WMSDK-0000: fix(coredata): use enqueued timestamp for CDEvent.timestamp
#599
Conversation
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
… refactor loader; add contract tests - Add `DataBaseLoading` protocol (`loadPersistentContainer` / `makeInMemoryContainer` / `destroy`). - Make `DataBaseLoader` conform to the protocol. - Startup flow: on-disk load → destroy+retry → in-memory fallback; no fatalError on loader init. - DI: register `DataBaseLoading`; use `StubLoader` on init failure (assertionFailure in DEBUG). No Core Data model changes; no public API changes beyond introducing the protocol.
…ack; remove fatalError in injection - Add DatabaseRepository protocol: - Lifecycle/limits: `limit`, `lifeLimitDate`, `deprecatedLimit`, `onObjectsDidChange`. - Metadata: `infoUpdateVersion`, `installVersion`, `instanceId`. - CRUD: `create(event:)`, `readEvent(by:)`, `update(event:)`, `delete(event:)`. - Queries/maintenance: `query(fetchLimit:retryDeadline:)`, `removeDeprecatedEventsIfNeeded()`, `countDeprecatedEvents()`, `erase()`, `countEvents()`. - Convenience default: `query(fetchLimit:)` uses `Constants.Database.retryDeadline`. - Make the existing Core Data repository conform to DatabaseRepository. - DI/Composition: depend on `DatabaseRepository` instead of a concrete class. - On construction failure, inject a safe Noop `StubDatabaseRepository` (logs via MindboxLogger; `assertionFailure` in DEBUG). - Remove `fatalError` from repository injection paths to avoid startup crashes and allow graceful degradation. Notes - No Core Data model changes. - Public API impact limited to adding the DatabaseRepository protocol and conformances. - DatabaseLoader changes and tests will follow in separate commits.
…a salvage, destroy+retry, in-memory fallback - Add low-disk gate: compute free space via filesystem attributes and bail out to in-memory store if `freeSize < diskSpaceRepairThreshold` (300 MB) without touching on-disk store. - Implement read-only metadata salvage (`NSReadOnlyPersistentStoreOption: true`) to preserve `install`, `infoUpdate`, `instanceId` before any destructive action; re-apply metadata to the new/retried store. - Define clear startup flow: 1. try loadPersistentStores() → return on success; 2. on failure → salvage metadata; 3. if low disk → in-memory fallback (no destroy); 4. else destroy on-disk store and retry loading; 5. if retry fails → in-memory fallback. - Centralize Core Data store options in `applyStandardOptions`: synchronous add, `NSFileProtectionNone`, automatic/inferred migration. - `destroy()` uses `destroyPersistentStore` (iOS 15+ / legacy path for older OSes) and logs before/after states. - Keep `StubLoader` for DI fallback; no `fatalError` paths in loader code.
…; add `NoopDatabaseRepository` & retry semantics coverage
- Migrate tests from concrete `MBDatabaseRepository` to the `DatabaseRepository` protocol
- Update test fixtures/factories to accept `DatabaseRepository` instead of `MBDatabaseRepository`
- Add a dedicated test suite for `NoopDatabaseRepository`
- Verify CRUD methods are no-ops and never throw
- Assert `countEvents()` and `countDeprecatedEvents()` return `0`
- Assert metadata getters return `nil` and setters are safe (no crash, no side effects)
- Ensure `erase()` is a no-op and safe to call repeatedly
- Update `Event` contract in tests to cover retry support
- Extend fixtures with `retryTimestamp: TimeInterval`
- Add computed `isRetry` and pin its behavior:
- `isRetry == false` when `retryTimestamp == 0` (new/default events)
- `isRetry == true` when `retryTimestamp != 0` (after update)
- Use a fixed clock/deterministic timestamps to avoid flakiness
- Notes
- Test-only changes; no production code or Core Data model changes
- Prepares the suite for the new loader/fallback flow by decoupling tests from the concrete repository
… `UserDefaults.standard` when `UserDefaults(suiteName:)` returns `nil` - Avoids a hard crash in production while still surfacing the problem in debug. - Keeps the app functional if the App Group identifier is missing/misconfigured. - Adds a guard around UserDefaults(suiteName:) and returns `.standard` on failure.
- Add `DatabaseLoaderFlowTests` with Spy subclass to cover all branches: - straight success (no salvage/destroy), - load failure + low disk → in-memory + metadata applied, - load failure + enough disk → destroy+retry + metadata applied, - repair failure → in-memory fallback.
…bserver lifecycle and refactor `DatabaseRepository` injection
- Add memory-warning handling to `MBDatabaseRepository` that’s enabled **only** for the
`NSInMemoryStoreType` store.
- Subscribe to `UIApplication.didReceiveMemoryWarningNotification` (`queue: .main`) in `init`
and remove the observer in `deinit`.
- On warning:
- Create a background context and aggressively prune all `CDEvent` rows
(`includesPropertyValues = false`), then `save()` and `reset()` the context.
- Fire `onObjectsDidChange` on the **main** queue after pruning completes.
- Guard against re-entrancy with `isPruningOnWarning`; reset the flag on the main queue
once the prune finishes.
- No-op for SQLite-backed stores (warning is ignored there).
- Keep the main thread lightweight: the heavy work is performed on a background context;
main is used only to enqueue the job and deliver the callback.
Tests:
- Add `MBDatabaseRepositoryMemoryWarningTests` covering:
- `test_InMemory_PrunesAll_OnMemoryWarning` — in-memory store is fully pruned on warning.
- `test_SQLite_IgnoresMemoryWarning` — SQLite store ignores warning.
- `test_InMemory_PruneIsIdempotent_WhenWarningsBurst` — multiple warnings don’t double-prune.
Notes:
- No public API changes. Behavior is active only when the SDK runs with the in-memory fallback,
reducing OOM risk by shedding data under pressure.
justSmK
force-pushed
the
fix/WMSDK-0000-fix-Event
branch
from
6676ea2 to
a3ff24e
Compare
Contributor
Author
|
Здесь не увидели каких-то проблем потенциальных или рисков. Можно поправить этот момент. |
Vailence
approved these changes
Sep 30, 2025
enotniy
approved these changes
Sep 30, 2025
justSmK
force-pushed
the
fix/WMSDK-0000-fix-Event
branch
from
a3ff24e to
7b2d2f9
Compare
Base automatically changed from
feature/WMSDK-518-fatalError-MBDatabaseRepository
to
develop
September 30, 2025 11:01
Previously we saved the current time (`Date().timeIntervalSince1970`). Now we persist the timestamp from `Event.enqueueTimeStamp` so the record reflects the actual enqueue time. Event model was not changed.
justSmK
force-pushed
the
fix/WMSDK-0000-fix-Event
branch
from
7b2d2f9 to
2c80d1d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Previously we saved the current time (
Date().timeIntervalSince1970). Now we persist the timestamp fromEvent.enqueueTimeStampso the record reflects the actual enqueue time. Event model was not changed.I'm not sure about this change / fix.