C2a F2, operator: "CultCache never destroys or relabels a record it doesn't
understand; keeps it byte-for-byte, refuses writes that would overwrite it."
- The registry's claim decision (Claimant) is split out of resolution. It
never decodes or compares shapes, so a store classifies a record from its
header. The last branch, "No local CultCache schema matches", now means
foreign instead of a refused load.
- Both store kinds keep a foreign record out of the cache and list it in
CacheBackingStore.ForeignRecords (key, schema id, catalog name, storedAt).
- Single-file store: a whole-view write lays back every foreign record the
file holds now, with its catalog entry, and keeps a marked header while it
carries one (it cannot see their ids). A staged write, a commit (onto the
file or whole view) and a whole view over a key the file holds as foreign
refuse with CultSchemaConflictException naming the key and id. A merge
judgement never decodes a foreign record, and one that replaced a held
record takes it away.
- Directory store: a foreign record's page is never read; a generation that
would replace or remove one refuses before any page is written.
- Contracts: hard rejection no longer lists "no compatible local schema
candidate"; the persistence format says what a foreign record is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A directory-store record's identity on disk is its manifest entry, which is what
a commit condition compares; the page holds its own copy of the id. A load now
takes StoredSchemaId from the manifest, so a manifest rewritten under an older
id (AConditionOnARecordUnderADeclaredOldIdHoldsInTheDirectoryStore) is the id
the condition expects. Canonicalisation had hidden the page and manifest
disagreeing in that fixture.
The whole-view test's variant overrides its name: a variant may not share its
base's.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each stored document carries StoredSchemaId: the id its record sits under in its
store, the one it was loaded under until a write puts it under its type's own id.
Expect, ExpectUnchanged, the reload and merge identity checks compare
(StoredSchemaId, storedAt) with the store's bytes exactly. CanonicalSchemaId is
deleted: canonicalising the disk id let an older version's whole-view rewrite
(same storedAt, restamped to its own id, a member shed) pass a newer cache's
condition (Soul F1), and failed every condition on a record resolved by name or
through a file entry listing the local id (F3). A whole-view write marks every
record it wrote as stored under its type's id, in the store and the cache.
Tests: Soul's rolling-deploy probe as AWholeViewRewriteByAnOlderVersionFails...
(Mismatch for Expect and ExpectUnchanged, nothing derived lands, a pull shows the
rewrite and the retry commits); F3's H1, H1b, H2 pass Expect and ExpectUnchanged
when nothing changed; conditions after a whole-view commit hold for a plain
record and a variant. BareDeck is emitted at run time (EmittedDocumentTypes) so
the shared registry never sees it beside DeclaringDeck; the CanonicalSchemaId
tests and their Claimant fixtures go with the function.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CultDocumentAttribute.CompatibleSchemaIds, as TypeScript and Python already name it. The
registered catalog entry lists the ids after its own, the registry resolves a record under one
to the type (its own id first, then the first registered type that declares it), and a
condition on a record compares schema ids through that resolution, so a record still under a
declared old id counts as the record the cache holds. The cache already stamps every record it
writes with the descriptor's own id; the store writers still change no identity, so a record a
commit does not rewrite stays under the id the file holds it under, published by the registered
entry. Without the declaration such a commit is refused, typed, as before.
Contract and XML docs describe the declaration.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Deleted the heuristic that rewrote a record the cache held "unchanged since read" (same key and
storedAt) under the registered id, in the single-file commit and in the directory generation. It
could overwrite a record another writer changed while keeping its storedAt. A commit onto a file
now writes its batch and leaves every other record as the file holds it; a record under an id only
an arrived entry lists is refused, typed, with the file left as it was. Deleted the directory
store's _durableCatalog, a second stale-able source of arrived entries beside the manifest read
under the lease.
Tests: the lost-update probe, the two refusals that replace the restamp tests, the
arrived-overwrites-registered guard in both id orders, the content-hash tie-break against the
compatible-list order, and refusal determinism across record arrival order.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Every writer (C# single-file and directory, TypeScript, Python; Rust already did) chooses, for each schema id a record carries, one
entry that publishes it: the entry that owns the id (a registered descriptor, else an arrived entry), else one that lists it as
compatible. A registered descriptor wins over an arrived entry with the same own id. Entries are written as chosen, with no union of
compatible ids and no dependence on arrival order; an entry no record needs is not written. Two entries of one tier that share an own id
and disagree on the schema name, records of different types under one id, or a record no chosen entry publishes refuse the write with a
typed error (CultSchemaConflictException, SchemaConflictError) naming the id, the names and a record key, and the store is left as it was.
CultSchemaCatalogEntry.MergeById and the Python and TypeScript merge paths are gone. A cache-held record unchanged since the file was read
is rewritten under its registered id (C# single-file and directory), so an entry that lists an older id is not needed to publish it.
