Polecat: consume JasperFx rc.2 / Weasel alpha.8 dedupe + lift wave (#125–#138), cut 4.0.0-rc.1#139
Merged
Merged
Conversation
…ncyStyle (#127), DcbConcurrencyException+ProgressionProgressOutOfOrderException (#128), IdentityAttribute+FindIdMember (#135) Foundation re-pin from the alpha line to the JasperFx 2.0 rc: JasperFx 2.0.0-alpha.20 -> 2.0.0-rc.2 JasperFx.Events 2.0.0-alpha.20 -> 2.0.0-rc.2 JasperFx.RuntimeCompiler 5.0.0-alpha.8 -> 5.0.0-rc.2 JasperFx.SourceGeneration 2.0.0-alpha.9 -> 2.0.0-rc.2 JasperFx.Events.SourceGenerator 2.0.0-alpha.12 -> 2.0.0-rc.2 Weasel.SqlServer 9.0.0-alpha.7 -> 9.0.0-alpha.8 Weasel.EntityFrameworkCore 9.0.0-alpha.7 -> 9.0.0-alpha.8 The rc lifted three families of types Polecat declared locally, so the bump does not compile until all three collisions are resolved — bundled here as one green checkpoint (the rest of the consume/dedupe issues land one per commit on this branch). #127 (jasperfx#327) — TenancyStyle + DeleteStyle: - Deleted Polecat.Metadata.DeleteStyle and the inline StoreOptions TenancyStyle enum (both ordinal-identical to the lifted enums: Remove=0/SoftDelete=1, Single=0/Conjoined=1). - Added src/Polecat/GlobalUsings.cs aliasing TenancyStyle -> JasperFx.MultiTenancy.TenancyStyle and DeleteStyle -> JasperFx.DeleteStyle, mirroring Marten's GlobalUsings.cs. Fixed two fully-qualified Metadata.DeleteStyle refs in QuerySession. #128 (jasperfx#328) — DcbConcurrencyException + ProgressionProgressOutOfOrderException: - Deleted Polecat.Events.Dcb.DcbConcurrencyException (byte-identical to the lifted JasperFx.Events type) and Polecat.Exceptions.ProgressionProgressOutOfOrderException (the lifted JasperFx.Events.Daemon type folds in Polecat's richer 3-arg ctor + ProjectionName/ExpectedFloor/AttemptedCeiling props verbatim). - Global aliases for both; dropped the now-dead using in UpdateProjectionProgress. The inline catch(2627) insert-conflict mapping is untouched. #135 (jasperfx#335) — IdentityAttribute + ID resolution: - Deleted Polecat.Attributes.IdentityAttribute; aliased IdentityAttribute -> JasperFx.IdentityAttribute (so [Identity] keeps resolving). Rewrote the test that used the fully-qualified [Polecat.Attributes.Identity] to [JasperFx.Identity]. - DocumentMapping.FindIdProperty now delegates to the side-effect-free JasperFx.DocumentIdentity.FindIdMember(type, predicate), passing Polecat's own valid-id-type predicate (SupportedIdTypes + TryResolveValueTypeId, both nullable-aware) so strong-typed-id "Id" members are still recognized — the shared helper's default ValidIdTypes set is scalar-only. Result filtered to PropertyInfo to preserve Polecat's property-only contract (both call sites need PropertyInfo for .PropertyType / .SetValue). Test-assembly GlobalUsings.cs added to Polecat.Tests, Polecat.EntityFrameworkCore, and Polecat.EntityFrameworkCore.Tests for the lifted names those assemblies reference unqualified (global usings don't cross assembly boundaries). Validated: Polecat.Tests (net10.0): 1133 pass / 3 pre-existing skips / 0 fail Polecat.EntityFrameworkCore.Tests (net10.0): 37 pass / 0 fail Polecat.AotSmoke (net10.0): builds + runs clean All projects build clean under -c Debug/Release Closes #127, #128, #135. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…eConstants
Two "adopt the already-shared thing" cleanups from the Critter Stack 2026
dedupe pillar (jasperfx#214), both Polecat-side only (no upstream work).
