Releases: bazer/DataLinq
DataLinq v0.6.8 - Cache Diagnostics, Correctness, and Test Infrastructure
This is a maintenance-heavy release, but it is not filler. The important work here is cache correctness and observability: a real cache-eviction bug was fixed, cache-notification cleanup is more reliable, and DataLinq now ships its first real metrics API with a hierarchy that matches the actual runtime shape. Around that, this release also starts and finishes a large test-infrastructure transition: the project moves from the old mixed xUnit setup to a TUnit-centered test architecture with new CLI and CI support, and adds a proper benchmark harness.
Highlights
-
Fixed a real cache correctness bug where table-specific cache cleanup could evict rows from every table.
A table-level row limit inRemoveRowsBySettingswas incorrectly routed through the database-wide cleanup path. In practice, that meant cache limits configured for one table could cause unnecessary evictions and extra churn in unrelated tables. That behavior is now fixed, and regression coverage was added to make sure it stays fixed. -
Cache-notification cleanup is more robust and much easier to diagnose.
CacheNotificationManagernow compacts dead weak subscribers correctly again in read-heavy workloads, andNotify()/Clean()were tightened so they do not race each other and lose invalidations. DataLinq also now exposes live notification telemetry such as queue depth, sweep sizes, dropped dead subscribers, and peak queue growth. -
DataLinq now ships a real hierarchical runtime metrics API.
DataLinqMetricsis new in this release and reports:- runtime totals
- per-provider-instance metrics
- per-table metrics within each provider
That matters because query execution, row-cache behavior, relation loading, and cache-notification churn do not belong to the same scope. The shipped API avoids misleading aggregation and makes multi-provider diagnostics much more trustworthy from day one.
Runtime Fixes and Observability
- Added a new public diagnostics surface under
DataLinq.Diagnostics.DataLinqMetrics, including typed snapshot models for runtime, provider, table, query, relation, row-cache, and cache-notification metrics. - Scoped cache-notification telemetry by provider instance and table so multiple loaded providers with the same logical database name are tracked independently.
- Added stable provider telemetry instance ids so aggregation does not collapse unrelated provider instances together.
- Added live and cumulative cache-notification metrics, including current queue depth, last notify/clean sweep values, sweep totals, dropped dead references, busy clean skips, and approximate peak queue depth.
- Added a dedicated diagnostics and metrics documentation page that explains how to interpret the new hierarchy and which values are counters, gauges, sums, or maxima, without pretending there was an earlier released flat metrics API.
- Fixed a
ThreadWorkerteardown race during fast disposal, improving shutdown reliability in scenarios that rapidly create and dispose providers.
Benchmarking, Testing, and Tooling
- This release contains a full test suite migration from xUnit to TUnit. The project moved to a TUnit-centered structure across unit, compliance, and generator coverage.
- Added a cross-platform
DataLinq.Testing.CLIworkflow for bringing test infrastructure up/down, waiting, resetting, running suites, listing targets, and validating legacy-to-TUnit parity. - Moved the test suite completely to Podman containers, with support for all the current LTS versions of MySQL and MariaDB.
- Added a parity gate so legacy xUnit coverage cannot silently disappear during the migration.
- Cleaned up the test structure and provider matrix so the suite is easier to reason about and less dependent on ad hoc local scripts.
- Added real CI for the project, including a main automated lane plus broader matrix coverage, instead of relying on purely local validation.
- Built that CI around the new testing workflow with more resilient teardown, dedicated MySQL/MariaDB coverage, and machine-readable summaries for badges and reporting.
- Replaced the old benchmark stub with a real BenchmarkDotNet harness, including deterministic SQLite-backed employee benchmarks for cold/warm primary-key fetches and relation traversal.
Documentation and Maintenance
- Added first-class documentation for diagnostics and metrics, and linked it from the README, site index, and usage docs so it is actually discoverable.
- Reorganized development-plan docs and refined roadmap/async planning material.
- Refreshed NuGet dependencies across the solution.
Full Changelog
DataLinq v0.6.7 - Generator Reliability, Default Handling, and Release Tooling
This release is mostly about correctness and maintainability, and that is exactly what it needed to be. The biggest themes are a cleaner source-generator pipeline, much better handling of default values across providers, several SQLite and MySQL/MariaDB correctness fixes, a large documentation overhaul, and a far more practical local NuGet publishing workflow.
Highlights
-
Replaced the old SGF-based generator pipeline with a native Roslyn incremental generator.
This is the most important internal change in the release. It reduces moving parts, aligns the generator with the platform it actually runs on, and gives DataLinq a more stable foundation for future analyzer and generation work. -
Default value handling is significantly more correct across generation, metadata parsing, and SQL output.