Changelogs for cultnet-py (removed export) and cultcache-py.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Entries that share a schema id are one schema seen by different writers, so the writer lists every id any of them lists, in any
order: C# single-file and directory (CultSchemaCatalogEntry.MergeById), TypeScript and Python. Rust synthesises one entry per id and
has no such merge. An id that one entry owns and another lists as compatible names the owner in Rust, TypeScript, Python and the TS
inspector, as C# already did (own-id-over-compatible-v3). Also: a directory manifest in an unreadable format is the typed refusal
naming it, cultnet-py applies and reports a record under the local schema id, and Rust's error prints its cause once.
Vectors for the Mesh layout discriminator: a nil and an integer at catalog slot 5.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The single-file and directory writers dropped an entry whose own id no record carried, leaving a record under a compatible id with
no catalog entry that publishes it: the same store nothing could open. An entry now stays while it publishes any id a record
carries. The conditional-commit readability test over compatible-id-only-v3 pins it.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A record whose schema id the catalog lists only as a compatible id now opens (the cache resolved a persisted schema by entry id
alone). CultMesh's fallback to the older catalog-entry reader fires only for a store in that layout, decided by the entry's slot 5,
not for any decode failure, so a current-layout store with a malformed slot is refused as CultCache refuses it.
CultStoreUnreadableException.Path names the refused file. New vectors: compatible-id-only-v3 (accepted by all four) and three
current-layout catalogs with one malformed slot. The contract states the writer's invariant.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
C# read such a store on flush and commit and refused it on open, with an untyped InvalidOperationException from the cache. The
store reader now refuses it with CultStoreUnreadableException, so open, flush and commit give one verdict, as Rust, TypeScript
and Python already do. The readability open test builds a cache with a codec, so a variant store opens as it would in use.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
CultStoreUnreadableException (a NotSupportedException) is the one refusal for a store file this runtime cannot read, thrown
by the single-file reader, the rewriting paths and CultMesh's reads, with the cause as the inner exception. The reader is exactly
one array of three slots (or an empty array): a header alone and an extra top-level slot are refused instead of read as empty or
skipped. CultMesh's older-catalog fallback is the same reader with the older entry reader, so it also requires one array.
Vectors: bad-body, header-only, missing-schema, variant-slot, extra-top-slot in the readability set, and a legacy-catalog store
with and without bytes after it for CultMesh. The persistence contract says what is true.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The Phacelle merge added the public CultMath value type CultPhasor without
its Caching adapters, failing the every-public-type tests. Register it the
way CultCellular is: {"cos","sin"} objects in JSON (which also gives the
dictionary-key form) and a [cos, sin] array in MessagePack. The adapters are
Caching code and stay MIT; only CultPhasor.cs itself is MPL-2.0.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A rewrite replaces a file only when the runtime's own reader opens it. C#
drops ReadStoreHeader (a second, header-only verdict that let a flush replace
files the reader refuses) and asks DeserializeSnapshot from open, flush and
commit; the reader now requires exactly one array, so C# refuses bytes after
the store on read as TS and Python do. A zero-byte file is an empty store for
all three paths. CultMesh.WriteSingleFileDocument reads the file it replaces
through the Mesh reader before writing. Rust reads and rewrites through one
classification (trailing bytes refused on read; a non-store first slot must be a
legacy envelope array the reader accepts). TypeScript checks a legacy array with
pullAll's own decoder before replacing it. Python already conformed.
Shared byte vectors and per-runtime verdicts live in
tests/vectors/document-variants-c2a/readability; every runtime's test reads
them. Holds gains second-dictionary-value and second-list-item cases. The
persistence format contract now marks v3 by content and states the readability
rule.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
CultElementIds.Holds is the one fact a store marks its header by; a test now
drives it directly over a dictionary value, a list of lists, a list under an
unset element and an element, and over unset ids that hold nothing, so a
mutant that skips any of those branches fails.
C# ReadStoreHeader refused a first-slot string without the cultcache.store.
prefix and then ran RequireSingleFileFormat, which refuses every string that
is not v1, v2 or v3 anyway. The prefix check is gone; a non-header string is
refused as a format this runtime does not read.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A rewrite read the header of the file it replaces but not whether the file was
a store: C# ReadStoreHeader skipped one value and ignored bytes after it, and
took a non-array value or an array whose first slot is a string that is not a
store header for a headerless legacy file (so did the C# Mesh raw write, which
uses it); TypeScript's pushAll took any non-array value or non-header first slot
for v1; Rust ignored bytes after the array.
The rule, one per runtime: a file is a store on rewrite only when it is exactly
one complete MessagePack array. Bytes after it, a value that is not an array,
and a first slot that is a string but not a header this runtime reads are
refused, and the file is left untouched. An empty file and a legacy envelope
array (first slot not a string, empty included) carry no header and are still
rewritten.