1. IProjectionCoordinator — stop re-declaring it
- Polecat.Events.Daemon.Coordination.IProjectionCoordinator now inherits
the member-for-member-identical JasperFx.Events.Daemon.IProjectionCoordinator
(lifted from Marten); the five duplicated members are deleted. Mirrors
Marten's empty inheriting interface.
- The generic marker IProjectionCoordinator<T> inherits the lifted
IProjectionCoordinator<T> and relaxes its constraint from
`where T : IDocumentStore` to the shared `where T : class`. The concrete
ProjectionCoordinator<T> gains `class` (now `where T : class, IDocumentStore`)
so it still satisfies the relaxed interface; the DI registration in
PolecatConfigurationExpression already used `class, IDocumentStore`.
2. JasperFx.StorageConstants
- Deleted Polecat.Tenancy (its only member was DefaultTenantId = "*DEFAULT*",
byte-identical to JasperFx.StorageConstants.DefaultTenantId). All consumers
(~19 refs across sessions, bulk insert, event-store explorer, projection
replay, table DDL, ITenancy impls, + 4 test files) now reference
JasperFx.StorageConstants.DefaultTenantId. Removes the silent-drift hazard
if the sentinel ever changes upstream.
- The "tenant_id" string literal in the 6 schema/table builders (DocumentTable,
DocumentIndex, EventsTable, StreamsTable, NaturalKeyTable, EventTagTable)
now uses JasperFx.StorageConstants.TenantIdColumn. Operations/LINQ column
references keep the literal (out of the issue's "table classes" scope; same
resolved value, lower drift risk).
Validated:
All projects build clean (Debug)
Polecat.Tests Daemon+Schema+MultiTenancy+Documents (net10.0): 152 pass / 3 skip / 0 fail
(schema tests confirm the tenant_id constant emits identical DDL)
Closes #125.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…class (consume jasperfx#326)
jasperfx#326 (PR jasperfx#336) lifted the projection-coordinator execute loop
into JasperFx.Events.Daemon.ProjectionCoordinatorBase. Polecat's ExecuteAsync
was a slot-for-slot mirror of Marten's, so this is the Polecat-side consumption.
Polecat.Events.Daemon.Coordination.ProjectionCoordinator now subclasses
ProjectionCoordinatorBase (and still declares Polecat's empty IProjectionCoordinator
sub-interface, mirroring Marten):
- Base ctor supplied with: BuildDistributor(store, loggerFactory) [now static so
it can run in the base-ctor argument list], the StoreOptions.ResiliencePipeline,
TimeProvider.System, and the three settings converted to TimeSpan
(LeadershipPollingTime is int-ms -> TimeSpan.FromMilliseconds; AgentPauseTime
and HealthCheckPollingTime are already TimeSpans on DaemonSettings).
- Abstract seams implemented: ResolveDaemon(set) -> DaemonFor((PolecatDatabase)set.Database)
keeping the per-database dictionary cache in the subclass; ResolvedDaemons()
snapshots that cache; the three tenancy-aware daemon accessors
(DaemonForMainDatabase / DaemonForDatabase / AllDaemonsAsync) are overrides.
- Deleted the lifted-into-base members: ExecuteAsync, the two agent helpers
(StartAgentsIfNecessaryAsync / StopAgentsIfNecessaryAsync), StopAllAgentsAsync,
and the bespoke StartAsync/StopAsync/PauseAsync/ResumeAsync lifecycle with its
_paused / _runCts / _runTask tracking. The lazy Distributor property is gone —
the base exposes Distributor (set once in the ctor).
Behavioral change (the #326 fix): the shared loop wraps StartAgentAsync in the
ResiliencePipeline and, on failure, ejects the paused shard and releases the
set's lock so another node can pick it up. Polecat previously started agents raw
with no lock-release-on-failure — this closes that gap. Pause/Resume/Stop
bookkeeping is normalized on the base's single cancellation-source + awaited
runner model.