A large portion of this release fixes subtle but important bugs around default values:- generated models now preserve source defaults more accurately, including overridden property types
- default literal escaping has been fixed in generated models
- MySQL, MariaDB, and SQLite now parse and emit default values more reliably
- typed default compatibility is validated more aggressively during generation
-
SQLite behavior is more consistent and less fragile.
This release fixes several SQLite-specific issues:- in-memory database lifetime and test isolation were improved
Guidparameter matching forTEXTcolumns was corrected- millisecond precision handling was aligned more closely with .NET
DateTimebehavior - SQLite default value parsing and SQL generation were expanded and tightened up
-
MySQL and MariaDB SQL/default handling got a substantial correctness pass.
Multiple fixes in this release address quoted defaults, typed model properties, date defaults, enum defaults, view parsing fallback behavior, and SQL generation for provider-specific edge cases.
LINQ and Query Fixes
- Fixed LINQ
charequality translation across SQLite, MySQL, and MariaDB. - Corrected several provider-level query and metadata edge cases that were previously easy to miss but could produce the wrong SQL or incorrect defaults.
Source Generator and Analyzer Improvements
- Added analyzer release tracking for
DLG000. - Improved validation for model default values.
- Tightened generator test coverage around defaults, syntax parsing, and model generation behavior.
- Fixed transitive Roslyn/source-generator packaging issues so the NuGet experience is more reliable in Visual Studio and downstream projects.
Packaging and Tooling
- Added a new local
publish-nuget.ps1release script for packing and publishing public packages. - The script now stages release artifacts in a fresh folder, prompts for the NuGet API key at publish time, and publishes packages and symbol packages explicitly.
- Fixed
DataLinqsymbol packaging so.snupkgfiles actually contain real PDBs and can be published successfully. - Improved the local release flow for
DataLinq,DataLinq.SQLite,DataLinq.MySql,DataLinq.CLI, andDataLinq.Tools.
Documentation
- Performed a broad documentation overhaul and website restructuring.
- Added or substantially improved docs for:
- installation and getting started
- configuration and model generation
- CLI usage
- LINQ query support
- transactions
- troubleshooting
- backend-specific behavior for SQLite and MySQL/MariaDB
- Fixed docfx homepage routing and cleaned up the site structure.
Full Changelog
DataLinq v0.6.6 - Performance Improvements and SQLite Logging
This maintenance release improves core query and materialization performance, reduces memory overhead in hot paths, and makes SQLite logging behavior more consistent. It also includes a dependency refresh, source generator cleanup, and a large set of internal planning documents for upcoming releases.
Highlights
- Faster Primary-Key Lookups and LINQ Execution: Several internal optimizations make common read scenarios faster and cheaper. [31fd6f3]
Selectcan now detect simple primary-key predicates and short-circuit to a direct lookup instead of building and executing a full query.- Expression evaluation now uses a reflection-based fast path for local variable access, avoiding unnecessary lambda compilation in many cases.
QueryExecutornow caches standard identity projection delegates to reduce repeated overhead for common queries.
- Lower-Overhead
RowDataStorage:RowDatahas been redesigned to use an indexed object array instead of a dictionary, giving O(1) column access and reducing allocations during row materialization. [2ce6b41]- This change is supported by new column indexing metadata assigned during metadata parsing, ensuring correct alignment even for partial
SELECTqueries.
- This change is supported by new column indexing metadata assigned during metadata parsing, ensuring correct alignment even for partial
- Improved SQLite Logging Integration: Logging configuration is now propagated more consistently through SQLite database and transaction classes, improving diagnostics and making SQLite behavior more aligned with the other providers. [db25a31] [07fc896]
Internal Improvements
- Source Generator Cleanup: Refactored
ModelGeneratorto streamline class declaration processing and improve metadata caching in the generator pipeline. [3300f0f] - Dependency Refresh: Updated NuGet package dependencies across the runtime, generator, tooling, benchmark, and test projects to newer stable versions. [d16ae98]
- Minor Code Cleanup: Removed unused
usingdirectives and small bits of dead code as part of the performance work. [ba088a7]
Documentation & Planning
- Added a substantial new set of development-plan documents covering batched mutations and optimistic concurrency, in-memory provider support, JSON data type support, metadata architecture, migrations and validation, performance benchmarking, projections and views, query pipeline abstraction, source generator optimizations, SQL generation optimization, result-set caching, testing infrastructure, and recommended application patterns. [59f3de0] [2242bc3] [c2f3547] [ef00311]
- Updated the documentation workflow to use .NET 10 for static site generation. [ebf70a7]
Full Changelog: 0.6.5...0.6.6
DataLinq v0.6.5 - LINQ Enhancements & Multi-Targeting
This release expands framework support to include .NET 8, 9, and 10, introduces significant improvements to the LINQ query parser for string manipulation and collection handling, and includes internal optimizations for newer .NET runtimes.