- C#: ReadStoreHeader refuses non-array values, trailing bytes and a non-header
first string, with a NotSupportedException naming which.
- TypeScript: storeHeader (store-format.ts) decides the rewrite header;
@msgpack/msgpack's decode already refuses trailing bytes.
- Rust: store_header(bytes, whole) walks with a cursor; the rewrite path asks
for the whole file, the read paths keep reading as before.
- Python: push_all already reads the whole file (unpackb refuses trailing
bytes, the legacy decoder refuses anything but a list of maps); pinned by a
test.
Tests in each runtime run the same file matrix (truncated, trailing, scalar,
map, string first; empty, empty array, legacy), plus a C# unconditional commit
over a store with trailing bytes and the C# Mesh raw write.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SingleFileBackingStore.PushAll replaced the file without reading it, so a
truncated store or one in a format this runtime does not read was clobbered,
where Rust, TypeScript and Python refuse. PushAll now reads the header of a
non-empty file under the lock before it writes; a file it cannot read throws
and is left as it was. An empty file and a legacy envelope array carry no
header and are still replaced.
The header is read by the store's codec: SingleFileBackingStore gains a
protected virtual ReadStoreHeader (default: read the whole snapshot), and the
MessagePack store overrides it with
CultDocumentMessagePackSerialization.ReadStoreHeader. An unconditional commit
reads the file it rewrites through the same check.
Doc text: last-writer-wins now says it holds only over a file the writer can
read (CultCache.cs PushAll comment, cultcache-persistence-format.md
"Conditional commit is the only safe multi-process write" and writer rule 4,
cultcache-store-composition.md "A plain flush and an unconditional commit are
last-writer-wins").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One conflict, at the end of Admit in CultCache.cs. The branch had added the
IdsInMemoryOnly marking of a load's minted ids before Apply; main had turned
the thread's hold into a Hold carrying Changes and Failures and run the
journals right after Apply. Both are kept, in that order: the marking touches
only the records and runs no user code, then Apply mints the Sequences, the
hold records the changes, and the journals run under the gate.
Id planning and Project do not collide. Ids are planned and applied first,
Resolve (variants, then Project, then the name and index rules) reads the
filled documents, land runs, and only then Apply mints. A Project getter that
throws is caught by the same handler that undoes the id plans, so a refused
admission changes no document and mints nothing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ExistingHeader read the whole snapshot and swallowed every failure into "no header": a
truncated v3 file got a v1 raw write and an unknown v9 file was overwritten, where
TypeScript, Python and Rust refuse. CultDocumentMessagePackSerialization.ReadStoreHeader
reads the prefix header, refuses a file that is not one complete MessagePack value or whose
format this runtime cannot read, and the raw write keeps a whole marked file marked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CacheBackingStore.HeaderFor(holdsIds, existingHeader, wholeStore, directoryStore) replaces FormatFor,
DirectoryFormatFor and the writers' private marker logic. A writer that rewrites the whole store
decides by content; a writer that touches part of it keeps a header already marked. Flush,
unconditional and conditional commit, the directory store and CultMesh's single-file write all call
it. A typed CultMesh write replaces the file with a document it can see, so it decides by content;
a raw payload is opaque and keeps a marked file marked.
A record's HoldsIds is what its stored form persists: a load keeps a variant's delta as read, so
the ids a load mints in an override exist only in memory and no longer mark the store. A
conditional commit onto the file no longer reads this cache's view of what the file held, so a
stale view cannot over-mark it. IdPlan.HoldsIds is derived from HeldIds and the fills.
Reverts InternalsVisibleTo GameCult.Caching.MessagePack and the public DirectoryFormatV4/V5
constants: the directory store reaches the decision through the protected member it inherits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A flattened variant's HoldsIds is read off the document re-resolved against the post-batch base, not carried from the
variant or from the document the flatten was planned against. The conditional commit onto a file decides by the same
rule as the unconditional commit (ids in the records the file will hold, ids only this cache minted excepted, a marked
file staying marked). The directory manifest's header choice moves beside FormatFor. Pins the three-batch flatten
probe on the commit and flush paths, the conditional commit, the last id leaving, an id-bearing override that is not
the last, and a failed CultMesh write giving its ids back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A store is written v3 (v5 for a directory) only when a record in it holds a non-empty
element id. CultElementIds.Plan reports that as IdPlan.HoldsIds; the admission stamps it
on the stored record and every store reads it. The descriptor's CarriesElementIds is
deleted. A file already marked on disk stays marked. CultPersistedStoreSnapshot.FormatFor
is the one header decision.
