Changelog #

0.5.4 #

Typed writes now read ids through generated id getters instead of requiring generated document maps to contain the id field.
Decoded schema-backed documents are rehydrated with the external collection-key id on read paths, preserving typed object ids while keeping ids out of stored generated document payloads.
Typed putAll now reuses caller-provided lists instead of copying them before entering the native batch writer.
Native typed batch writes now follow Isar's insert path more closely by avoiding duplicate-id set prevalidation before UPSERT.
MDBX trusted schema batch inserts skip old-index lookup work when the schema has no indexes.
MDBX native query-plan updates now skip per-document change id collection when no watcher is registered, reducing unindexed bulk update overhead.
Native filters avoid repeated schema field lookups after query-plan preparation and use a faster single-byte string contains path.
Sorted MDBX query plans now order borrowed planned documents directly, avoiding intermediate order and option vectors on filtered sort queries.
Updated the development generator dependency constraint to ^0.5.4.
Bumped the native ABI to 29.

0.5.3 #

Fixed MDBX query-plan updates for compact native documents that store ids in the collection key instead of inside the document payload.
Kept compact sortById queries ordered by the external document id without reintroducing embedded ids into compact documents.
Hydrates generated compact binary documents with ids from the collection key on Dart fallback reads.

0.5.2 #

Updated the annotations package constraint to the 0.5.2 documentation and release line.
Aligned generated native MDBX typed writes with the collection-key id path, keeping ids out of compact document payloads and reducing insert size.
Added native document id reads for generated typed hydration.
Improved typed MDBX getAll reads by keeping a native read cursor alive for the batch reader, avoiding full batch materialization before Dart hydration.
Kept the existing batch reader as the fallback for non-MDBX paths and active write transaction reads.
Added a streaming native query reader for unsorted MDBX query-plan reads so typed filter queries can hydrate results as the cursor advances.
Extended the MDBX query reader to stream indexed equality results directly from the index cursor instead of materializing the full result batch first.
Streamed sorted MDBX query results through the native reader, avoiding full document batch copies after sorting.
Added a native string-list reader for generated typed hydration so MDBX getAll can hydrate List<String> fields without a child reader and one FFI call per element.
Added a direct compact bool equality scan for sorted indexed query plans.
Specialized prepared native bool equality filters for compact MDBX binary documents.
Avoided building local watcher change sets for native writes when no watcher is registered, reducing delete and write overhead on the common path.
Bumped the native ABI to 28.

0.5.1 #

Updated the Cindel API package to the 0.5.1 patch line so Flutter app dependency sets can pair it with cindel_flutter_libs 0.5.1, which contains the regenerated native runtime libraries required by the 0.5 API.

0.5.0 #

Updated the Cindel API package and hosted Cindel package constraints to the coordinated 0.5.0 release line.
Prepared native MDBX filters against schema field layouts before scan evaluation, reducing repeated field lookup work in unindexed filter queries.
Aligned native sorted query tie handling with the core comparator path and removed typed reader string interning work from generated native hydration.
Reduced generated typed getAll hydration overhead for native list fields by reusing reader scratch buffers between parent documents and child lists.
Fixed schema binary document decoding for compact native string-list payloads.
Tightened native List<String> payloads to use the compact nested list layout while keeping backward-compatible reads for the previous marker format.
Added native list readers for generated typed hydration, so primitive list fields no longer force query results back through generic Dart binary document decoding.
Reduced unindexed MDBX filter query overhead by applying offset/limit while iterating unsorted native query plans and evaluating string/list contains filters directly from compact binary payloads.
Cached native function resolution across Cindel database openings.
Registered schemas during MDBX native open so typed opens avoid the extra Dart-to-native schema registration round trip.
Cached generated native field layouts for typed MDBX reads and writes.
Reduced MDBX query-plan filter + sort overhead by keeping sorted query documents in borrowed native storage form longer and specializing compact string sort keys.
Reduced MDBX batch delete overhead for compact documents with one simple non-unique bool value index by deleting the index entry directly with prepared cursors.
Reused one MDBX document cursor for batched getAll reads instead of opening a table lookup for every requested id.
Trusted compact static layout metadata for generated MDBX getAll readers, matching the native query-plan reader path.
Added a direct MDBX query-update path for simple non-unique bool value indexes, updating compact document bytes in place and moving the index entry with positioned cursors.
Added native query-plan updates for compact MDBX binary documents, enabling property updates without Dart object hydration.
Fixed geometry and table count settings.
Raised the MDBX default map limit to support large batch updates and deletes without forcing application-level chunking.
Skipped MDBX index rewrites when schema-backed binary updates leave derived index entries unchanged.
Skipped MDBX index derivation during schema-backed batch deletes when the collection schema declares no indexes.
Reused MDBX cursors while cleaning indexes during schema-backed batch deletes.
Reused MDBX document cursors during native batch writes.
Reduced schema binary document encode overhead by avoiding per-dynamic-field header allocations.
Reduced compact binary string index preparation by reading string payloads directly instead of materializing and cloning intermediate values.
Reused prepared MDBX index cursors during schema-backed batch writes.
Deleted schema-backed MDBX index entries through positioned cursors during batch updates.
Reduced native batch writer reallocations by reusing observed document capacity for subsequent documents.
Added nested native list writers for generated typed MDBX batch inserts, including direct string value-record writing for primitive lists.
Reduced generated typed MDBX insert overhead for List<String> fields by storing native string lists with a compact offset table and by sharing the native string buffer between parent and list writers.
Bumped the native ABI to 25.