Highlights
- Multi-Targeting Support: DataLinq now explicitly targets .NET 8.0, .NET 9.0, and .NET 10.0.
- Performance Optimization on .NET 9+: Implemented conditional compilation to utilize the new
System.Threading.Lockon .NET 9 and greater, improving thread synchronization performance inImmutableRelationandImmutableForeignKey. - Advanced LINQ Chains: Added support for chained string functions in queries. You can now write LINQ expressions like
x.Name.Trim().Length, and they will correctly translate to the corresponding SQL.
LINQ & Query Engine
- Chained String Functions: The
QueryBuildernow supports parsing and generating SQL for chained string operations (e.g.,Trim().ToUpper().Length). - Enhanced Collection Handling: Improved translation logic for
ContainsandAnymethods.- Added robust handling for empty lists in
Containsqueries (resolving to1=0or1=1). - Fixed handling of negated
Containsconditions. - Added support for
op_Implicitcalls wrapping arrays or spans within queries.
- Added robust handling for empty lists in
- Entity Selection: Improved
QueryExecutorto handle selecting the full entity directly in projections (e.g.,source.Select(x => x)). - Expression Evaluation: Updated the
Evaluatorto safely handle non-reducible expressions (likeQuerySourceReferenceExpression), preventing runtime errors during partial evaluation of query trees.
Bug Fixes & Internal Improvements
- Dependency Update: Updated
ThrowAwaypackage to version 0.3.1 and updated various test dependencies (Bogus, xUnit, etc.). - SQL Generation: Added argument validation for
Substringfunctions in SQL providers to ensure correct usage. - Test Suite Reliability: Adjusted transaction tests to correctly account for isolation level differences between SQLite (which may expose uncommitted writes) and MySQL/MariaDB.
- Type Safety: Added specific handling to skip unsupported tests in MariaDB relating to GUID formats until upstream connector support is clarified.
Full Changelog: 0.6.4...0.6.5
DataLinq v0.6.4 - Critical Concurrency & Performance Fixes
This is a high-priority release that resolves critical performance and stability issues related to the relation caching system under high thread contention. It introduces a more robust, leak-free, and highly performant pattern for handling cache invalidation notifications.
🚀 Highlights
- Fixed Critical Threading & Performance Issue: A major bug has been fixed where applications with a high number of loaded relations and concurrent threads could experience severe performance degradation or hangs.
- The
CacheNotificationManagerhas been re-engineered to use a "fire-and-forget" pattern with aConcurrentQueue. This makes theSubscribeoperation a lock-free, O(1) action, drastically improving performance in scenarios with many relation accesses. - The
ImmutableRelationandImmutableForeignKeyclasses have been hardened with a robust double-checked locking pattern usingvolatile, ensuring that lazy-loaded data is fetched only once and is safe from race conditions, while keeping the "hot path" for accessing already-loaded data lock-free and extremely fast.
- The
🐛 Bug Fixes & Internal Improvements
- Resolved High-Contention Concurrency Bugs: Replaced the previous cache notification logic with a new, more robust implementation to prevent thread starvation and potential hangs. This completely overhauls the internal mechanics of relation cache invalidation for better performance and stability. [c1f7380]
- Fixed Test Suite Initialization: Corrected a bug in the
DatabaseFixturethat could prevent test databases from being set up correctly in certain configurations. [8a1c76c]
Full Changelog: 0.6.3...0.6.4
DataLinq v0.6.3 - Improved Key Handling and Robustness
This release focuses on improving the internal robustness of the core data access logic by addressing a key deficiency in index handling and providing a more resilient and correct foundation for future development.
🚀 Highlights
- Enhanced Key Generation and Indexing: Improved handling of key types in the index. The
KeyFactoryhas been refactored to correctly generate and compare primary keys for all supported data types. This fixes a potential issue where indexes on specific column types (byte[], enums, and other value types) might lead to incorrect results or stability problems.
