rpc_dart_data 3.4.4

Data service layer (CRUD + querying + change streams) built on rpc_dart.

3.4.4

• SQLite: chunked reads switch to keyset pagination to avoid large OFFSET scans; queryCollection now uses keyset pagination for cursors (OFFSET kept only when explicitly requested); bulk FTS updates insert per chunk in a single statement after one delete.
• SQLite: batched writeRecords checks existing versions once before upsert and throws conflicts without an extra SELECT per conflict.
• Postgres: chunked reads and queryCollection use keyset pagination (OFFSET only when explicitly requested); batched writes prefetch versions once to surface conflicts without a second query.

3.4.3

• Web: SqliteConnectionOptions adds webVfsMode (opfs/inMemory/custom), webFileName, and optional webCustomVfs. Web connections now follow the drift-style path: OPFS default, fall back to IndexedDB, then in-memory when OPFS is unavailable.

3.4.2

• IDataStorageAdapter now requires ensureReady, Postgres adapter aligns readiness with schema/journal/index validation.
• Migration runner accepts a migration-capable repository interface (works with Postgres/SQLite), and Postgres repository exposes migration helpers.
• README/doc tweaks for migration helper wording.

3.4.1

• Wasm compatibility

3.4.0

• Postgres adapter: per-collection JSONB tables with server-side filters/sort/pagination/FTS in Postgres, JSON-path indexes, change journal and schema registry tables.
• Import from SQLite into Postgres without changing contracts (export/import stay at repository level).
• All system timestamps (createdAt/updatedAt/occurredAt) are stored in UTC; added UTC coverage for SQLite and Postgres.
• JSON path validation in the Postgres adapter (segment sanitization) to block SQL injection via sort/filter/index fields.
• Postgres integration tests (CRUD, filters/indexes, FTS, UTC) plus SQLite→Postgres import test; SQLite tests now assert UTC.

3.3.1

• Update rpc_dart to 2.4.0

3.3.0

• Breaking: removed skipValidation flags from all RPC models and repository paths; schema validation is always enforced when a schema is enabled/requireValidation=true.
• Added SchemaValidationConfig.allowOverrideMigrations guard; override migrations now require explicit config and still validate transformed payloads, while recording schema history.
• Migration RPC surface removed; migrations run only server-side via repository/helpers.
• Initial migrations can supply a transformer; schema overrides write to collection_schema_history.
• Docs/examples updated; schema toggle now via policy only (no skip flags); maxPayloadBytes guard removed.

3.2.0

• Breaking: removed offline sync/command queue APIs (syncChanges/pushAndAwaitAck/OfflineCommandQueue, DataCommand/SyncChange models) to focus on online CRUD/search/watch; RPC contract, caller/responder, and repository helpers are trimmed accordingly.
• Breaking: removed aggregate metrics API and adapter implementations; clients should compute metrics externally if needed.

3.1.0

• Schema validation: enforced on create/update/patch/bulk/import (including RPC); per-collection policy flags with defaults, nested object/array support, detailed validation errors.
• Schema registry: SQLite tables for active/history schemas, policy, migration checkpoints/logs/errors; export/import includes schemas and validates on import.
• Migrations: server-side runner with checkpoints/logging, index/FTS rebuild after migration, fail-fast vs collect errors (errors block activation). Migrations are now bound to a specific collection (MigrationDefinition.collection), and a single SchemaValidationEngine is shared in SqliteDataRepository.
• RPC surface trimmed: migration RPC methods removed from public client; list/get/setPolicy remain for diagnostics, migrations executed server-side via repository/runner helpers.
• Examples/tests updated to use MigrationRunnerHelper and schema validation over RPC; added patch validation coverage and nested schema tests.

3.0.0

• Breaking: upgraded to sqlite3 3.x, which loads native binaries via hooks/native assets; the old DynamicLibrary override path is gone.
• Encryption: SQLCipher works out of the box via SQLite3MultipleCiphers (hooks.user_defines.sqlite3.source: sqlite3mc), no system libsqlcipher required; web uses sqlite3mc.wasm.
• API: replaced dispose() with close() on sqlite objects per sqlite3 3.x requirements.
• SQLCipher helper now sets SQLCipher compatibility pragmas (cipher='sqlcipher', legacy=4) before applying the key and detects support in sqlite3mc builds.

2.2.1

• Web: fixed dart2js/wasm build failures by parsing 64-bit FNV constants as BigInt, keeping stable index-name hashes without exceeding JS number precision.

