/packages/cultcache-rs/README.md b/packages/cultcache-rs/README.md index ed9701d5..da4185e2 100644 --- a/packages/cultcache-rs/README.md +++ b/packages/cultcache-rs/README.md @@ -384,10 +384,9 @@ an empty store. Any other failure to reach it, such as a symlink loop, a file where a parent directory should be, or a missing permission, is an I/O error. A store file that exists and holds zero bytes is `Undecodable`: no CultCache writer leaves one. A symbolic link whose target is gone is not a missing store: -reading it is an I/O error, so a read-then-write (`push`, `delete`, the exchanges) -fails before writing. Writes do not yet follow links: a write through a live link, -and a `push_all` through a dangling one, replace the link with a regular file and -leave its target alone. A refusal names no +reading it is an I/O error, so a read-then-write (`push`, `push_all`, `delete`, the exchanges) +fails before writing. Writes do not yet follow links: a write through a live link +replaces the link with a regular file and leaves its target alone. A refusal names no value read from the file; it names a record only by its key and schema id, and a header only when it has the shape `cultcache.store.v`. diff --git a/packages/cultcache-rs/src/lib.rs b/packages/cultcache-rs/src/lib.rs index 47a3dd18..678ff4d6 100644 --- a/packages/cultcache-rs/src/lib.rs +++ b/packages/cultcache-rs/src/lib.rs @@ -1253,12 +1253,14 @@ pub trait CacheBackingStore: Send { fn push(&mut self, entry: &CultCacheEnvelope) -> Result<()>; fn delete(&mut self, entry: &CultCacheEnvelope) -> Result<()>; - /// Replaces everything the store holds with `entries` in one atomic step: - /// on `Err` the store still holds exactly what it held before. A cache - /// batch (`put_prepared_batch`) is all-or-nothing only through this. There - /// is no default, because a replace assembled from `delete` and `push` - /// calls leaves the store emptied or partial when one of them fails; a - /// store without an atomic replace does not compile: + /// Writes `entries` onto the store as it is when the write lands, in one + /// atomic step: a record under the same identity is replaced and no other + /// record is removed or changed, so a writer whose image is stale cannot + /// drop a record it never saw. On `Err` the store still holds exactly what + /// it held before. A cache batch (`put_prepared_batch`) is all-or-nothing + /// only through this. There is no default, because a write assembled from + /// `push` calls leaves the store partial when one of them fails; a store + /// without an atomic write does not compile: /// /// ```compile_fail,E0046 /// use cultcache_rs::{CacheBackingStore, CultCacheEnvelope}; @@ -1675,13 +1677,7 @@ impl SingleFileMessagePackBackingStore { // payloads this store passes through may hold ids), one that is gone or legacy is written v1. The file is read by // the same reader as an open, so a store this runtime cannot read completely, a variant included, is refused // and never overwritten. - let disk = match self.read_store_unlocked() { - // A dangling link holds no store, so a whole-file replace has no header to keep and replaces the link (R3 pins - // this). The only NotFound the reader returns is that one: a merging writer has already refused it when it - // read the current entries. - Err(error) if io_not_found(&error) => DiskStore::absent(), - other => other?, - }; + let disk = self.read_store_unlocked()?; let entries = settle_one_record_per_key(entries, &disk)?; let bytes = encode_store_snapshot(&entries, &disk) .and_then(|snapshot| rmp_serde::to_vec(&snapshot).context("failed to encode MessagePack")) @@ -1962,14 +1958,18 @@ impl CacheBackingStore for SingleFileMessagePackBackingStore { } // Of entries sharing a key and a schema id the last one given is the record, whatever their type names (a type - // renamed under a stable id); a second schema id under a key reaches the commit guard. + // renamed under a stable id); it replaces the held record under that key and schema id. A second schema id + // under a key reaches the commit guard, which judges the merged store. let mut by_id = BTreeMap::<(String, String), &CultCacheEnvelope>::new(); for entry in entries { by_id.insert((schema_identity(entry).to_string(), entry.key.clone()), entry); } - let mut entries: Vec = by_id.into_values().cloned().collect(); - entries.sort_by_key(entry_id); - self.write_all_unlocked(&entries) + let mut merged = self.read_all_unlocked()?; + let replaced: BTreeSet<_> = by_id.values().map(|entry| entry_id(entry)).collect(); + merged.retain(|held| !replaced.contains(&entry_id(held))); + merged.extend(by_id.into_values().cloned()); + merged.sort_by_key(entry_id); + self.write_all_unlocked(&merged) }) } } @@ -2757,13 +2757,6 @@ impl CacheBackingStore for OwnedRedbMessagePackBackingStore { let write = self.inner.database.begin_write()?; { let mut table = write.open_table(REDB_ENVELOPES)?; - let keys = table - .iter()? - .map(|row| Ok(row?.0.value().to_vec())) - .collect::>>()?; - for key in keys { - table.remove(key.as_slice())?; - } for entry in entries { insert_redb_entry(&mut table, entry)?; } @@ -2899,13 +2892,6 @@ impl CacheBackingStore for RedbMessagePackBackingStore { let write = database.begin_write()?; { let mut table = write.open_table(REDB_ENVELOPES)?; - let keys = table - .iter()? - .map(|row| Ok(row?.0.value().to_vec())) - .collect::>>()?; - for key in keys { - table.remove(key.as_slice())?; - } for entry in entries { insert_redb_entry(&mut table, entry)?; } @@ -3602,18 +3588,12 @@ impl