🐛 Bug Fixes
- Fixed Incorrect Indexing with
byte[]and Enums: Resolved a critical bug in theKeyFactorywhere the comparison logic forbyte[]and enum-based primary keys did not correctly consider the content of thebyte[]data or the enum value. This could lead to incorrect lookups in caches and index maintenance, especially for the indexes that reference those columns. [91a8eaa]
Full Changelog: 0.6.2...0.6.3
DataLinq v0.6.2 - Performance, Stability, and LINQ Fixes
This is a focused maintenance release that addresses critical performance issues under high thread contention, improves the correctness of the code generation CLI, and adds a commonly requested feature to the LINQ provider.
🚀 Highlights
- Fixed Critical Performance Issue in Cache Notifications: Resolved a thread starvation bug in the
CacheNotificationManagerthat occurred under high-contention scenarios (many threads, thousands of relations). The previous lock-freeSubscribemethod has been replaced with a more robust and efficient lock-based write and lock-free read pattern, eliminating excessive CPU usage and potential application hangs. [11323da] - Implemented
string.Lengthin LINQ Queries: Added support for using the.Lengthproperty on string columns within LINQWHEREclauses. [ff3d480]
🐛 Bug Fixes & Improvements
- Corrected Nullable Reference Type Generation in CLI: The
datalinq create-modelscommand now correctly respects the"UseNullableReferenceTypes": falsesetting indatalinq.json. It will no longer incorrectly add a?to nullable reference types (likestring) when the feature is disabled, ensuring the generated abstract models are correct for non-NRT projects. [d122ed9] - Improved CLI Help Text: Corrected a misleading error message in the CLI that suggested an unimplemented
--alloption, preventing user confusion. [a26a8f5]
Full Changelog: 0.6.1...0.6.2
DataLinq v0.6.1 - Stability and Code Generation Fixes
This is a maintenance release that focuses on improving the correctness and robustness of the metadata parsing and source generation engines. It resolves critical bugs related to recursive table relationships and properties that serve as both a primary and foreign key. It also cleans up all known C# compiler warnings in the generated model code for a smoother developer experience.
🐛 Bug Fixes
- Fixed Recursive Relation Parsing: A critical bug was fixed where a table with a self-referencing foreign key (e.g., an
employeetable with amanager_id) would cause the metadata parser to generate incorrect and duplicate relation properties. [d33c7bc]- The relation parsing logic in the
MetadataFactoryhas been refactored to be direction-aware. It now correctly generates one single-entity property for the "many-to-one" side (e.g.,Manager) and one collection property for the "one-to-many" side (e.g.,Subordinates) with the appropriateIImmutableRelation<T>type. - This fixes crashes in both the
datalinq create-modelscommand and the source generator when encountering this common database pattern.
- The relation parsing logic in the
- Corrected
requiredMembers for Primary Keys that are also Foreign Keys: Fixed a bug in the source generator where a column that was both a[PrimaryKey]and a[ForeignKey]was incorrectly omitted from the required members' constructor in mutable classes. The generator now correctly identifies these properties as required. [dda5e8e]
🛠️ Code Generation & Developer Experience Improvements
- Resolved All Known Nullability Warnings in Generated Code:
- Fixed
CS8603('Possible null reference return') for non-nullable relation properties (e.g.,public override Employee employees) by correctly applying the null-forgiving operator (!) only when nullable reference types are enabled and the property is non-nullable. [8efe6b1] - Fixed
CS8618warnings in the getters ofrequiredproperties by using the null-forgiving operator on theGetValue(...)cast, assuring the compiler that the value will not be null after construction. [00af93d] - Suppressed
CS8618warnings ('Non-nullable property must contain a non-null value') in mutable constructors that take an immutable object, acknowledging that the base constructor correctly initializes all required members. [8efe6b1]
- Fixed
- Resolved Member Hiding Warnings: Added the
newkeyword to generated properties (likeIsDeleted) that intentionally hide methods from theMutable<T>base class, resolvingCS0108compiler warnings. [8efe6b1] - Improved Nullability Metadata Parsing: The logic for parsing relation properties from source files (
SyntaxParser) now correctly detects and stores whether a relation is declared as nullable (e.g.,public abstract Employee? Manager). [8efe6b1]
Full Changelog: 0.6.0...0.6.1
DataLinq v0.6.0 - Powerful Queries, Dedicated Backends
This is a major feature release that significantly enhances the power and expressiveness of the LINQ provider, introduces dedicated first-class support for both MariaDB and MySQL, improves the developer experience with more convenient data access methods, and includes a deep refactoring of the query generation engine for greater stability and extensibility.