2.2.0

• Breaking: removed tenantId from DataRecord, adapters, and SQLite schema/indexing. Multi-tenant isolation now belongs to higher layers; table footprint and indexes are smaller.
• Correct optimistic locking: SQLite writes now use conditional INSERT/UPDATE/DELETE (version-guarded upsert, version-aware delete) and surface RpcDataError.conflict when the incoming version is stale. Bulk upsert detects per-record conflicts.
• SQLite search ranks by relevance (bm25) then id, fixing alphabetical ordering, and RecordFilter.containsTerms is supported for parity with in-memory adapter.
• Delete APIs accept expectedVersion end-to-end (storage + repository) to prevent stale deletions.

2.1.1

• Expanded IDataService API docs in English with detailed semantics for pagination, optimistic locking, bulk ops, exports/imports, search/aggregate, indexes, change streams, offline sync, and shutdown guidance.
• Breaking: DataServiceCaller, DataServiceResponder, and the facade/server factories now require an explicit RpcDataTransferMode in constructors; wire this through when instantiating clients/servers.

2.1.0

• SQLCipher: when a SqlCipherKey is provided, the library now automatically overrides the loaded SQLite binary on macOS/Linux to use SQLCipher (SQLITE3_LIB_DIR/SQLITE3_LIB_NAME or common Homebrew paths) before opening the database; still fails fast when cipher pragmas are missing so encrypted DBs can’t silently fall back to plaintext.
• Web: switched the WASM loader to sqlite3mc.wasm (SQLite3MultipleCiphers) by default so cipher-enabled builds work consistently across platforms; still respects a custom webSqliteWasmUri if supplied.
• Refined SQLCipher helpers and tests to support repeated open/close cycles on the same encrypted file (idempotent table creation, stable paths).

2.0.0

• Added ListCollectionsRequest/ListCollectionsResponse plus RPC support (IDataServiceContract, caller, responder, IDataService) so remote clients can ask “what collections exist?” without scanning data manually.
• Extended repositories/adapters with listCollections() and wired the RPC integration test to cover the new flow, ensuring collection names come straight from storage.
• Broadened RPC coverage by refactoring codecs, contracts, and facades while introducing suites for change journals, models, JSON helpers, SQLite/daemon helpers, SQLCipher, and RPC authorization/exception paths—lifting previously low coverage areas.
• Confirmed SqliteDataStorageAdapter already provides FTS5 and exposes SqliteSetupHook so callers can enable WAL/PRAGMA hooks before traffic hits the database.
• Documented that SqliteDataDatabase.transaction wraps BEGIN/COMMIT/ROLLBACK, giving true SQLite transactions for batch updates and journaled change streams.
• Removed the drift and rpc_dart_transports runtime dependencies to keep the package lean—everything now routes through the core storage adapters and RPC helpers already shipped here.

1.2.0

• Added streaming NDJSON exports with payloadStream plus the ability to skip the in-memory string via ExportDatabaseRequest(includePayloadString: false). Export now prefetches collection chunks and respects consumer backpressure.
• Imports validate the snapshot stream before mutating data and process records in databaseImportBatchSize chunks, which keeps memory flat even for very large dumps.
• DataServiceClient gained high-level helpers: listAllRecords, bulkUpsertStream, pushAndAwaitAck, createOfflineQueue, and close.
• Storage adapters can expose custom SQLite setup logic through SqliteSetupHook, and the SQLite adapter batches UPSERT statements plus supports chunked readRecords to reduce I/O.
• Introduced RpcStreamIterator (based on StreamQueue) so all streaming exports respect consumer demand.
• Documentation rewritten in English and aligned with the new streaming/offline features.

1.1.0

• Добавлен персистентный журнал изменений для watch() и sync() с поддержкой восстановления курсоров после рестарта.
• SQLite-хранилище научилось создавать индексы, выполнять фильтрацию и пагинацию на стороне SQL, а также использовать SQLCipher-ключ из PASERK.
• Для подготовки multi-tenant сценариев таблицы коллекций теперь включают столбец tenantId с индексом и поддержкой фильтрации/сортировки.
• CLI serve объединён в универсальный ServeCli с поддержкой SQLCipher, SecureWrap и белых списков bearer-токенов для backend-сценариев.
• README уточняет, что прямой доступ конечных клиентов не поддерживается из коробки: сервис предполагает работу через доверенный backend.
• Добавлено руководство по изоляции данных между клиентами (tenant-ами) при backend-only использовании, включая рекомендации по производительности при использовании отдельных коллекций.

1.0.0

• Initial version.