CultCache { } let change = self.judge_unique(&staged.iter().collect::>())?; if let Some(store_index) = store_index { - // The home store receives its own records only: those it already holds in this cache plus the staged ones, and - // the records it carries that no registered type claims: push_all replaces the store's records, so a foreign - // record is laid back as the store read it. - let snapshot: Vec<_> = next - .values() - .filter(|entry| self.resolve_route_index(&entry.r#type) == Some(store_index)) - .chain(self.foreign.iter().filter(|(home, _)| *home == store_index).map(|(_, entry)| entry)) - .cloned() - .collect(); + // The home store is handed the staged records alone and applies them onto what it holds now. + // A key staged twice lands as its last record, whatever the store. + let landing: BTreeMap<_, _> = staged.iter().map(|entry| (entry_id(entry), entry.clone())).collect(); self.stores[store_index] .store - .push_all(&snapshot, PushAllOptions::default())?; + .push_all(&landing.into_values().collect::>(), PushAllOptions::default())?; } self.entries = next; change.apply(&mut self.index_holders); @@ -4324,12 +4304,6 @@ fn release_lock(lock: File, result: Result) -> Result { result } -fn io_not_found(error: &anyhow::Error) -> bool { - error - .chain() - .any(|cause| cause.downcast_ref::().is_some_and(|io| io.kind() == std::io::ErrorKind::NotFound)) -} - /// True only when nothing is at `path`; a symbolic link, dangling or not, is something. /// A path that cannot be inspected (a file where a directory should be, no permission) /// is an error, never absent. @@ -5400,6 +5374,12 @@ mod tests { Ok(()) } + fn io_not_found(error: &anyhow::Error) -> bool { + error + .chain() + .any(|cause| cause.downcast_ref::().is_some_and(|io| io.kind() == std::io::ErrorKind::NotFound)) + } + fn test_envelope(r#type: &str, key: &str, payload: &[u8]) -> CultCacheEnvelope { CultCacheEnvelope { key: key.to_string(), @@ -7980,15 +7960,17 @@ mod tests { #[cfg(unix)] #[test] - fn r3_decides_resolve_or_refuse_a_push_all_through_a_dangling_link_replaces_the_link() -> Result<()> { + fn a_push_all_through_a_dangling_link_is_refused_like_push_and_writes_nothing() -> Result<()> { let temp = tempfile::tempdir()?; let volume = temp.path().join("volume"); fs::create_dir(&volume)?; let link = temp.path().join("store.cc"); std::os::unix::fs::symlink(volume.join("store.cc"), &link)?; - SingleFileMessagePackBackingStore::new(&link).push_all(&[snapshot_envelope("a", b"1")], PushAllOptions::default())?; - assert!(fs::symlink_metadata(&link)?.file_type().is_file(), "the link was replaced by a regular file"); + let mut store = SingleFileMessagePackBackingStore::new(&link); + assert!(store.push_all(&[snapshot_envelope("a", b"1")], PushAllOptions::default()).is_err()); + assert!(store.push(&snapshot_envelope("a", b"1")).is_err()); + assert!(fs::symlink_metadata(&link)?.file_type().is_symlink(), "the link is left as it was"); assert_eq!(fs::read_dir(&volume)?.count(), 0, "nothing reached the link's volume"); Ok(()) } @@ -8191,4 +8173,89 @@ mod tests { routed.put("again", &settings)?; Ok(()) } + + #[test] + fn a_stale_cache_batch_keeps_the_records_another_writer_added() -> Result<()> { + let temp = tempfile::tempdir()?; + let single = temp.path().join("single.cc"); + let redb = temp.path().join("redb.db"); + let open = |path: &Path| -> Result { + let mut cache = CultCache::new(); + cache.register_registry(TestEntries)?; + if path.extension().is_some_and(|extension| extension == "db") { + cache.add_generic_backing_store(RedbMessagePackBackingStore::new(path)?)?; + } else { + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(path))?; + } + cache.pull_all_backing_stores()?; + Ok(cache) + }; + for path in [&single, &redb] { + let mut seed = open(path)?; + seed.put("held", &Settings { theme: "held".to_string(), retries: 1 })?; + let mut stale = open(path)?; + let mut other = open(path)?; + other.put("held", &Settings { theme: "other-newer".to_string(), retries: 3 })?; + other.put("added", &Settings { theme: "other".to_string(), retries: 2 })?; + let (staged, _) = stale.prepare_entry("mine", &Note { title: "t".to_string(), body: "b".to_string() })?; + stale.put_prepared_batch(vec![staged])?; + let reader = open(path)?; + assert_eq!(reader.get_required::("added")?.theme, "other", "{path:?}"); + assert_eq!(reader.get_required::("held")?.theme, "other-newer", "a record another writer updated is not reverted"); + assert_eq!(reader.get_required::("mine")?.title, "t", "{path:?}"); + } + Ok(()) + } + + #[test] + fn every_store_merges_push_all_and_lands_the_last_of_a_repeated_key() -> Result<()> { + let temp = tempfile::tempdir()?; + let stores: Vec> = vec![ + Box::new(SingleFileMessagePackBackingStore::new(temp.path().join("single.cc"))), + Box::new(RedbMessagePackBackingStore::new(temp.path().join("redb.db"))?), + Box::new(OwnedRedbMessagePackBackingStore::new(temp.path().join("owned.db"))?), + ]; + for mut store in stores { + store.push(&test_envelope("pushall.merge", "kept", b"kept"))?; + store.push_all(&[test_envelope("pushall.merge", "new", b"new")], PushAllOptions::default())?; + let keys: Vec<_> = store.pull_all()?.into_iter().map(|entry| entry.key).collect(); + assert_eq!(keys, ["kept", "new"]); + } + // A batch naming a key twice lands the last, in the single-file and the redb stores alike. + for path in [temp.path().join("dup.cc"), temp.path().join("dup.db")] { + let mut cache = CultCache::new(); + cache.register_registry(TestEntries)?; + if path.extension().is_some_and(|extension| extension == "db") { + cache.add_generic_backing_store(RedbMessagePackBackingStore::new(&path)?)