Changelog #

1.0.0 #

Released the first stable Cindel package version.
Updated internal package constraints to the 1.0.0 release line.

0.9.3 #

Fixed native generated serialization for List<String> fields inside embedded objects, including embedded objects stored inside lists, so MDBX and SQLite native backends hydrate those values as lists instead of string payloads.
Fixed MDBX CindelQuery.watch() for native-planned queries so typed document snapshots are materialized correctly instead of reporting a missing generated native reader.
Fixed SQLite Web native generated rows for embedded object and embedded object-list fields, case-insensitive unique replace lookups, indexed equality fallbacks, nested embedded filters, and anyOf/allOf modifier queries.
Bumped the native ABI to 34 for the typed native writer symbols that preserve dynamic value kinds for string, list, and object payloads.

0.9.2 #

Fixed the Web database facade so CindelDatabase.deleteAll(collection, ids) is exposed like native backends while still using the SQLite-native typed row delete path.

0.9.1 #

Published in lockstep with cindel_annotations, cindel_generator, and cindel_flutter_libs 0.9.1 so pub.dev can resolve the sync-enabled release outside the workspace.
Updated internal package constraints to require the matching 0.9.1 release line.

0.9.0 #

Added experimental Sync, enabled only when opening the database with Cindel.open(..., sync: CindelSyncConfig(...)).
Added backend-neutral CindelSyncAdapter push/pull contracts, sync status callbacks, durable internal outbox storage, checkpoint persistence, and automatic background scheduling without exposing public sync commands.
Sync now captures typed puts, bulk puts, unique-index puts, deletes, query deletes, and link replacements across SQLite native, MDBX, and SQLite Web.
Remote changes apply through internal guarded writes so watchers update while remote apply does not re-enqueue outgoing mutations.
Query update operations now throw while sync is enabled because they do not yet produce canonical per-document mutations.

0.8.0 #

Added CindelLink<T> and CindelLinks<T> relation containers with explicit load, save, and reset operations.
Added persisted forward links and read-only backlinks across SQLite native, MDBX, and SQLite Web/OPFS, including relation validation, source-collection watcher notifications, and native ABI/Web worker bindings.
Added generated link schema metadata and runtime binding hooks so hydrated objects can load and save relations through the typed API.

0.7.0 #

Added public open-time data migration tooling with database-level migration versions, ordered migration steps, before/after verification callbacks, paged typed export helpers, typed/document import helpers, migrated schema registration, and optional backend compaction.
Added native migration primitives for SQLite, MDBX, and Web SQLite/OPFS: migrated schema registration, migration version persistence, and compact requests through the Dart FFI/Web worker surfaces.
SQLite migrated schema registration now rebuilds target collection tables so renamed, removed, or type-changed fields can be rewritten safely by the migration callback before the app receives the final handle.
MDBX migrated schema registration now clears target collection documents and indexes before importing rewritten target data while preserving normal compatibility checks for non-migration opens.
Added migration coverage across SQLite native, MDBX, and the Web public worker contract.
Added CindelDatabase.documentIdsPage for bounded, ascending document-id scans across SQLite native, MDBX, and Web SQLite so backup/export tooling does not need to materialize entire collections.
Added public typed full-database backup/export/import streams with JSONL v1 archives, checksum verification, progress callbacks, native gzip compression, Web-compatible uncompressed archives, and empty-target restore validation.
Bumped the native ABI to 32 for the paged document-id FFI/Web surface.

0.6.5 #

Updated package metadata so the published documentation link points to the extended Cindel API guide.
Updated the hooks dependency constraint to support the latest stable release used by pub.dev scoring.
Fixed a dangling Web worker bridge library documentation comment reported by static analysis.

