tecfy_database 1.2.3

A fast, realtime, JSON-based, index-driven local database for Flutter, built on SQLite. Schemaless documents with native indexed queries and reactive streams.

1.2.3

Fixed

• Stream listener leak: stream(), count() and doc().stream() now register their listener only while the stream has subscribers and unregister it when the last one cancels. Previously every call added a listener that was never removed, so streams created inside a widget's build() piled up and each write re-ran all of their queries.
• Document lock could stay held forever: the global lock around doc().get()/update()/delete() is now a FIFO mutex that is always released, even when the query throws. Waiters no longer poll every 50 ms.

Added

• stream(limit:): cap a collection stream to the first N rows, e.g. stream(orderBy: 'time DESC', limit: 100) for a paged live list.

Changed

• Re-queries triggered while a listener's query is already running are coalesced into one follow-up run, so a burst of writes costs at most two queries per stream and results can't arrive out of order.
• Query errors inside a stream are now delivered to the stream as errors instead of escaping as unhandled async errors.
• get()/search() results of 500+ rows are JSON-decoded on a background isolate (not on web).
• TecfyDatabase.dbLock is now read-only.

1.2.0

Added

• Optional TecfyDatabase(databaseFactory:, inMemory:) parameters — open with a custom sqflite factory and/or an in-memory database. Fully backward compatible; existing constructors are unaffected.
• New exports: DatabaseFactory, databaseFactoryFfi, sqfliteFfiInit, and inMemoryDatabasePath (useful for writing your own tests).
• Full unit test suite (init/isReady, CRUD, document ops, primary keys, schema migration, field types, filters, search/pagination, batch, streams, error paths) with ~90% line coverage.
• GitHub Actions CI: format check, analyze, tests with coverage, and pub publish dry-run.
• Runnable benchmark harness (benchmark/tecfy_benchmark.dart) with indicative results published in the README.
• Dartdoc comments across the public API.
• Expanded example app into a gallery: complex queries, batch, pagination, realtime streams, and error handling.
• Documentation: FAQ, Troubleshooting, Migration guide, per-platform limitations, and a Production Readiness guide (doc/production_readiness.md).
• Community files: CONTRIBUTING.md, issue templates, and a pull-request template.
• .pubignore to keep internal planning docs and build artifacts out of the published package.

Changed

• dispose() now returns Future<void> so callers can await a clean close. Existing dispose(); call sites remain valid.
• Expanded the pubspec.yaml description and added repository / issue_tracker metadata.

Fixed

• Custom primary keys: a collection declared with a custom primaryField (non-id) now resolves correctly for doc() lookups, updates, deletes, and read-back. Previously such collections queried/returned the literal id column and inserted the key column as NULL.
• No-filter aggregates: searchCount(), searchAny(), and the no-filter count() stream now count all rows. Previously they built WHERE null and always returned 0 / false.
• Corrected the README datetime-filter example to pass an epoch integer.

Known issues

• Indexed datetime columns store microseconds internally while the document body uses milliseconds; the add/get round trip is correct, but datetime filter values must be passed as epoch integers (yourDate.millisecondsSinceEpoch). See the README Troubleshooting section.

1.1.3

• isReady() now waits until every collection's table and indexes are actually created (fixes "no such table" races on fresh installs).
• Creating a second TecfyDatabase instance (e.g. after logout/login) no longer hangs isReady() forever; it reuses the open database and rebuilds the collection operations.
• dispose() unregisters the GetIt database singleton and clears operations so a later re-init works.
• commitBatch() notifies listeners only after the commit completes, so listeners can't re-query before the new rows are visible.
• Export ConflictAlgorithm and Batch so callers can request upsert behavior without importing sqflite directly.

0.0.8

• enhance performance

0.0.1

• Describe initial release.