?; + } else { + cache.add_generic_backing_store(SingleFileMessagePackBackingStore::new(&path))?; + } + cache.pull_all_backing_stores()?; + let (first, _) = cache.prepare_entry("k", &Settings { theme: "first".to_string(), retries: 1 })?; + let (last, _) = cache.prepare_entry("k", &Settings { theme: "last".to_string(), retries: 2 })?; + cache.put_prepared_batch(vec![first, last])?; + assert_eq!(cache.get_required::("k")?.theme, "last"); + } + Ok(()) + } + + #[test] + fn push_all_replaces_a_record_held_under_another_schema_id_as_push_does() -> Result<()> { + let temp = tempfile::tempdir()?; + let mut store = SingleFileMessagePackBackingStore::new(temp.path().join("store.cc")); + let mut held = snapshot_envelope("k", b"old"); + held.schema_id = Some("legacy.id".to_string()); + store.push(&held)?; + let mut replacement = snapshot_envelope("k", b"new"); + replacement.schema_id = None; + let mut by_push = store.clone(); + by_push.push(&replacement)?; + assert_eq!(store.pull_all()?.len(), 1); + store.push(&held)?; + store.push_all(&[replacement.clone()], PushAllOptions::default())?; + let held_now = store.pull_all()?; + assert_eq!(held_now.len(), 1, "one record under the key"); + assert_eq!(held_now[0].payload, replacement.payload); + Ok(()) + } } From b098d5b7da5a884fc1af5488aaccdc5c1931812f Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 09:47:33 +0000 Subject: [PATCH 10/12] TypeScript: pushAll writes onto the store instead of replacing it (ts-flush-replaces-store) (#136) --- packages/cultcache-ts/CHANGELOG.md | 1 + packages/cultcache-ts/README.md | 2 +- packages/cultcache-ts/src/cult-cache.ts | 7 +++++ .../single-file-messagepack-backing-store.ts | 22 +++++++------ packages/cultcache-ts/test/cult-cache.test.ts | 31 ++++++++++++++++--- packages/cultcache-ts/test/store-file.test.ts | 11 +++++-- packages/cultmesh-ts/src/index.ts | 2 +- packages/cultmesh-ts/test/cultmesh.test.ts | 30 +++++++++++++++++- 8 files changed, 88 insertions(+), 18 deletions(-) diff --git a/packages/cultcache-ts/CHANGELOG.md b/packages/cultcache-ts/CHANGELOG.md index 2da49fe9..dcbfd12e 100644 --- a/packages/cultcache-ts/CHANGELOG.md +++ b/packages/cultcache-ts/CHANGELOG.md @@ -7,6 +7,7 @@ All notable changes to `@gamecult/cultcache-ts` are recorded here, following ### Breaking +- `SingleFileMessagePackBackingStore.pushAll` writes the given entries onto the store as it is under the lock, as `push` does: a record under the same key is replaced and no other is removed. It used to replace the whole file with the entries, so `CultMesh.flush()` from a cache that held less than the store (a node created with `pullOnStart: false`, or one whose image was stale) wiped the records it never read. A caller that relied on `pushAll` to reset a store removes the file first. A soft flush and the dangling-link case are unchanged. `CultCache.storedSnapshot()` is the image as the stores hold it (a global loaded from a legacy key under that key) and is what `CultMesh.flush` persists, so a flush no longer leaves such a global under two keys; `snapshot()` is unchanged. - A single-file store marked `cultcache.store.v3` (it holds an element id, written by the C# runtime) is read, and a rewrite keeps the mark and the ids; this runtime does not decide when a document carries ids. A runtime from before this change refuses a v3 store by name, so rolling back to an older release after a store has been written as v3 is one-way: there is no carry-back, so restore from a copy taken before upgrading. - `indexes` are unique, as C# `[CultIndex]`: a write or a load that leaves two records of a type holding one non-blank value under one alias is refused naming both keys and the alias, and writes nothing. Index text is the C# canonical diff --git a/packages/cultcache-ts/README.md b/packages/cultcache-ts/README.md index a081cd92..625d28dd 100644 --- a/packages/cultcache-ts/README.md +++ b/packages/cultcache-ts/README.md @@ -174,7 +174,7 @@ const settings = cache.getRequiredGlobal(settingsDocument); - `updateGlobal(...)` - `delete(...)` - `deleteGlobal(...)` -- `snapshot()` +- `snapshot()` (the logical image) and `storedSnapshot()` (the image as the stores hold it, for a whole flush) - `getEnvelope(...)` - `getRequiredEnvelope(...)` - `getGlobalEnvelope(...)` diff --git a/packages/cultcache-ts/src/cult-cache.ts b/packages/cultcache-ts/src/cult-cache.ts index 6cf7a8cb..a88bd66d 100644 --- a/packages/cultcache-ts/src/cult-cache.ts +++ b/packages/cultcache-ts/src/cult-cache.ts @@ -636,6 +636,13 @@ export class CultCache { return [...this.#hydrated.entries.values()].map((entry) => this.#toEnvelope(entry)); } + // The image as the stores hold it, for persisting it whole (a flush): a global loaded from a legacy key is handed out under + // that key until its first write retires it, so a flush cannot leave the store holding it under two keys. `snapshot` is the + // logical image and always names a global by CultCache.GLOBAL_KEY. + storedSnapshot(): CultCacheEnvelope[] { + return [...this.#hydrated.entries.values()].map((entry) => this.#hydrated.legacyGlobals.get(entry.type) ?? this.