0.6.4 #

Aligned MDBX, SQLite native, and SQLite Web around the generated typed public API. Applications now use the same Cindel.open(...), typed collections, typed queries, transactions, and watchers across supported platforms.
Added the SQLite Web/OPFS backend behind the normal Cindel.open(...) path, loading packaged Worker/Wasm assets from cindel_flutter_libs.
Removed the separate public Web entrypoint and removed untyped collection-level document APIs from the current public direction.
Completed the Web typed runtime surface for CRUD, bulk reads/writes, unique-replace helpers, query plans, projections, aggregates, updates, deletes, transactions, collection revisions, change-set draining, and single-tab watchers.
Optimized SQLite native/Web typed writes and getAll reads with direct generated-row batch encoding, prepared-statement reuse, chunked id queries, and stable requested-order restoration.
Added SQLite native query-plan projection and aggregate execution for typed schema-backed rows, including count, min, max, sum, and average.
Kept MDBX as the default native backend and reference path while aligning SQLite and Web SQLite behavior with the same typed app workloads.

0.6.3 #

Added the native Web SQLite baseline behind the web feature, keeping MDBX as the default native backend.
Added OPFS SAHPool setup for Web workers and explicit SQLite opfs-sahpool opening for persistent Web storage.
Added the minimal Web worker transport bridge for request ids, structured errors, close cleanup, and transferable payloads.
Added Web schema manifest encoding and a Wasm Web engine opener that registers SQLite schema metadata persistently.
Kept runtime collection schemas loaded after persisted Web SQLite opens so native document cursors remain available after OPFS reopen.
Added the Web typed CRUD surface on CindelWebEngine and the package worker template: id allocation, put/putAll, get/getAll, getStored/getAllStored, delete/deleteAll, documentIds, and SQLite-native document write/delete batches.
Added Web CRUD wire helpers for id lists, indexed document batches, optional get results, and SQLite-native generated document batches.
Added internal release packaging support for the SQLite-only Wasm runtime assets emitted under packages/cindel_flutter_libs/web/pkg.

0.6.2 #

Added @Name support for persisted collection and field names so Dart model identifiers can differ from the stored schema names.
Added unique replace indexes with @Index(unique: true, replace: true) and CompositeIndex(..., unique: true, replace: true).
Generated putBy... and putAllBy... helpers for unique replace indexes so typed objects can upsert by natural keys without a manual query.
Updated SQLite, MDBX, schema metadata, and wire schema encoding to preserve replace-index metadata and delete conflicting documents during replace writes.
Persisted MDBX auto-increment high-water counters so deleted high ids and allocate-only ids are not reused after reopening the database.
Updated the annotations dependency constraint to ^0.5.4 and the development generator dependency constraint to ^0.6.1.

0.6.1 #

Added documented Cindel-specific runtime errors exported from the public API, while keeping them compatible with existing StateError handling.

0.6.0 #

Added CindelQuery.optional, CindelQuery.anyOf, and CindelQuery.allOf dynamic query modifiers for conditional and repeated filter composition.
Added list-length and list-empty filter predicates, plus generated list helpers such as tagsIsEmpty, tagsLengthGreaterThan, and tagsLengthBetween.
Added generated nested filter support for fields inside embedded object list elements, such as recipientsElement((recipient) => ...).
Added a Rust native embedded object writer/reader path for generated typed serializers, avoiding intermediate Dart document maps for embedded object fields, embedded object lists, and nested embedded objects.
Added native FFI entry points for embedded object child writers and readers, including list-element embedded objects and nested embedded payloads.
Made embedded object equality and embedded-list element equality compare stored map/list values deeply instead of by Dart object identity.
Added current-document native reader FFI entry points for dynamic bytes and list fields, keeping generated MDBX point reads aligned with the active native cursor document.
Improved generated typed MDBX getAll hydration by avoiding indexed dynamic field reads after the cursor has already selected the current document.
Updated the development generator dependency constraint to ^0.6.0.

0.5.13 #

Declared iOS and macOS as supported Flutter prebuilt platforms through cindel_flutter_libs.
Added automated tag-based release publishing support for the coordinated cindel and cindel_flutter_libs release line.