Validated:
All projects build clean (Debug)
Polecat.Tests.Daemon (net10.0): 53 pass / 3 pre-existing skips / 0 fail
(covers async_daemon_tests + multi_tenant_daemon_tests hot-cold handover on SQL Server)
Closes #126.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…(consume jasperfx#329) jasperfx#329 (PR jasperfx#340) lifted SkippedEventsCountObserver, ResilientEventLoader, and EventLoaderException into JasperFx.Events.Daemon. - Deleted Polecat.Events.Daemon.SkippedEventsCountAugmenter; PolecatDatabase now subscribes the lifted SkippedEventsCountObserver. The lifted observer adds a `ShardName == HighWaterMark` guard (strict superset) — harmless here since Polecat only subscribes it for the HWM tracker. - Wrapped PolecatEventLoader with the lifted ResilientEventLoader(Options.ResiliencePipeline, inner, database) at the IEventStore.BuildEventLoader site (database is already the IEventDatabase the decorator wants). Dropped the inline _resilience.ExecuteAsync(...) Polly call from PolecatEventLoader.LoadAsync — the loader is now the bare inner loader; resilience is layered by the decorator. Removed the now-unused _resilience field + `using Polly;`. - Bonus parity: Polecat now reports loading metrics via EventRequest.Metrics.TrackLoading() (provided by the decorator) — it did no loading metrics before. Validated: All projects build clean (Debug) Polecat.Tests.Daemon (net10.0): 53 pass / 3 pre-existing skips / 0 fail (covers event_loader_tests, event_loader_resiliency_tests, and the skipped-events-count tracker behavior) Closes #129. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…perfx#330)
jasperfx#330 (PR jasperfx#342) lifted ISoftDeleted, IVersioned, ITracked to
JasperFx.Metadata.
- Deleted Polecat.Metadata.{ISoftDeleted,IVersioned,ITracked}; aliased all
three to the JasperFx.Metadata types in src/Polecat + test-assembly
GlobalUsings.cs (the names are referenced unqualified throughout). ISoftDeleted
and IVersioned (Guid Version) are byte-identical straight rebinds. Fixed one
short-qualified `Metadata.ITracked` pattern-match in DocumentSessionBase.SyncMetadata.
- ITracked was lifted with non-nullable string members (Marten's shape) vs
Polecat's string?. Per the issue, adjusted the concrete classes rather than the
interface: the TrackedDoc / FullMetadataDoc test documents now declare
non-nullable `string ... = string.Empty` for CorrelationId/CausationId/
LastModifiedBy (SyncMetadata always copies non-null session values onto them;
no test asserts a null default). This clears the CS8766 return-nullability
warnings the rebind surfaced.
IRevisioned is intentionally NOT included (tracked in jasperfx#341 pending the
int/long revision unification — see #136/#138).
Validated:
All projects build clean (Debug); no new nullability warnings (the 3 remaining
CS8767 on TenantId/IHasTenantId are pre-existing, unrelated to #130)
Polecat.Tests Metadata+SoftDeletes+Concurrency+Versioning (net10.0): 59 pass / 0 fail
Closes #130.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…onsume jasperfx#331)
jasperfx#331 (PR jasperfx#343) lifted IPatchExpression<T> as Marten's superset
+ RemoveAction into JasperFx.Events. Polecat's interface was a subset, so
adoption grows the implementation.
- Deleted Polecat.Patching.IPatchExpression<T> + RemoveAction. RemoveAction
(ordinals identical: RemoveFirst=0/RemoveAll=1) is aliased in src + test
GlobalUsings.cs. IPatchExpression<T> is an open generic (can't be a using
alias), so the two files that name it unqualified — PatchExpression.cs and
PatchingExtensions.cs — gain `using JasperFx.Events;`.
- Grew PatchExpression<T> to the full lifted interface:
* Rename(string, Expression<Func<T, object>>) — replaces Polecat's generic
Rename<TElement>. ToPath already unwraps the Func<T,object> boxing Convert;
a small UnwrapMemberType helper recovers the real member type to keep the
scalar/complex JSON_MODIFY distinction. (No test exercised the old generic
Rename, so it's dropped rather than kept as a convenience method.)