#toEnvelope(entry)); + } + #registerNow(definition: AnyCultCacheDocumentDefinition): void { // Every registration refusal is a SchemaConflictError naming the new definition's schema id and both schema names. const refuse = (holder: RegisteredDefinition, claimed: CultCacheSchemaCatalogEntry, what: string) => diff --git a/packages/cultcache-ts/src/single-file-messagepack-backing-store.ts b/packages/cultcache-ts/src/single-file-messagepack-backing-store.ts index 5dfcfc74..298f1bee 100644 --- a/packages/cultcache-ts/src/single-file-messagepack-backing-store.ts +++ b/packages/cultcache-ts/src/single-file-messagepack-backing-store.ts @@ -82,10 +82,12 @@ export class SingleFileMessagePackBackingStore implements CacheBackingStore { async pushAll(entries: CultCacheEnvelope[], options: PushAllOptions = {}): Promise { await this.#enqueue(async () => { - // A flush of the whole store writes the header the file on disk carries, read now: a file marked for element ids - // stays marked, and one that is not (or is gone or legacy) is written unmarked. The file is read by the same - // reader as `pullAll`, so one it would refuse (not exactly one store, a variant, a body it cannot decode) is refused - // and left as it is. A soft flush asks that reader too: nothing at the path is written, a store is left alone. + // A flush writes `entries` onto the store as it is under the lock, the way a push does: a record under the same key + // is replaced and no other is removed, so a cache whose image is stale (or empty, never pulled) cannot drop what + // another writer added. The store is read by the same reader as `pullAll`, so one it would refuse (not exactly one + // store, a variant, a body it cannot decode) is refused and left as it is. A soft flush asks that reader too: nothing + // at the path is written, a store is left alone. + let held: CultCacheEnvelope[] = []; try { const disk = this.#readDisk(); if (options.soft && !disk.absent) { @@ -93,10 +95,11 @@ export class SingleFileMessagePackBackingStore implements CacheBackingStore { } this.#format = disk.format; + held = disk.envelopes; } catch (error) { - // Only a dangling link reaches here with ENOENT: it holds no store, so a whole-store flush has no header to keep and - // replaces the link (R3 pins this). A soft flush does not replace what it cannot read, and every merging writer - // has already refused the link when it read the current entries. + // Only a dangling link reaches here with ENOENT: it holds no store, so a flush has nothing to merge with and no + // header to keep, and replaces the link (R3 pins this). A soft flush does not replace what it cannot read, and every + // merging writer has already refused the link when it read the current entries. if (options.soft || (error as NodeJS.ErrnoException).code !== "ENOENT") { throw error; } @@ -104,9 +107,10 @@ export class SingleFileMessagePackBackingStore implements CacheBackingStore { this.#format = STORE_FORMAT_VERSION; } - requireOneTypePerKey(this.filePath, entries, []); // A key is held once: of same-type entries sharing a key in one batch, the last is kept. - await this.#writeAll([...new Map(entries.map((entry) => [entry.key, entry])).values()], []); + const supplied = [...new Map(entries.map((entry) => [entry.key, entry])).values()]; + requireOneTypePerKey(this.filePath, entries, held); + await this.#writeAll(supplied, held); }); } diff --git a/packages/cultcache-ts/test/cult-cache.test.ts b/packages/cultcache-ts/test/cult-cache.test.ts index b575170f..2636a9d3 100644 --- a/packages/cultcache-ts/test/cult-cache.test.ts +++ b/packages/cultcache-ts/test/cult-cache.test.ts @@ -1019,6 +1019,7 @@ test("legacyKeyGlobalLoadsWithoutWritingAndFirstWriteLeavesOneGlobal", async () assert.deepEqual(reloaded.getRequiredGlobal(settingsDocument), { t: "new" }); // Deleting an adopted global removes the legacy record. + await rm(storePath); await new SingleFileMessagePackBackingStore(storePath).pushAll([legacy]); const deleting = build(); await deleting.pullAllBackingStores(); @@ -1026,10 +1027,30 @@ test("legacyKeyGlobalLoadsWithoutWritingAndFirstWriteLeavesOneGlobal", async () assert.deepEqual(await new SingleFileMessagePackBackingStore(storePath).pullAll(), []); // Two globals of one type on disk, under any keys, are refused. + await rm(storePath); await new SingleFileMessagePackBackingStore(storePath).pushAll([legacy, { ...legacy, key: CultCache.GLOBAL_KEY }]); await assert.rejects(build().pullAllBackingStores(), /has multiple persisted entries/u); }); +test("a flush of a cache that loaded a legacy-key global leaves the store reopenable", async () => { + const settingsDocument = defineDocumentType({ type: "settings", schema: z.object({ t: z.string() }), global: true }); + const storePath = join(await mkdtemp(join(tmpdir(), "cultcache-legacy-flush-")), "generic.cc"); + const store = new SingleFileMessagePackBackingStore(storePath); + const build = () => CultCache.builder().withRegistry(defineDocumentRegistry(settingsDocument)).withGenericStore(store).build(); + const seed = CultCache.builder().withRegistry(defineDocumentRegistry(settingsDocument)).build(); + await seed.putGlobal(settingsDocument, { t: "old" }); + await store.push({ ...seed.getRequiredEnvelope(settingsDocument, CultCache.GLOBAL_KEY), key: "legacy" }); + + const cache = build(); + await cache.pullAllBackingStores(); + assert.equal(cache.snapshot()[0]!.key, CultCache.GLOBAL_KEY, "the logical image names the global by the global key"); + await store.pushAll(cache.storedSnapshot()); + assert.deepEqual((await store.pullAll()).map((entry) => entry.key), ["legacy"], "the flush wrote the record under the key the store holds it"); + const reopened = build(); + await reopened.pullAllBackingStores(); + assert.deepEqual(reopened.getRequiredGlobal(settingsDocument), { t: "old" }); +}); + test("concurrentPutAndAttachCannotLandRecordInNonHomeStore", async () => { const settingsDocument = defineDocumentType({ type: "settings", schema: z.object({ t: z.string() }) }); const tempDir = await mkdtemp(join(tmpdir(), "cultcache-put-attach-race-")); @@ -2261,7 +2282,7 @@ test("a stamp this runtime reads, followed by bytes that are not UTF-8, gets the } }); -test("pushAll leaves a store it reads alone when soft, and replaces one it reads otherwise", async () => { +test("pushAll leaves a store it reads alone when soft, and writes onto one it reads otherwise", async () => { const vector = await readFile(resolve(cultLibRoot, "tests", "vectors", "document-variants-c2a", "v3-base.msgpack")); const envelope = { key: "probe", type: "probe", payload: documentPayloadFromBytes(new Uint8Array([0x80])), storedAt: "2026-10-22T00:00:00Z" }; const existing = join(await mkdtemp(join(tmpdir(), "cultcache-pushall-soft-")), "store.msgpack"); @@ -2273,7 +2294,7 @@ test("pushAll leaves a store it reads alone when soft, and replaces one it reads const store = new SingleFileMessagePackBackingStore(own); await store.pushAll([envelope]); await store.pushAll([{ ...envelope, key: "second" }]); - assert.deepEqual((await store.pullAll()).map((entry) => entry.key), ["second"]); + assert.deepEqual((await store.pullAll()).map((entry) => entry.key), ["probe", "second"]); }); test("SingleFileMessagePackBackingStore reads only a missing store as empty", async () => { @@ -2575,7 +2596,7 @@ test("SingleFileMessagePackBackingStore pushAll refuses a store whose header it // A file is replaced by a rewrite exactly when this runtime's own reader opens it: one verdict per file, asked by pullAll, // pushAll, soft pushAll, push and delete alike; a soft flush leaves the files that read as they are. The bytes and every runtime's verdict: tests/vectors/document-variants-c2a/readability. -test("SingleFileMessagePackBackingStore replaces a file exactly when it reads", async () => { +test("SingleFileMessagePackBackingStore rewrites a file exactly when it reads", async () => { const root = join(c2aVectors, "readability"); const rows = (await readFile(join(root, "manifest.txt"), "utf8")) .split(/\r?\n/u) @@ -2612,6 +2633,7 @@ test("SingleFileMessagePackBackingStore replaces a file exactly when it reads", ? store.push(envelopes[0]!) : store.delete(envelopes[0]!); if (reads) { + const heldBefore = operation === "pushAll" ? (await new SingleFileMessagePackBackingStore(file).pullAll()).map((entry) => entry.key) : []; await run(); if (operation === "pullAll") { continue; @@ -2628,7 +2650,8 @@ test("SingleFileMessagePackBackingStore replaces a file exactly when it reads", assert.equal((decode(await readFile(file)) as unknown[])[0], vector.includes("v3") ? "cultcache.store.v3" : "cultcache.store.v1", `${vector} ${operation}`); const keys = (await new SingleFileMessagePackBackingStore(file).pullAll()).map((entry) => entry.key); if (operation === "pushAll") { - assert.deepEqual(keys, envelopes.map((entry) => entry.key).sort(), `${vector} ${operation}`); + assert.deepEqual(envelopes.map((entry) => entry.key).filter((key) => !keys.includes(key)), [], `${vector} ${operation}: every written key is held`); + assert.deepEqual(heldBefore.filter((key) => !keys.includes(key)), [], `${vector} ${operation}: nothing else is removed`); } else if (operation === "push") { assert.ok(keys.includes(envelopes[0]!.key), `${vector} ${operation}`); } else { diff --git a/packages/cultcache-ts/test/store-file.test.ts b/packages/cultcache-ts/test/store-file.test.ts index eb41b77b..39436e30 100644 --- a/packages/cultcache-ts/test/store-file.test.ts +++ b/packages/cultcache-ts/test/store-file.test.ts @@ -670,7 +670,7 @@ test("delete removes only the record at that type and key", async () => { } }); -test("pushAll replaces the whole store, and a soft pushAll leaves an existing store alone", async () => { +test("pushAll writes onto the store as it is: it replaces by key and removes nothing, and a soft pushAll leaves an existing store alone", async () => { const directory = await scratch(); try { const path = join(directory, "store.cc"); @@ -681,7 +681,14 @@ test("pushAll replaces the whole store, and a soft pushAll leaves an existing st await store.pushAll([entryOf("b", "t", 2)], { soft: true }); assert.deepEqual([...(await readFile(path))], [...before], "soft leaves an existing store byte for byte"); await store.pushAll([entryOf("b", "t", 2)]); - assert.deepEqual(await snapshotOf(store), ["t/b:2"]); + assert.deepEqual(await snapshotOf(store), ["t/a:1", "t/b:2"], "a record the batch does not name stays"); + await store.pushAll([entryOf("a", "t", 7)]); + assert.deepEqual(await snapshotOf(store), ["t/a:7", "t/b:2"], "a record the batch names is replaced"); + await store.pushAll([]); + assert.deepEqual(await snapshotOf(store), ["t/a:7", "t/b:2"], "an empty flush from a cache that holds nothing wipes nothing"); + const before2 = [...(await readFile(path))]; + await assert.rejects(() => store.pushAll([entryOf("a", "u", 1)]), SchemaConflictError); + assert.deepEqual([...(await readFile(path))], before2, "a batch that retypes a held key is refused and the file is unchanged"); } finally { await rm(directory, { recursive: true, force: true }); } diff --git a/packages/cultmesh-ts/src/index.ts b/packages/cultmesh-ts/src/index.ts index 2f734cfd..e6f71e21 100644 --- a/packages/cultmesh-ts/src/index.ts +++ b/packages/cultmesh-ts/src/index.ts @@ -3933,7 +3933,7 @@ export class CultMeshNode { } public async flush(soft = false): Promise { - await this.store.pushAll?.(this.cache.snapshot(), { soft }); + await this.store.pushAll?.