0.4.0 #

Removed the legacy native benchmark module and optional benchmarks Cargo feature from the committed native package sources.
Aligned the MDBX open profile by storing the physical database as cindel.mdbx with no-subdir mode, NoMetaSync, coalescing, 1 MiB minimum map size, 128 MiB default maximum map size, 5 MiB growth steps, and 20 MiB shrink threshold.
Changed the MDBX default table limit from 2048 to 512.
Removed MDBX_ACCEDE from the default MDBX open flags.
Shortened MDBX document table names and marked the storage layout as mdbx-v4.
Added schema-specific typed binary document storage with a compact static/dynamic layout.
Marked generated typed document storage as binary-v2.
Compacted MDBX index table names and table-local index keys, marking the storage layout as mdbx-v5.
Removed MDBX reverse-index metadata writes from the schema-backed binary document path, marking the storage layout as mdbx-v6.
Optimized MDBX indexed batch writes by preparing document and index tables once per native putAll.
Aligned native schema registration with generated binary document field slots.
Added native compact-document writer and reader handles for generated FFI typed paths.
Reworked the native compact-document writer to use a single object buffer and static null template.
Added an MDBX fast path for no-schema/no-index batch document writes.
Added generated id getters for typed schemas and preserved generated field order in native schema registration.
Added generated native typed writer and reader serializers for supported schema fields.
Reduced native index extraction reparsing for schema-backed binary documents.
Streamed schema-backed indexed native putAll without prepared document batches.
Optimized compact-document index iteration for typed MDBX writes.
Reduced native writer string marshaling overhead for large repeated strings.
Skipped native change-set id tracking for generated writes when no watchers are registered.
Treated generated native-writer documents as trusted schema documents during MDBX typed batch writes.
Optimized ASCII word-index extraction for generated binary documents.
Replaced per-entry MDBX index table hash lookups with prepared table lookup.
Reused MDBX index key buffers during generated typed batch writes.
Added native query-plan document readers for generated typed findAll results.
Added ASCII string reads and native repeated-large-string interning to the generated document reader.
Skipped redundant query-plan presence checks during generated typed materialization.
Added generated direct binary document hydration without intermediate field lists.
Optimized lazy collection and query watchers to avoid eager snapshot hydration.
Aligned typed query watcher emissions with potential-change notifications.
Changed the native release profile from size-optimized to speed-optimized.
Bumped the native ABI to 23.

0.3.4 #

Removed the default-runtime serde_json dependency from the native Rust core.
Replaced legacy native projection and aggregate JSON result payloads with CindelWireV1 ProjectionRows and Scalar buffers.
Replaced SQLite stable list-index JSON keys with canonical binary WireIndexValue bytes encoded as hex.
Removed Dart runtime jsonDecode usage from FFI projection and aggregate paths while keeping public query APIs source-compatible.
Bumped the native ABI to 17.