* AppendIfNotExists / InsertIfNotExists / Remove predicate overloads —
these require translating an Expression<Func<TElement,bool>> into a JSON
array search, which Polecat's JSON_MODIFY patch translation doesn't do yet.
Per the issue, they throw NotSupportedException with a clear message
pointing at the supported non-predicate overload, so the interface members
exist without silently misbehaving.
Validated:
All projects build clean (Debug)
Polecat.Tests.Patching (net10.0): 43 pass / 0 fail
(the 42 existing patch ops via Set/Append/Insert/Remove/Delete/Increment/
Duplicate + non-predicate IfNotExists are unchanged; new predicate overloads
are NotSupported and not exercised)
Closes #131.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…Level (consume jasperfx#332)
jasperfx#332 (PR jasperfx#344) lifted the common OpenTelemetryOptions base +
TrackLevel to JasperFx.OpenTelemetry.
- Polecat.Internal.OpenTelemetry.OpenTelemetryOptions had nothing beyond the
base (TrackConnections + a "Polecat"-named Meter), so it's now a thin subclass
of JasperFx.OpenTelemetry.OpenTelemetryOptions with `: base("Polecat")`. Kept
as a subclass (option 2 in the issue) rather than consuming the base directly
so the public name + parameterless construction survive — StoreOptions does
`OpenTelemetry { get; } = new()` and a test does `new OpenTelemetryOptions()`.
- Deleted Polecat's TrackLevel enum (ordinals identical: None=0/Normal=1/
Verbose=2); aliased TrackLevel -> JasperFx.OpenTelemetry.TrackLevel in src +
test GlobalUsings.cs. TrackConnections + Meter now come from the base.
- PolecatTracing ActivitySource holder intentionally left as-is (per the issue
and the upstream pass-N audit — per-library OTel sources are wanted).
Validated:
All projects build clean (Debug)
Polecat.Tests OpenTelemetry + document_store_usage (net10.0): 13 pass / 0 fail
(covers Normal/Verbose/None track-level tracing + the flattened
OpenTelemetryTrackConnections store-usage description)
Closes #132.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
… jasperfx#333)
jasperfx#333 (PR jasperfx#345) lifted IDocumentSchemaResolver to JasperFx.Events
preemptively — Polecat had no equivalent. This implements it as the cross-store
"where does this document live" surface.
- New Internal.PolecatDocumentSchemaResolver : IDocumentSchemaResolver maps
Polecat's SQL Server conventions: documents -> pc_doc_{typename}, event store ->
pc_events / pc_streams / pc_event_progression, all under the single
DatabaseSchemaName (events share the document schema). Qualified names are
bracket-quoted [schema].[table]; bare names are the unbracketed identifier
(matching DocumentMapping.TableName). For<T> / For(Type) cover documents and
projection aggregates via the same pc_doc_ convention.
- Exposed as StoreOptions.SchemaResolver (named SchemaResolver, not Schema,
because StoreOptions.Schema is already Polecat's SchemaConfiguration). Lazily
constructed + cached.
- Wired into DescribeConfiguration: the store-usage description now carries a
"Schema" child reporting the resolved event-store table locations — the
schema-diagnostics surface the lift is about.
Validated:
All projects build clean (Debug)
Polecat.Tests Schema resolver + document_store_usage (net10.0): 11 pass / 0 fail
(4 new resolver tests: schema names, qualified/bare event tables, document
table by type, custom schema)
Closes #133.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…cs (consume jasperfx#334) jasperfx#334 (PR jasperfx#346) lifted IInitialData<TStore>, InitialDataCollection<TStore>, IConfigureStore<TOptions>, and IAsyncConfigureStore<TOptions> to JasperFx core. - Polecat.IInitialData now inherits JasperFx.IInitialData<IDocumentStore> (empty marker preserving the Polecat.IInitialData name); implementers' Populate(IDocumentStore, CancellationToken) satisfies the lifted member. - Polecat.InitialDataCollection now inherits JasperFx.InitialDataCollection<IDocumentStore> — the lambda Add overload (originally Polecat's contribution) lives in the lifted base, so the empty subclass keeps StoreOptions.InitialData and PolecatActivator's `foreach (var d in InitialData) await d.Populate(store, ct)` working unchanged. - Polecat.IConfigurePolecat now inherits JasperFx.IConfigureStore<StoreOptions>; the AddPolecat factory's GetServices<IConfigurePolecat>() + the LambdaConfigurePolecat impl bind unchanged. Old names type-forwarded via inheritance (no rename, no source break). The generic IConfigurePolecat<T> is Polecat-specific and out of scope (the lifted contract is options-typed, not store-typed). The optional async configure hook (IAsyncConfigureStore<StoreOptions>) is available but not wired — no current need. Validated: All projects build clean (Debug) Polecat.Tests Seeding + DependencyInjection (net10.0): 22 pass / 0 fail Closes #134. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…ion enums (consume weasel#287)
weasel#287 lifted the Hi-Lo sequence contract and the serialization storage
enums into Weasel.Core; both were member-identical between Marten/Polecat and
Weasel.Core (verified ordinal/flag values against the pinned 9.0.0-alpha.8 dll).