(this.cache.storedSnapshot(), { soft }); } } diff --git a/packages/cultmesh-ts/test/cultmesh.test.ts b/packages/cultmesh-ts/test/cultmesh.test.ts index 10de21b9..341a35b3 100644 --- a/packages/cultmesh-ts/test/cultmesh.test.ts +++ b/packages/cultmesh-ts/test/cultmesh.test.ts @@ -6,7 +6,7 @@ import test from "node:test"; import assert from "node:assert/strict"; import { decodeWireValue, encodeWireValue } from "@gamecult/cultnet-ts/wire"; import { z } from "zod"; -import { SchemaConflictError, defineDocumentType, documentPayloadBytes, documentPayloadFromBytes } from "@gamecult/cultcache-ts"; +import { SchemaConflictError, SingleFileMessagePackBackingStore, defineDocumentType, documentPayloadBytes, documentPayloadFromBytes } from "@gamecult/cultcache-ts"; import { CultNetDocumentRegistry, CultNetPeer, @@ -109,6 +109,34 @@ test("CultMesh TS opens a durable local node and registers document bindings", a assert.ok(reopened.documents.getBySchemaId("cultmesh.note.v0")); }); +test("CultMesh TS flush from a node that never pulled leaves the records the store holds", async () => { + const filePath = join(await mkdtemp(join(tmpdir(), "cultmesh-ts-flush-")), "node.ccmp"); + const first = await CultMesh.startNode(filePath, { documents: [noteDocument] }); + await first.put(noteDocument, "note:held", { noteId: "note:held", body: "already in the store" }); + await first.flush(); + + const unpulled = await CultMesh.startNode(filePath, { documents: [noteDocument], pullOnStart: false }); + await unpulled.flush(); + + const reopened = await CultMesh.startNode(filePath, { documents: [noteDocument] }); + assert.equal(reopened.getRequired(noteDocument, "note:held").body, "already in the store"); +}); + +test("CultMesh TS flush of a node that loaded a legacy-key global leaves the store reopenable", async () => { + const settings = defineDocumentType({ type: "cultmesh.settings", schema: z.object({ theme: z.string() }), global: true }); + const filePath = join(await mkdtemp(join(tmpdir(), "cultmesh-ts-legacy-flush-")), "node.ccmp"); + const seed = await CultMesh.startNode(join(await mkdtemp(join(tmpdir(), "cultmesh-ts-seed-")), "seed.ccmp"), { documents: [settings] }); + await seed.cache.putGlobal(settings, { theme: "iron" }); + const legacy = { ...seed.cache.getRequiredEnvelope(settings, "__global__"), key: "legacy" }; + await new SingleFileMessagePackBackingStore(filePath).push(legacy); + + const node = await CultMesh.startNode(filePath, { documents: [settings] }); + await node.flush(); + + const reopened = await CultMesh.startNode(filePath, { documents: [settings] }); + assert.deepEqual(reopened.cache.getRequiredGlobal(settings), { theme: "iron" }); +}); + test("CultMesh TS document handles hide local cache plumbing behind typed reactive reads", async () => { const filePath = join(await mkdtemp(join(tmpdir(), "cultmesh-ts-doc-")), "node.ccmp"); const node = await CultMesh.startNode(filePath, { From 2fd80c274b5f8691d360297f5cf95f2aeb2488e8 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 09:47:38 +0000 Subject: [PATCH 11/12] Rust mesh: a raw snapshot pull is one cache commit (mesh-snapshot-atomic) (#134) --- packages/cultcache-rs/CHANGELOG.md | 2 +- packages/cultmesh-rs/src/lib.rs | 40 ++++++++++++++++++++++-------- 2 files changed, 30 insertions(+), 12 deletions(-) diff --git a/packages/cultcache-rs/CHANGELOG.md b/packages/cultcache-rs/CHANGELOG.md index b43c13c6..e1fad67d 100644 --- a/packages/cultcache-rs/CHANGELOG.md +++ b/packages/cultcache-rs/CHANGELOG.md @@ -11,7 +11,7 @@ only in git history. Semver: this is a breaking change in the 0.x line (0.3.x to 0.4.0). - A record whose payload carries unread slots is read-only. A payload with a slot past its type's declared slots, at any depth (record, nested record, list element, union arm, unit arm), still decodes, and every write or removal of its key is now refused before any store is touched: `put`, `put_envelope`, `put_raw_envelope`, `put_prepared_batch` (the whole batch) and `delete`, so `update` too. The refusal is `SchemaConflictError` with the new `unread_slots`; its message names the type and both slot counts and never a slot value. Holders of such a record used to overwrite it and drop the extra slots silently. The payload nesting refusal still runs first. A held payload this build cannot read at all (an arm it lacks from a newer writer, or a corrupt record) is read-only too: every write or removal is refused with an error naming the type and key (a newer writer's record is read with that build; a corrupt one is restored from a backup or removed at the store itself, as no cache call removes it), and `is_writable` answers false. `load_soa` is a write of every held key it omits or changes, so it refuses to drop or replace a read-only record. `is_writable` also answers false for a key a store carries as a foreign record and for a type other than the one holding the key, as the writes do. Each write over a held key decodes the held payload once to ask (about 40% of a `put` of a 200,000-element record, measured in release); keys nothing holds pay nothing. -- cultmesh-rs `pull_rudp_catalog_snapshot` applies documents one by one with `put_raw_envelope`, so a snapshot document for a key the node holds read-only ends the pull at that document and the documents before it stay applied. C# applies a raw snapshot as one cache commit, so there a refusal applies nothing: this is a divergence, owed by follow-up `mesh-snapshot-atomic` (stage the pull through `put_prepared_batch`). +- cultmesh-rs `pull_rudp_catalog_snapshot` applies a snapshot as one cache commit (`put_prepared_batch`), as C# does: a snapshot document for a key the node holds read-only, or any other refusal, applies nothing. It used to apply document by document, so a refusal ended the pull with the earlier documents kept. A snapshot whose types route to different backing stores is now refused whole, as a batch cannot span stores. - `SchemaConflictError` gains the public field `unread_slots: Option`; code that builds it with a struct literal must set it (`None`). - A unit arm with slots (`[N, [extra...]]