Plan refuses one element object sitting twice in a list at write, naming the record and
list, instead of writing both ids and refusing the store at load.
CultMesh single-file write mints, refuses and marks as a cache write does; the read applies
the load-time plan, so a write and a load refuse the same things.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Named literals now follow AllowReadingFromString and WriteAsString as
System.Text.Json's float does; quoted numbers refuse whitespace;
UnmappedMemberHandling.Disallow and AllowDuplicateProperties=false are honoured
and a repeated property replaces the whole value; dictionary keys read only the
canonical spelling. Components live in stack spans, property names are
pre-encoded, and there is no per-property closure, so a float3 read allocates
nothing (424 bytes before). Tests pin unknown nested values, de-DE and sv-SE
style cultures, and a distinct negative-zero key.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The id rule is one plan, made before anything is written: Plan decides every fill and
refuses a duplicate, a random id that is not 12 lowercase hex characters, and a derived
id whose source is null or empty, naming the record and list (CultElementIdException).
A load refuses what a write refuses. Apply and Undo make minting part of the admission:
a refused or uncommitted write leaves the caller's documents as they were.
Override values mint inside Resolve, into the decoded copy, so the public
CultVariantOverride.Set, C1-era overrides and Override() all get ids on write and on load;
Override() no longer mints. Interface unions are judged on their concrete types, the same
rule registration and minting share, and a subtype reusing the id's key is refused at
registration. Dictionary keys format with the invariant culture.
Records minted at load are marked per entry (IdsInMemoryOnly) instead of listed by key, so
a deleted record leaves the mark with it, and a flush of the whole store clears it. A store
whose types reach an object list declares cultcache.store.v3 (directory: v5), which every
reader older than ids refuses; a marked file stays marked when a writer merges onto it.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
One Scalar plus Node description per type now drives reading, writing and the
dictionary-key spelling, replacing the per-value JsonDocument parse. NaN and the
infinities follow JsonNumberHandling (named literals or JsonException), floats
beyond float.MaxValue are refused instead of becoming infinity, out-of-range
components throw JsonException instead of FormatException, property matching
honours PropertyNameCaseInsensitive, and every type works as a dictionary key
spelled as comma-joined components.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
CultMathJson is the JSON twin of CultMathResolver: vectors, quaternion and
Color32 are objects keyed by field name, matrices are arrays of row objects,
rect is {min,max}. Consumers add them with AddCultMathConverters(); CultLib owns
no JSON options. CultCircle and CultSphere serialize and deserialize through
them, with [JsonConstructor] so the struct constructor is used. The Caching
README documents each MessagePack and JSON shape in one table.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Golden bytes now cover double2, double3, int3, int4, bool3, bool4, quaternion,
Color32, Random, float3x3 and CultCellular. A missing Color32 alpha decodes as
255, the constructor default. A test proves a consumer resolver registered for
the assembly wins over CultMathResolver. The redundant nil check is deleted;
ReadArrayHeader already refuses nil.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
GameCult.Geometry no longer carries vector or rectangle values. CultCircle and
CultSphere hold float2/float3, the isosurface and surface-nets extractors take
float3 origin, and the per-call conversion to float3 is gone. Their MessagePack
bytes are unchanged (pinned against the pre-collapse encodings), now produced
by CultMathResolver through the default CultCache options.
This is a breaking change to the GameCult.Geometry public API.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The first commit's resolver needs the CultMath project reference and scoped
span parameters to compile; the golden test file also had two literal slips.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
CultMathResolver formats float2/3/4, double2/3, int2/3/4, bool2/3/4,
quaternion, Color32, Random, float2x2, float3x3, rect and CultCellular as
component arrays, and sits in the default CultCache resolver chain so no
consumer registers anything. It lives in GameCult.Caching.MessagePack, which
already owns that chain; CultMath core stays dependency-free.
Vector, rect and Aetheria-formatter byte shapes are pinned by golden tests, a
reflection test fails when a public CultMath value type has no formatter, and
decode skips extra elements and refuses nil.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The held index kept one key per value, last writer winning, and RemoveKey dropped
a value outright. Two consequences: removing or changing one of two plain
holders left the other unfindable, and the variant index rule judged a variant
against that lossy view, so a variant could land on a value a plain record still
held and the store then refused to reopen.
- _indexes maps a value to the keys holding it, oldest first. Index, Unindex and
the removal maintain the set. GetByIndex finds the most recently indexed
holder of a value, so removing the winner finds the other (behaviour change for
plain duplicates: it used to find nothing).
- One rule, RefuseVariantIndexSharing, serves both write orders. It reads the
truthful index minus what the commit replaces or removes, plus what it lands.
ResolveOne no longer judges indexes; ClaimIncomingIndexValues, HeldByVariants
and RefusePlainSharingWithVariant are gone.