0.3.3 #

Added CindelWireV1 native watcher change-set payloads for collection name, post-commit revision, and affected document ids.
Wired Dart watchers to committed SQLite and MDBX native change sets so local writes can wake subscriptions without first rereading collection revision.
Kept the public watcher API unchanged and retained pollInterval as the fallback for external database-handle changes.
Added change-set codec fixtures in Dart and Rust plus watcher coverage for compact batched writes and delete id reporting.
Bumped the native ABI to 16.

0.3.2 #

Added CindelWireV1 native query plan payloads for MDBX binary-document queries.
Moved supported findAll, count, deleteFirst, deleteAll, property projection, and property aggregate query shapes onto native plan execution.
Let native planning handle filter, sort, distinct, offset, and limit before Dart object hydration when no Dart-only source filter is required.
Kept the public Dart query builder API unchanged and kept SQLite on the compatibility fallback path.
Added query-plan codec fixtures in Dart and Rust plus backend coverage for native plan behavior.
Bumped the native ABI to 15.

0.3.1 #

Replaced schema registration JSON with CindelWireV1 binary schema manifests.
Replaced SQLite and MDBX collection schema JSON records with explicitly versioned binary schema metadata.
Replaced MDBX reverse document-index JSON metadata with binary IndexEntryList records.
Added fail-closed rejection for JSON-era preview schema metadata, JSON-era reverse-index metadata, future schema metadata versions, and corrupt binary metadata.
Bumped the native ABI to 14 while keeping the public Dart API unchanged.

0.3.0 #

Replaced manual CindelDocument JSON runtime storage with GenericDocumentV1 binary payloads.
Switched getMany FFI results from JSON arrays to binary document batches.
Kept generated typed models on the existing typed binary document format, with manual reads normalizing typed payloads through generated mappers when needed.
Added GenericDocumentV1 Dart tests for nested values, UTF-8 key ordering, and invalid payload rejection.
Bumped the native ABI to 13 while keeping the public Dart API unchanged.

0.2.18 #

Replaced native query filter JSON payloads with CindelWireV1 binary filter AST buffers.
Added WireFilter codec coverage in Dart and Rust so field/group/not filters are tested byte-for-byte across the FFI boundary.
Preserved current filter semantics for null equality, numeric comparisons, string contains/prefix/suffix filters, list contains, object equality, and all/any/not groups.
Bumped the native ABI to 12 while keeping the public Dart API unchanged.

0.2.17 #

Replaced JSON FFI payloads for index values, index entry lists, and manual indexed document write batches with CindelWireV1 binary buffers.
Added WireIndexedDocumentWrite codec coverage in Dart and Rust so indexed batch writes are tested byte-for-byte across the FFI boundary.
Changed hash index canonicalization to hash binary WireIndexValue bytes instead of JSON strings.
Removed the obsolete mdbx-v2-spike benchmark backend and its dedicated storage module now that the real MDBX backend carries the optimized layout.
Bumped the native ABI to 11 while keeping the public Dart API unchanged.

0.2.16 #

Optimized MDBX get-family reads by caching collection schema manifests after registration, avoiding repeated schema metadata reads and JSON parsing during get and getMany style lookups.
Expanded the benchmark harness to measure manual get, raw stored-byte reads, generated binary-document reads, typed get, getAll, batch-size variants, query hydration, and readTxn get loops separately.
Added typed collection coverage confirming MDBX generated writes store the Cindel binary document format used by the fast typed read path.
Kept the public Dart API and native ABI unchanged.

0.2.15 #

Replaced JSON array id-list FFI payloads with CindelWireV1 binary buffers for document ids, getMany, getManyStored, deleteMany, indexed query id results, native filter candidate/result ids, projection candidate ids, and aggregate candidate ids.
Reused CindelWireV1 DocumentWriteBatch for generated binary batch writes.
Bumped the native ABI to 10 for the binary id-list contract.
Kept the public Dart API unchanged; callers still use getAll, documentIds, query builders, projections, aggregates, and deletes the same way.