Part A — Hi-Lo sequence:
- HiloSequence now derives from Weasel.Core.Sequences.HiloSequenceBase, which
owns the EntityName / CurrentHi / CurrentLo / MaxLo / Settings state, the
NextInt/NextLong lock, AdvanceValue, ShouldAdvanceHi, and TrySetCurrentHi.
The subclass keeps only the SQL Server I/O: AdvanceToNextHi(Sync),
SetFloor, EnsureHiloTable*, the optimistic UPDATE/INSERT TryGetNextHi* with
the 2627 retry, and the ResiliencePipeline wrapping. (TrySetCurrentHi now
takes the boxed long the loop already passes.)
- Deleted Polecat.Schema.Identity.Sequences.{ISequence, IReadOnlyHiloSettings,
HiloSettings} and Polecat.Exceptions.HiloSequenceAdvanceToNextHiAttemptsExceededException;
global-aliased the three sequence types in src/Polecat/GlobalUsings.cs so
SequenceFactory / StoreOptions / DocumentMapping / AdvancedOperations keep
resolving them unqualified. Behavior unchanged (the lifted exception message
differs only by a trailing period; no test asserts it).
Part B — serialization enums:
- Deleted Polecat.Serialization.{Casing, CollectionStorage, NonPublicMembersStorage,
EnumStorage}; consume the Weasel.Core enums. No global EnumStorage alias (it
would shadow files that already see Weasel.Core.EnumStorage); instead added
`using Weasel.Core;` to the six consumers (ISerializer, Serializer, StoreOptions,
EventDataMemberFactory, EnumMember, MemberFactory) + the serialization test.
Validated:
All gate projects build clean (Debug): Polecat, Polecat.Tests, Polecat.AspNetCore,
Polecat.EntityFrameworkCore(.Tests), Polecat.AotSmoke
Polecat.Tests HiLo + serialization_configuration (net10.0): 68 pass / 0 fail
Closes #137.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
…column (consume jasperfx#336, jasperfx#338)
jasperfx#338 lifted the ILongVersioned (long Version) contract; jasperfx#336 is the
fuller feature. Implemented once, closing both: documents can now track a 64-bit
revision, recommended for MultiStreamProjection-derived views whose Version is the
global event sequence number (which can climb past Int32.MaxValue).
- Detection: DocumentMapping recognizes JasperFx.ILongVersioned alongside IRevisioned —
both set UseNumericRevisions; ILongVersioned also sets the new UseLongRevisions flag.
- Column (Decision D2 = bigint always): DocumentTable `version` is now `bigint` with no
default (every write sets it explicitly; the old DEFAULT was dead weight and blocked
the in-place ALTER). IRevisioned (int) values fit and are downcast on read.
- Non-destructive widening migration: DocumentTableEnsurer.WidenVersionColumnIfNeededAsync
drops any lingering version default constraint and ALTERs an existing int column to
bigint in place before Weasel diffs the table — never a drop/recreate, so rows survive.
SQL Server rejects ALTER COLUMN while a default references the column and Weasel emits
only a bare ALTER COLUMN, so the default handling lives here. Idempotent.