- A conditional commit merged onto the file is judged by the cache's own
resolution (CacheBackingStore.Judging) on the merged set: records other writers
added, records they removed, and the batch. Another writer's variant sharing a
value with the batch, or a base removal that strands another writer's variant,
is refused with the named refusal instead of writing an unopenable store.
- Shedding is one decision at load. ToStoredDocument resolves the schema once and
returns the report with it; a variant's override of a slot the type lacks is
dropped from the delta there, so flush needs no predicate and every store
reports through the same surface (ReportVariantSheds and OverridesTheTypeHas
are gone). The directory store refuses variants and loads through the same call.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
F3: an override drifts exactly as the same member would on a plain record. A
slot the type dropped is ignored and reported by the store's schema resolution
(LastSchemaMigrationReports), the one surface a plain record uses; a value that
no longer decodes refuses the load, naming the variant and member. The second
report surface (VariantMigrationReports, the per-variant report map, the
catalog slot-name plumbing) is deleted.
F4: an override for a slot the type does not have, or a value that does not
decode as the member, is refused at write. Drift is for type changes after the
fact, not for bad writes. F5: two overrides of one slot are refused.
F2: a variant that shares an indexed value with another record is refused,
naming both keys and the index (the R6 rule extended to [CultIndex]); plain
record duplicate handling is unchanged.
F7: a flatten now resolves against the same batch, so a base edited in that
batch is in the flattened record; and the directory store refuses to load a
variant page as it refuses to write one.
The C1 vector's big variant now overrides its indexed code, as the unique-index
rule requires.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
A variant record stores [baseKey, overrides[]] in record slot 4 and an empty
payload; the cache resolves it against its base inside the hold that admits it or
any record on its base chain, and re-resolves dependents in the same Apply, so a
base change reaches its variants with contiguous sequences and no stamp. Plain
Upsert at a variant key is refused (Flatten is the explicit way); cycle, missing
base, different concrete type, evicted base with survivors and a name that
shadows the base are refused naming the keys. Overrides of a dropped slot or an
undecodable value are ignored with a warning and reported as compatible drift.
The store writes cultcache.store.v2 only when it holds a variant and reads v1 and
v2. The codec pair CultInspectorModel took becomes CultCodec, shared with the
cache, which needs it to clone documents and overlay override values. The
directory store, CultMesh single-file reads and CultNet row building refuse a
variant by name until their own cuts.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Delete the trailing Skip loop in the record reader, the ?? v1 header default,
and the mesh reader's copy of the record decoder. The single-file store and
CultMesh single-file reads accept cultcache.store.v1 only. Persistence-format
and parity docs state the rule.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
452f928 made nil the unset wire form, including dictionary keys: a
Dictionary<CultRecordRef<H>, float> holding the empty key encoded it as C0.
Python msgpack refuses nil map keys under strict_map_key, its default, so
cultcache-py cannot decode such a record. Nil cannot be the one canonical
form; "" can, and every runtime reads it.
The writer always writes "". The reader still takes nil as unset, because
452f928/b85a828 stores and Aetheria's Cut 9 import wrote nil; such a record
converts to "" on its next save and is byte-stable after that. The
store-composition contract names "" for value and key position, requires
readers to accept nil, and tells other runtimes to write "".
The tests pin payload bytes, so each claim is consequential: an unset ref
writes "", reloads through a fresh cache and resaves byte-identical; a
nil-written legacy payload reads as unset and resaves as ""; a ref-keyed
dictionary with an empty key writes "", reads back unset and resaves
byte-identical. The nil writer fails every byte assertion, a reader taking
nil as a set key fails the legacy pair, and a reader taking "" as a set key
fails the read-back assertions.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The formatter wrote default(CultRecordRef<T>) as nil but read nil back as
"", so re-upserting a loaded record changed its bytes (0xC0 became 0xA0).
One in-memory unset value: CultRecordKey.Value is never null, so
default(CultRecordRef<T>) equals a ref built from "". One wire form: an
empty key writes nil, and nil or a legacy "" reads back as default.
Load then save is byte-stable; a store holding "" converges to nil on its
next save.
Rust, TypeScript and Python have no reference type: a reference member is a
string in the consumer's own payload, so there is no sibling codec to align.
The store-composition contract now names nil as the one unset wire form.
The new test loads a record with an unset ref through a fresh cache,
re-upserts and flushes, and requires identical payload bytes; it fails on
the old formatter.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
An exception thrown inside a `when` filter is swallowed as false, so a manifest
re-read failing mid-File.Replace turned a race that should retry into a thrown
torn-page error. The re-read now runs in the handler, and a failed re-read
counts as moved. Parallel.For's AggregateException is unwrapped where the pages
are read, so the page's own failure propagates and the torn check no longer
walks wrappers.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Key identity serialized a key with the options of the key type's own
assembly, but the store writes it inside the document with the document
assembly's options, [CultCacheFormatterResolver] resolvers included. A key
that failed to serialize fell back to comparison by ToString.