0.5.12 #

Streamed MDBX query-plan count and property aggregate execution for supported plans, avoiding full planned-document materialization while preserving the existing fallback for distinct and unsupported plans.
Streamed MDBX query-plan property projection for supported plans, writing projection rows directly while preserving the existing fallback for distinct and unsupported plans.
Made vendored MDBX bindgen generation pass Cargo's target explicitly so Windows native builds use the correct FFI type layout.

0.5.11 #

Avoided creating Cindel-only SQLite storage metadata for generated schema-aware SQLite opens, keeping the typed collection table layout while still validating existing metadata when present.
Added a direct native string-list writer used by generated typed serializers to reduce MDBX List<String> batch insert overhead.
Added MDBX native query update support for compact dynamic fields, including generated List<String> fields.
Avoided an MDBX document copy while generated native readers hydrate point reads and unsorted query results.
Added MDBX streaming for unsorted generated index range query results.
Added a fast generated SQLite string-list hydration path for native reads, while preserving Unicode and escaped-string fallbacks.
Kept generated typed APIs on document hydration after generic writes to the same collection.
Updated the development generator dependency constraint to ^0.5.9.

0.5.10 #

Added native typed reader and writer support for embedded object fields and embedded object lists.
Added generated nested filters for fields inside a single embedded object.
Kept embedded object-list nested query helpers unsupported.
Fixed SQLite native typed reads and projections for generated schemas that store embedded/list payloads in collection tables.
Updated the development generator dependency constraint to ^0.5.8.

0.5.9 #

Updated the development generator dependency constraint to ^0.5.7.
Added generated schema coverage for Freezed primary-factory collection models, including factory-parameter @Index and @ignore annotations.

0.5.8 #

Changed the generated collection id convention from id to dbId, leaving id free for API/domain fields in user models.
Removed the old CDBF document compatibility path so generated compact documents use the current backend format as the only supported storage path.
Kept generated document ids in the storage key path while allowing normal indexed id fields to be stored, queried, sorted, projected, and aggregated like any other model field.
Expanded typed filter helpers so generated contains filters can represent string, list, and embedded-value containment without narrowing the public filter value to String.
Updated tests and documentation for the dbId id field convention.

0.5.7 #

Completed the generated typed SQLite app backend path for the current public surface: schema-backed open, typed putAll, getAll, query updates, query deletes, filter queries, sorted queries, and direct native typed reads now use SQLite collection tables.
Removed the remaining generated SQLite query fallback to generic stored documents for supported typed query plans.
Added native SQLite query update and filter execution for generated collections.
Reduced extra MDBX typed insert work by reusing cached schema metadata during trusted unindexed batch writes.
Updated the development generator dependency constraint to ^0.5.5.

0.5.6 #

Aligned typed SQLite getAll with Isar's cursor path by keeping a prepared SQLite statement alive across the batch, binding each requested id, and reading generated fields directly from the active SQLite row.
Wrapped the generated SQLite getAll cursor in a native read transaction, matching Isar's getTxn read path for cursor reads.
Added direct-return native reader calls for generated id, bool, int, double, and string reads so the Dart hydration path mirrors Isar's reader FFI shape.
Removed the generated SQLite reader's fallback to the generic stored-document batch path; generated SQLite reads now use the schema collection table path.

0.5.5 #

Aligned the SQLite schema-aware open path with Isar by registering generated schemas during native open instead of opening first and registering schemas through a second native call.
Configured SQLite mmap_size during open to match the Isar SQLite benchmark profile.
Wrapped SQLite open-time table, metadata, and schema setup in one native transaction and added coverage for opening SQLite with schemas already available.
Aligned typed SQLite putAll with Isar's collection-table insert path by routing generated native writers into per-collection SQLite tables instead of the generic documents blob table.
Added coverage for native SQLite inserts proving generated values land in collection columns without creating generic document rows.

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 settling the optimized storage format.
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 1.0.0 cindel: ^1.0.0 copied to clipboard

Metadata