- Reader paths widened to GetInt64 then assign-long (ILongVersioned) / downcast-to-int
(IRevisioned): Insert/Update/UpsertOperation.PostprocessAsync, QuerySession.SyncVersionProperties,
DeserializingSelector, AdvancedSqlResultReader. Expected-revision plumbing (DocumentProvider
BuildUpsert/BuildUpdate + the two operations) is now long; int callers widen implicitly.
- API: added UpdateRevision<T>(T, long) overload (IDocumentOperations + DocumentSessionBase
+ NestedTenantSession delegate). JasperFx.ConcurrencyException already covers both.
- Docs: documents/concurrency.md gains an ILongVersioned section + the int-vs-long guidance;
storage docs updated to bigint.
Validated:
All gate projects build clean (Debug); no new warnings
Polecat.Tests full suite (net10.0): 1144 pass / 3 pre-existing skip / 0 fail
(Versioning slice 27 pass — includes long_versioned_operations with a Version >
Int32.MaxValue round-trip + concurrency, and version_column_widening_migration
proving int->bigint preserves rows)
Polecat.EntityFrameworkCore.Tests (net10.0): 37 pass / 0 fail (incl. forced widening path)
Closes #136 #138.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
Graduates Polecat to its first release candidate on the foundation already pinned in this branch (JasperFx 2.0.0-rc.2 / Weasel 9.0.0-alpha.8) after the #125–#138 consume + dedupe wave. (main had independently moved to 4.0.0-alpha.10; rc.1 supersedes it.) Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
This was referenced May 20, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Consumes the Critter Stack 2026 dedupe/lift wave: types Polecat used to declare locally were lifted into JasperFx 2.0.0-rc.2 / JasperFx.Events and Weasel.Core/SqlServer 9.0.0-alpha.8, which this branch re-pins to. Each issue is one commit; type-forwards use global-using aliases (one
GlobalUsings.csper assembly) so unqualified references keep resolving to the canonical lifted types.Per-issue commits
72f05ac9acda3eIProjectionCoordinator+JasperFx.StorageConstants3e08f0aProjectionCoordinatorto aProjectionCoordinatorBasesubclassf7c4b5cResilientEventLoader+SkippedEventsCountObserver9df48e9ISoftDeleted/IVersioned/ITracked) toJasperFx.Metadata5e73547IPatchExpression<T>superset +RemoveAction838af8bOpenTelemetryOptionsbase + type-forwardTrackLevela00c774IDocumentSchemaResolverfrom JasperFx.Events0866839IInitialData<TStore>+IConfigureStore<TOptions>genericsd86db84HiloSequenceBase+ serialization enums325d3eaILongVersioned(long) document revisions +bigintversion column747e00aFoundation bump
Pins move to
JasperFx* 2.0.0-rc.2andWeasel* 9.0.0-alpha.8. Polecat itself graduates to its first release candidate, 4.0.0-rc.1 (mainhad independently moved to4.0.0-alpha.10; rc.1 supersedes it).Decision D2 makes the document
versioncolumnbigintalways (carryingIRevisionedint — downcast on read — andILongVersionedlong). Existing tables with anintversion column are widened in place byDocumentTableEnsurer.WidenVersionColumnIfNeededAsync:versiondefault constraint, then runsALTER COLUMN ... bigint— never a drop/recreate, so rows are preserved. (SQL Server rejectsALTER COLUMNwhile a default references the column, and Weasel emits only a bareALTER COLUMN, so the default handling lives here.) It is idempotent.versioncolumn now has no default — every write sets it explicitly, the default was dead weight, and removing it eliminates default-constraint migration churn.version_column_widening_migration(proves int→bigint preserves data) andlong_versioned_operations(aVersion > Int32.MaxValueround-trip + concurrency).Validation
Polecat.Testsfull suite (net10.0): 1144 pass / 3 pre-existing skip / 0 fail.Polecat.EntityFrameworkCore.Tests(net10.0): 37 pass / 0 fail (including a forced widening path).Closes #125 #126 #127 #128 #129 #130 #131 #132 #133 #134 #135 #136 #137 #138
🤖 Generated with Claude Code