The inspector codec now takes the owning document type, ReplaceKey and
FreshKey take it, and a key serializes as its declared key type under that
document's options, the store's own path. A key the document cannot
serialize is refused with a notice; the text fallback is deleted.
The comment and README said "the dictionary's own comparer"; it is the
rebuilt dictionary type's default comparer, and now says so.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A missing or hash-failing page under an unchanged manifest is corruption,
not a writer mid-commit, but every InvalidDataException was retried five
times and reported as "did not settle" with the real cause buried. The
retry now also requires the manifest bytes to have changed; otherwise the
corruption error is thrown on the first attempt. A vanished page directory
is a race under the same condition.
The reload test's writer now commits under every unleased read, so it only
settles if the retry takes the lease that commit created.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A pull takes the commit lease only when .commit.lock exists. Without it
(no writer has created one yet, or the store was copied without it) the
reader loaded with no lease, and a writer committing between its manifest
read and its page reads left it with a vanished page or a stale manifest.
An unleased pull now keeps its load only if the manifest bytes it read are
still current afterwards; a changed manifest, or a named page that is
missing or fails its content hash, reloads it, at most five times, and a
pull that has not settled by then throws. A retry takes the lease once a
writer has created it. Opening still creates nothing. The lease comment no
longer claims a missing lock means no writer ever committed, and the store
composition contract states the rule.
The test deletes the lock from a seeded store and commits from a second
cache through an internal seam between the manifest read and the page
reads, once replacing the page the reader named and once only adding a key.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A refused key and an unavailable fresh key are now the model's answer:
ReplaceKey returns the kept key with the notice text, FreshKey returns
null with its notice, and IMGUI only shows them.
A key is taken when another serializes the same or when the dictionary's
own comparer calls them equal, so 0.0 against -0.0 is refused instead of
throwing in BuildDictionary.
Keys never actually compared by bytes: the Studio handed the model the
document codec, whose registry guard throws for any non-document value,
so every key fell back to ToString. CultCacheMessagePack now builds the
model with a value serializer, and the Studio and tests both use it.
Tests pin the comparer rule, the unchanged-key rule (an old empty
reference key survives an edit of its value) and serialized identity
(keys whose ToString collides); each fails with its rule removed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Opening a directory store, read-only or writable, created the .records
folder and its .commit.lock before reading a manifest that might not
exist. Only a write creates them now: PullAll takes the commit lease with
create: false, which opens an existing lock and returns no lease when
there is none (no writer has ever committed there, so there is nothing to
exclude). CommitBatch and PushAll create the folder and lock as before.
Test DirectoryMessagePackBackingStore_Open_CreatesNothing_UntilAWrite:
opening a missing store and a manifest-less .records folder, read-only
and writable, leaves the filesystem unchanged; the first flush creates the
manifest, the lock and one page. With the previous lease code restored
the test fails on the first read-only open.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The reflective CultDocumentRegistry is now the only descriptor authority,
and payloads always go through MessagePackSerializer with
OptionsFor(document assembly). No consumer builds AOT, IL2CPP or trimmed,
Unity always ran the reflective path, and no benchmark showed a gain,
while the generator duplicated the registry's shape rules.
Removed: GameCult.Caching.MessagePack.Generator and .Analyzers (solution,
project and analyzer references, NuGet pack list), CultGeneratedDocumentMetadata
(provider attribute, interface, definitions, loader), the generated descriptor
and codec branches in CultCache and CultDocumentMessagePackSerialization,
DocumentShapeSweepTests and the generator half of the unkeyed test, the
Microsoft.CodeAnalysis.CSharp test reference and InternalsVisibleTo.
The registry rejects, with a message naming the fix, what MessagePackSerializer
cannot serialize and what the generator used to paper over: a document without
[MessagePackObject], a non-public document without AllowPrivate = true, and a
non-public member of an AllowPrivate document that is not [IgnoreMember].
DocumentShapeTests pins every rejection message and the accepted shapes'
round trips against MessagePackSerializer bytes (keeps the 93A16E0305 pin).
Rejected shapes are emitted into a dynamic assembly, which registry
discovery skips.
Nested member type names use reflection's Outer+Inner form, not the
generator's dot form, so ReferenceHolderEntry's fixture id is re-pinned
(bd8506... -> 4ec0c5...). A scan of the schema catalogs in all 2786 .cc
stores under F:\Projects (AetheriaEve and CultLib-* worktrees, bin and obj
excluded) found no member type name naming a nested type in either form.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The generator and CultDocumentRegistry now discover members by the same
steps: most-derived declaration first, override chain collapsed to one
member, attributes read up the override chain, rejections in one order
with one text (GCC001 / InvalidOperationException).