0.2.14 #

Added the internal CindelWireV1 binary codec foundation shared by Dart and Rust.
Added byte-for-byte Dart and Rust fixture coverage for id lists, index values, scalar results, document write batches, projection rows, schema manifests, and reverse index entry lists.
Added malformed-payload validation coverage for truncated buffers, invalid tags, invalid UTF-8, invalid bool bytes, trailing bytes, and unsafe native item counts.
Kept the public Dart API, native ABI, runtime storage behavior, and prebuilt native binaries unchanged; runtime FFI paths still switch away from JSON in the later JSON-02 and JSON-03 stages.

0.2.13 #

Added CindelPropertyQuery aggregate helpers for count, min, max, sum, and average.
MDBX can execute supported property aggregates over native binary documents without hydrating full Dart objects.
Kept SQLite behavior compatible through fallback aggregate evaluation.
Added FFI bindings for native aggregate queries and bumped the native ABI to 9.
Declared Linux in pub.dev platform metadata now that Linux prebuilt generation is part of release validation.
Completed release hardening for the optimized MDBX default storage path.

0.2.12 #

Promoted the faster MDBX v2 table layout into the real MDBX backend path.
MDBX now stores documents in per-collection tables and indexes in per-index duplicate-sorted tables while preserving the existing public Dart API.
Kept binary documents, explicit transactions, unique/composite/multi-entry indexes, and optimized auto-increment counters on the converged backend.
Added watcher change sets for local writes, including changed document ids and written documents when Dart has them available.
Document and query watchers can skip safe local reads when a write does not affect the watched document or visible query result.
Left the Rust-only mdbx-v2-spike path as a temporary benchmark fixture behind the explicit --backend all-with-spike option.

0.2.11 #

Reduced MDBX allocations in id scans, range/equality query scans, collection stats, and multi-document reads.
Reused MDBX document-key buffers for get_many paths outside active write transactions.
Kept the existing safe MDBX cursor API; lower-level cursor work remains deferred until benchmarks prove it is needed.

0.2.10 #

Optimized MDBX auto-increment id allocation with shared in-memory counters initialized from document primary keys.
MDBX no longer writes persisted counter rows for the normal generated auto-increment insert path.
Added native benchmark operations for id allocation and auto-increment indexed writes.
Fixed explicit SQLite backend opening so it routes through the native backend selector instead of the default MDBX open symbol.

0.2.9 #

Added collection-level composite index schema metadata and generated equality query helpers.
Added multi-entry primitive list indexes for generated membership queries.
Added native composite index key encoding and SQLite compatibility handling.
Updated the annotations dependency to 0.2.1 and the generator dependency to 0.2.3.

0.2.8 #

Added a first native query planner path for MDBX binary-document queries.
Count and simple windowed list queries now use native candidate ids before document hydration when sort/distinct processing is not required.
Added native single-field projection over binary document bytes.
Added FFI bindings for native projection queries and bumped the native ABI to 8.

0.2.7 #

Added native MDBX filtering for generated Cindel filter predicates over binary document bytes.
Routed supported generated filter() and where().filter() queries through native filtering on MDBX, with the existing Dart filter path retained for SQLite and custom predicates.
Added FFI bindings for native filter queries and bumped the native ABI to 7.

0.2.6 #

Added generated binary document callbacks to collection schemas.
Added MDBX typed collection write and read paths that use Cindel binary document bytes instead of JSON map decoding.
Added FFI bindings for raw stored document reads and bumped the native ABI to 6.
Kept SQLite available on the existing JSON document path.
Updated the development generator dependency constraint to the 0.2.2 release line.

0.2.5 #

Switched MDBX schema-backed document storage to Cindel's native binary document format.
MDBX now derives index entries from binary document bytes for generated schema collections.
Made schema-backed MDBX writes strict: unknown fields are rejected instead of falling back to JSON.
Removed the public migration callback and dry-run APIs while Cindel remains pre-1.0 and the optimized storage format is still settling.
Made MDBX part of the default native Cargo build while keeping SQLite compiled as an explicit secondary backend.
Moved the native benchmark and experimental MDBX layout prototype behind the benchmarks Cargo feature so production builds stay smaller.
Updated the native MDBX default path so the Rust engine opens MDBX by default when the native library is compiled with MDBX support.
Bumped the native ABI to 5 for the migration API cleanup.