`) decodes as the unit arm and marks the record, where the derive refused it. - The registration map's value changes from the schema name to a private struct holding the schema name and the unread-slots check. diff --git a/packages/cultmesh-rs/src/lib.rs b/packages/cultmesh-rs/src/lib.rs index 18a02fcb..310629e3 100644 --- a/packages/cultmesh-rs/src/lib.rs +++ b/packages/cultmesh-rs/src/lib.rs @@ -735,30 +735,33 @@ impl CultMeshNode { self.apply_raw_snapshot_documents(documents) } - /// Applies each snapshot document of a registered schema with `put_raw_envelope`. The first refusal ends the pull - /// and the documents before it stay applied; a key this node holds read-only (unread slots from a newer peer, or a - /// payload this build cannot read) is refused like any other. C# applies a raw snapshot as one cache commit, so a - /// refusal there applies nothing: this is a divergence, owed by follow-up mesh-snapshot-atomic (stage the pull through - /// `put_prepared_batch`). + /// Applies the snapshot documents as one cache commit, as C# does: a refusal (a key this node holds read-only because + /// of unread slots from a newer peer or a payload this build cannot read, or any other the cache judges) applies + /// nothing. A schema id the registry does not know is skipped, except one with a `.vN` suffix, which is taken as + /// registered and refuses the commit when its type is not (C# throws on an unknown schema too). The commit lands in + /// one backing store, so a snapshot spanning stores is refused whole. fn apply_raw_snapshot_documents( &mut self, documents: Vec, ) -> Result { - let mut applied = 0usize; + let mut staged = Vec::new(); for document in documents { let Some(document_type) = registered_document_type_for_schema(&self.documents, &document.schema_id) else { continue; }; - self.cache.put_raw_envelope(CultCacheEnvelope { + staged.push(CultCacheEnvelope { key: document.record_key, r#type: document_type, payload: document.payload, stored_at: document.stored_at, schema_id: Some(document.schema_id), - })?; - applied += 1; + }); + } + let applied = staged.len(); + if applied > 0 { + self.cache.put_prepared_batch(staged)?; } Ok(applied) } @@ -1226,7 +1229,7 @@ mod tests { } #[test] - fn a_snapshot_document_for_a_read_only_key_ends_the_pull_unlike_csharp_which_applies_nothing() -> Result<()> { + fn a_snapshot_document_for_a_read_only_key_refuses_the_whole_pull_as_csharp_does() -> Result<()> { let temp = tempfile::tempdir()?; let mut node = CultMesh::create_node(temp.path().join("m.cc"), TestDocuments, Default::default())?; // A newer peer's note: one slot past the declared one. @@ -1246,12 +1249,27 @@ mod tests { assert_eq!(conflict.record_key, "held"); assert!(conflict.unread_slots.is_some()); assert!(!error.to_string().contains("extra slot")); - assert_eq!(node.get_required::("before")?.body, "b"); + assert!(node.get::("before")?.is_none()); assert_eq!(node.get_required::("held")?.body, "newer body"); assert!(node.get::("after")?.is_none()); Ok(()) } + #[test] + fn a_snapshot_applied_to_a_stale_node_keeps_the_records_another_writer_added() -> Result<()> { + let temp = tempfile::tempdir()?; + let path = temp.path().join("m.cc"); + let mut node = CultMesh::create_node(&path, TestDocuments, Default::default())?; + let mut other = CultMesh::create_node(&path, TestDocuments, Default::default())?; + other.put("added", &Note { body: "other writer".to_string() })?; + let plain = |body: &str| rmp_serde::to_vec(&(body,)); + assert_eq!(node.apply_raw_snapshot_documents(vec![raw_note("snap", plain("s")?)])?, 1); + let reopened = CultMesh::create_node(&path, TestDocuments, Default::default())?; + assert_eq!(reopened.get_required::("added")?.body, "other writer"); + assert_eq!(reopened.get_required::("snap")?.body, "s"); + Ok(()) + } + #[test] fn node_round_trips_registered_documents() -> Result<()> { let temp = tempfile::tempdir()?; From a23ac7755acff318b134300b7003d2383d9f28c3 Mon Sep 17 00:00:00 2001 From: Libby Date: Thu, 12 Nov 2026 10:31:13 +0000 Subject: [PATCH 12/12] Rust mesh: pin that a snapshot spanning backing stores is refused whole (#137) --- packages/cultcache-rs/CHANGELOG.md | 2 +- packages/cultmesh-rs/src/lib.rs | 48 +++++++++++++++++++++++++++++- 2 files changed, 48 insertions(+), 2 deletions(-) diff --git a/packages/cultcache-rs/CHANGELOG.md b/packages/cultcache-rs/CHANGELOG.md index e1fad67d..7f3caa73 100644 --- a/packages/cultcache-rs/CHANGELOG.md +++ b/packages/cultcache-rs/CHANGELOG.md @@ -11,7 +11,7 @@ only in git history. Semver: this is a breaking change in the 0.x line (0.3.x to 0.4.0). - A record whose payload carries unread slots is read-only. A payload with a slot past its type's declared slots, at any depth (record, nested record, list element, union arm, unit arm), still decodes, and every write or removal of its key is now refused before any store is touched: `put`, `put_envelope`, `put_raw_envelope`, `put_prepared_batch` (the whole batch) and `delete`, so `update` too. The refusal is `SchemaConflictError` with the new `unread_slots`; its message names the type and both slot counts and never a slot value. Holders of such a record used to overwrite it and drop the extra slots silently. The payload nesting refusal still runs first. A held payload this build cannot read at all (an arm it lacks from a newer writer, or a corrupt record) is read-only too: every write or removal is refused with an error naming the type and key (a newer writer's record is read with that build; a corrupt one is restored from a backup or removed at the store itself, as no cache call removes it), and `is_writable` answers false. `load_soa` is a write of every held key it omits or changes, so it refuses to drop or replace a read-only record. `is_writable` also answers false for a key a store carries as a foreign record and for a type other than the one holding the key, as the writes do. Each write over a held key decodes the held payload once to ask (about 40% of a `put` of a 200,000-element record, measured in release); keys nothing holds pay nothing. -- cultmesh-rs `pull_rudp_catalog_snapshot` applies a snapshot as one cache commit (`put_prepared_batch`), as C# does: a snapshot document for a key the node holds read-only, or any other refusal, applies nothing. It used to apply document by document, so a refusal ended the pull with the earlier documents kept. A snapshot whose types route to different backing stores is now refused whole, as a batch cannot span stores. +- cultmesh-rs `pull_rudp_catalog_snapshot` applies a snapshot as one cache commit (`put_prepared_batch`), as C# does: a snapshot document for a key the node holds read-only, or any other refusal, applies nothing. It used to apply document by document, so a refusal ended the pull with the earlier documents kept. A snapshot whose types route to different backing stores is now refused whole, as a batch cannot span stores. C# `CultCache.Commit` refuses a cross-store commit the same way (`The batch spans A and B; a commit lands in one home store`), so the two agree; a test pins it. - `SchemaConflictError` gains the public field `unread_slots: Option`; code that builds it with a struct literal must set it (`None`). - A unit arm with slots (`[N, [extra...]]`) decodes as the unit arm and marks the record, where the derive refused it. - The registration map's value changes from the schema name to a private struct holding the schema name and the unread-slots check. diff --git a/packages/cultmesh-rs/src/lib.rs b/packages/cultmesh-rs/src/lib.rs index 310629e3..b1ee613b 100644 --- a/packages/cultmesh-rs/src/lib.rs +++ b/packages/cultmesh-rs/src/lib.rs @@ -739,7 +739,8 @@ impl CultMeshNode { /// of unread slots from a newer peer or a payload this build cannot read, or any other the cache judges) applies /// nothing. A schema id the registry does not know is skipped, except one with a `.vN` suffix, which is taken as /// registered and refuses the commit when its type is not (C# throws on an unknown schema too). The commit lands in - /// one backing store, so a snapshot spanning stores is refused whole. + /// one backing store, so a snapshot spanning stores is refused whole; C# `CultCache.Commit` refuses a cross-store batch too + /// ("The batch spans A and B; a commit lands in one home store"), so the two agree. fn apply_raw_snapshot_documents( &mut self, documents: Vec, @@ -1210,10 +1211,22 @@ mod tests { body: String, } + #[derive(Clone, Debug, PartialEq, Eq, cultcache_rs::DatabaseEntry)] + #[cultcache(type = "cultmesh.test.memo", schema = "CultMeshTestMemo")] + struct Memo { + #[cultcache(key = 0)] + body: String, + } + cultmesh_documents!(TestDocuments { Note => "cultmesh.test.note.v0", }); + cultmesh_documents!(TwoTypeDocuments { + Note => "cultmesh.test.note.v0", + Memo => "cultmesh.test.memo.v0", + }); + fn raw_note(key: &str, payload: Vec) -> cultnet_rs::CultNetRawDocumentRecord { cultnet_rs::CultNetRawDocumentRecord { schema_id: "cultmesh.test.note.v0".to_string(), @@ -1255,6 +1268,39 @@ mod tests { Ok(()) } + #[test] + fn a_snapshot_spanning_two_backing_stores_is_refused_whole_and_applies_nothing() -> Result<()> { + let temp = tempfile::tempdir()?; + let mut node = CultMesh::create_node(temp.path().join("m.cc"), TwoTypeDocuments, Default::default())?; + // A node has one store through the public API; route Memo to a second one by hand (C# refuses the same commit). + node.cache.add_backing_store( + SingleFileMessagePackBackingStore::new(temp.path().join("memo.cc")), + [Memo::TYPE], + )?; + node.cache.pull_all_backing_stores()?; + let plain = |body: &str| rmp_serde::to_vec(&(body,)); + let memo = |key: &str, payload| { + let mut record = raw_note(key, payload); + record.schema_id = "cultmesh.test.memo.v0".to_string(); + record + }; + let error = node + .apply_raw_snapshot_documents(vec![ + raw_note("n", plain("note")?), + memo("m", plain("memo")?), + ]) + .unwrap_err(); + assert!(error.to_string().contains("cannot span backing stores"), "{error}"); + assert!(node.get::("n")?.is_none()); + assert!(node.get::("m")?.is_none()); + assert!(!temp.path().join("m.cc").exists() && !temp.path().join("memo.cc").exists()); + + // Each store alone still takes its own snapshot. + assert_eq!(node.apply_raw_snapshot_documents(vec![memo("m", plain("memo")?)])?, 1); + assert_eq!(node.get_required::("m")?.body, "memo"); + Ok(()) + } + #[test] fn a_snapshot_applied_to_a_stale_node_keeps_the_records_another_writer_added() -> Result<()> { let temp = tempfile::tempdir()?;