- Accessors are named per document, so documents sharing a [CultName] or
[CultIndex] base member no longer emit duplicate methods (CS0111).
- Duplicate [Key] slots are refused, including new-hidden fields and
properties; a hidden persisted member with a different slot is refused.
- String [Key("x")] gets its own message instead of "has no [Key]".
- A persisted property whose set accessor is private, protected or init,
or a readonly field, is refused instead of CS0272 in generated code.
- Overrides: MessagePack's DynamicObjectResolver reads the base
declaration's [Key]/[IgnoreMember] (it throws "key is duplicated" for a
re-keyed override and serializes an override marked [IgnoreMember]).
An override that changes them is refused in both paths; an override
that repeats or omits them is accepted and byte-pinned.
- Typed Serialize<T>/Deserialize<T> used static Options and skipped
generated codecs and declared resolvers; deleted. CultNetDocumentBinding
and CultMeshSnapshots route through the untyped path with their registry.
DocumentShapeSweepTests compiles each shape through the generator and
compares generated vs reflective schema id, catalog, name/index values
and payload against MessagePack's own bytes, or equal rejection text.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The static UntrustedData options are gone. An assembly declares its formatter
resolvers with [assembly: CultCacheFormatterResolver(typeof(R))];
CultDocumentMessagePackSerialization.OptionsFor(assembly) composes them ahead
of CultDocumentResolver and StandardResolver, once per assembly, and hands
assemblies that declare nothing the base Options. Generated codecs and the
reflective SerializeUntyped/DeserializeUntyped path both read
OptionsFor(typeof(X).Assembly). No registration call, no mutable options, no
load order.
CultMessagePackSecurity copies UntrustedData and supplies a collision-resistant
comparer over CultRecordRef<T>.Key.Value, so ref-keyed dictionaries
deserialize.
Tests: RefKeyedDictionaryRoundTrips failed on 825b4f7 ("No hash-resistant
equality comparer available"); DeclaredResolverEncodesValueTypeAsPositionalArray
and UnknownAssemblyGetsBaseOptions did not compile there (new API);
InteropNoteBytesUnchanged is a pin whose payload constant was captured on
825b4f7. BenchItem, the one unkeyed [CultDocument] in any consumer (in-memory
benchmark, no store), gains [Key]s in its existing slot order.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A direct MessagePackAnalyzer reference with PrivateAssets="all" in
Caching.MessagePack and Networking keeps MessagePack 3's generator out of
consumer builds, where it emitted a reference to a nonexistent
GeneratedMessagePackResolver formatter for Dictionary<CultRecordRef<T>, float>
(CS0426). CultLib composes StandardResolver itself and never registers a
generated resolver.
tests/GameCult.Caching.ConsumerBuild is RefKeyedDictionaryCompilesInAConsumer:
its build is the test. On 825b4f7 it failed with CS0426; now it builds and
obj has no MessagePack.SourceGenerator output.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The cache-wide unpublished list let any thread drain and deliver another
thread's changes: a throwing observer dropped the rest of the queue and threw
into an unrelated writer, writes returned before their change was published,
and a third-party load could wait indefinitely.
Every admission now runs in a hold (Held). The outermost hold on a thread
collects exactly the changes it admits and publishes them after it leaves the
gate, before it returns, on its caller's thread; nested holds append to it.
Every change is delivered even when an observer throws; the first exception
(AggregateException for several) is then rethrown, and a failing body's own
exception wins over observer exceptions.
Stores take the hold instead of locking a shared Gate object, so a direct
PullAll on an attached store publishes its own load and the empty follow-up
Loaded hand-over is gone. Gate is replaced by protected Held.
Tests: OwnChangeIsPublishedBeforeWriteReturns and
ThrowingObserverDoesNotLoseOtherChanges fail on 4eacd87 and pass here. The
repull tests now record a write that reaches the gate while the pull is paused
instead of inferring readiness from thread state; both fail when the pull does
not hold the gate.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Observers and OnUpdate never run under the cache gate. Admit queues changes
while the gate is held; the outermost entry point (attach, pull, commit,
upsert, remove) publishes them after releasing it. A store's direct PullAll
hands over an empty load outside its lock so the cache publishes then.
A throwing observer can no longer split store from cache: publication starts
only after the store has adopted its loaded view. The directory store adopts
its durable catalog only after the cache admits the load; the write-only
_durableIndex is deleted.
Single-file Push, Delete, PushAll and CommitBatch take the gate like the
directory store; Dispose flushes under it. Write mints storedAt and admits
under one gate hold, and Land re-stamps staged upserts under the gate that
admits them, so a concurrent pull cannot land a later storedAt in between.