0.2.4 #

Added internal storage metadata for layout and document format versions.
Added native storage verification helpers for layout, document-format, and index metadata.
Kept the public Dart API, production storage behavior, and native binary symbols unchanged.

0.2.3 #

Added an internal native binary document format prototype.
Documented the planned versioned object layout for future generated serializers.
Added native tests for round-tripping supported field shapes without JSON and reading one field by offset without full document decoding.
Kept the public Dart API, production storage layout, and native binary symbols unchanged.

0.2.2 #

Added an internal MDBX layout prototype for native performance benchmarks.
Added a benchmark backend that compares SQLite, the current MDBX layout, and the layout prototype with the same workload.
Kept the public Dart API, production storage layout, and native binary symbols unchanged.

0.2.1 #

Added an internal MDBX index abstraction boundary for key creation, unique checks, index replacement, deletion, collection clearing, and index stats.
Kept the existing MDBX storage layout and public Dart API unchanged.
Updated the development generator dependency constraint to the 0.2.1 release line.

0.2.0 #

Prepared the Cindel API package for the coordinated 0.2.0 release line.
Updated hosted Cindel package constraints to ^0.2.0.
Refreshed package documentation for MDBX as the default backend, SQLite as an explicit fallback, and the Android/Windows prebuilt binary scope.

0.1.18 #

Made MDBX the default storage backend for new Cindel databases.
Kept SQLite available through backend: CindelStorageBackend.sqlite.
Updated package documentation for the default backend switch.

0.1.17 #

Prepared the first pub.dev development preview.
Switched Cindel package dependencies from local paths to hosted development preview constraints.
Declared Android and Windows as the currently available prebuilt platforms.
Added optional libmdbx dependency and Windows build probe behind the native mdbx Cargo feature.
Documented MDBX-01 build feasibility and LLVM/libclang requirements.
Updated Cindel package dependency constraints for the next development preview.
Removed dates from changelog headings so pub.dev renders version labels cleanly.
Added the internal Rust backend selection boundary while keeping SQLite as the FFI default backend.
Fixed an analyzer lint reported by pub.dev for an unnecessary await.
Added a package example page for pub.dev.
Added a library-level documentation comment for dartdoc/pub.dev analysis.
Added the MDBX key encoding spike for document, index, unique index, and range-bound keys without changing the default SQLite backend.
Added the feature-gated minimal MdbxStorage prototype and benchmark backend selection for SQLite, MDBX, or both.
Expanded the native benchmark to cover open, schema registration, indexed writes, reads, indexed queries, batch writes, and deletes for the MDBX first adoption decision.
Added shared Rust StorageEngine contract tests for SQLite and MDBX.
Brought MDBX storage parity up to the non-explicit-transaction contract, including schema versions, compatible migrations, unique index enforcement, batch rollback, ids, revisions, and indexed reads/writes.
Added MDBX explicit transaction integration through an internal write log that commits staged writes inside one MDBX transaction without storing self-referential transaction handles.
Added shared Rust transaction contract tests for SQLite and MDBX covering commit, rollback, read transaction write rejection, nested transaction rejection, and id allocation rollback.
Added the public CindelStorageBackend option and FFI backend selector so callers can explicitly choose SQLite or MDBX while SQLite remains the default backend.
Added backend-parameterized Dart tests so the full Cindel package suite can run against SQLite or MDBX.
Added MDBX read-your-writes parity for staged write transactions and shared same-directory MDBX environments across multiple Dart database handles.
Updated package installation snippets for the MDBX-enabled Android and Windows prebuilt binaries.

0.1.16 #

Added putMany as a public alias for manual and typed atomic bulk writes.

0.1.15 #

Added pub.dev-oriented package metadata and documentation.
Added package-level README guidance for installation, generation, and core database features.

0.1.14 #

Added optimized native batch document reads.
Regenerated Windows and Android prebuilt native libraries for ABI 4.

cindel 0.5.4 cindel: ^0.5.4 copied to clipboard

Metadata