Tests: ObserverDoesNotRunUnderTheGate,
ThrowingOnUpdateHandlerDoesNotResurrectDeletedRecord,
RepullNeverErasesAStagedWriteOnDirectoryStore,
DirectCallsOnAnAttachedStoreShareTheCacheGate; the directory last-writer
result is pinned to {a,c,x}; the single-file re-pull test waits for the write
to reach the store or block instead of sleeping. Pausing tests always release
the pause so a failure cannot deadlock Dispose.
Docs: observer rule, direct store calls share the gate, directory TryCommit
contention and gate wait, per-key last-writer-wins, directory flush merge.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
An attached store takes its cache's gate as its own lock (CacheBackingStore.Gate),
so a pull's load callback into the cache can no longer take the gate after the
store lock. The directory store's private mutation gate is deleted; the
single-file pull now runs under the gate; PullAllBackingStoresAsync takes the
gate around each store's pull. Fixes the pull/upsert deadlock on directory
stores and the re-pull that erased a staged single-file write.
An unconditional single-file Commit writes the store's whole view plus the batch,
exactly as a flush would, instead of merging onto the file; it persists earlier
staged writes and leaves the store clean. Conditional commits still land onto
the file under the lock. Directory-store commit and flush already share one
generation writer (staged keys onto the current manifest) and are unchanged.
Contracts state both rules, the lock order, that OnUpdate fires for loads only,
that a missing single file keeps staged writes, and that a store failing to
hydrate on open throws and is left byte-identical.
Tests: PullAndUpsertOnDirectoryStoreDoNotDeadlock, RepullNeverErasesAStagedWrite,
UnconditionalCommitIsLastWriterWinsLikeFlush, UnconditionalCommitPersistsStagedWrites,
OpeningACorruptStoreThrowsAndLeavesItUntouched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Cut 3 of the CultCache migration (Aetheria docs/cultcache-migration-cut.md, 2.1
and 2.2). One commit, because routing, the batch that replaces the ambient
transaction, and conditional commit share one admission path.
Deleted first:
- globals: InitializeGlobals, MaterializeMissingGlobals, the two-argument
constructor, the initializeGlobals plumbing in CultCacheMessagePack,
ContainsDurableRecord (cache and directory store)
- replication: the push-to-every-store loops and the attach-time push of
existing records
- typeof(T) as the schema owner in AddAsync/UpsertAsync and in
CultNetDatabase.PutAsync/PutPredictedAsync
- the ambient transaction: ExecuteTransactionAsync x2, CultCacheTransaction,
CommitTransaction, VisibleStoredDocuments, the AsyncLocal, the SemaphoreSlim
gate, and every ambient check
- CultCacheOpenOptions.PullOnOpen; OpenAsync is Task.FromResult(Create(...))
- DirectoryMessagePackBackingStore.ResolveLegacyUncataloguedRecordCatalog: the
catalog precommit (fe55bff) is an ancestor of the v4 page format (eea8248),
so no v4 manifest can reach it
- CultCache.DeleteAsync<T>; its one caller, CultNetDatabaseSubscriptionClient,
called it by reflection and now calls Remove(key)
Added:
- AddBackingStore(store, params Type[] homes): attach validates (clean, not
already attached, one untyped store, no claimed home type, no admitted
record changes home) and then pulls under the gate
- Home(type): most derived routed home, else the untyped store
- Admit: every add and remove, from a write, a batch or a load, validates
(home, read-only, loaded-from-home, global singleton), lands durably, then
applies and publishes; a pull hands its whole delta over in one call, so a
refused record leaves the store and the cache unchanged
- read-only stores (CacheBackingStore(readOnly), both store constructors,
CultCacheOpenOptions.ReadOnly); flush skips them and writes only dirty stores
- assignable GetGlobal<T>/GetByName<T>/GetByIndex<T>/Watch<T>; more than one
candidate throws naming them
- Commit(Action<CultCacheBatch>) returning bool, TryCommit returning
CultCommitOutcome, CultCacheBatch (Upsert, Remove, Expect, ExpectUnchanged,
sealed after commit), CultCommitRequest; CacheBackingStore.CommitBatch
evaluates conditions under the store's exclusive lock
- single-file store: <path>.lock sidecar for CommitBatch and PushAll; a clean
store commits onto the file as it is, so records it never loaded survive
- storedAt minted in one place and bumped a tick when not later than the record
it replaces
Consumer adjustments in this repo: Mesh streaming tests drop PullOnOpen; the
Studio bridge stops reflecting on it; BackingStoreTests' global test creates
its global explicitly and its transaction tests are rewritten against Commit
(the two that proved ambient hydration rules are deleted); the pre-Cut 2
refusal test expects the attach to throw.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