🚀 Highlights
- Dedicated MariaDB & MySQL Providers: The previously unified MySQL provider has been split into two distinct, first-class providers. This major architectural change ensures more accurate, dialect-specific SQL generation and enables dedicated support for features unique to each database, like MariaDB's native
UUIDtype. - Massively Expanded LINQ
WHEREClause Support: The LINQ provider is now dramatically more powerful, with support for many common, real-world query patterns:- Member-to-Member Comparisons: You can now write queries that compare two columns directly (e.g.,
where x.ShippedDate > x.OrderDate). - Full Date & Time Property Support: You can now use all properties of
DateTime,DateOnly, andTimeOnlyin your queries (e.g.,where x.CreatedAt.Year == 2025orwhere x.LoginTime.Hour < 9). - String Function Support: Added support for common string functions like
.ToUpper(),.ToLower(),.Trim(), andstring.IsNullOrEmpty().
- Member-to-Member Comparisons: You can now write queries that compare two columns directly (e.g.,
- Major LINQ Provider Refactoring: The core query translation logic has been completely refactored from a monolithic
WhereVisitorinto a new, cleanerQueryBuilderclass. This makes the code more robust, easier to maintain, and significantly simplifies the process of adding new LINQ features in the future. - Source-Generated Static
GetMethods: Models now have a source-generated staticGet()method, allowing for a much cleaner and more discoverable way to fetch single entities by their primary key (e.g.,Employee.Get(123, transaction)).
✨ Features & Enhancements
- Backend Improvements:
- Dedicated MariaDB & MySQL Support: The single MySQL provider has been split to provide dedicated, robust support for both databases. This resolves previous inconsistencies and allows for better, dialect-specific feature implementation and testing.
- MariaDB Native UUID Support: The new MariaDB provider correctly parses and generates the native
UUIDdata type for MariaDB 10.7+.
- New Data Access Methods:
- A static, source-generated
Get(primaryKey, IDataSourceAccess)method is now available on all table models for direct and efficient entity retrieval. Transaction<T>now has aGet<M>(primaryKey)method for easily fetching entities within a transaction's scope.
- A static, source-generated
- Expanded LINQ Functionality:
- Date/Time Properties: Full support for
.Year,.Month,.Day,.DayOfYear,.DayOfWeek,.Hour,.Minute,.Second, and.MillisecondinWHEREclauses. - String Functions: Support for
.ToUpper(),.ToLower(),.Trim(),.Substring(),string.IsNullOrEmpty(), andstring.IsNullOrWhiteSpace()inWHEREclauses.
- Date/Time Properties: Full support for
- Corrected Nullability Logic:
- Nullable Booleans: The logic for handling nullable boolean comparisons (
x.IsDeleted != true) has been completely fixed to correctly includeNULLvalues, matching C# semantics. - Default Values: Fixed a regression where properties with a
[DefaultValue]attribute were incorrectly generated as non-nullable. They are now correctly nullable.
- Nullable Booleans: The logic for handling nullable boolean comparisons (
- Default UUID Values: Added support for
[DefaultNewUUID]attribute for models, which translates toUUID()in generated SQL for MySQL/MariaDB.
🛠️ Refactoring & Architectural Changes
WhereVisitortoQueryBuilder: The complex logic for parsing LINQWHEREclauses has been moved from theWhereVisitorinto a new, dedicatedQueryBuilderclass. This separates the concerns of expression tree traversal from SQL construction, improving code clarity and stability.- Instance Creation Refactoring: The internal logic for creating model instances now relies on
IRowDataandIDataSourceAccessinterfaces, improving testability and architectural consistency.
🐛 Bug Fixes
- Fixed a bug where relations for newly created models were not being added correctly during metadata transformation.
- Fixed incorrect casing for property names when using the
CapitaliseNamesoption, especially for relation properties. - Resolved an issue where
Transaction<T>.Commit()returned a non-genericTransaction, which has been corrected to returnTransaction<T>.
Full Changelog: 0.5.4...0.6.0
DataLinq v0.5.4 - Critical Memory Leak Fixes & Schema Generation Improvements
This is a high-priority maintenance and enhancement release focused on resolving critical memory leaks, improving the fidelity of the schema-to-model generation process, and giving developers more control over their code generation workflow.
🚀 Highlights
- Fixed Two Critical Memory Leaks: Identified and resolved two separate memory leaks related to cache clearing and event handling, drastically improving long-term stability and performance for applications under load.
- Greatly Improved Database-to-C# Type Mapping: The
create-modelsCLI tool is now much smarter, correctly generating unsigned types (uint,ulong,byte, etc.) and appropriately sized integer types (short,long) from MySQL/MariaDB schemas. - New
--overwrite-typesCLI Flag: Added a new powerful option to thecreate-modelscommand that allows developers to force the regeneration of C# property types directly from the database schema, perfect for a "schema-first" workflow.
🐛 Critical Bug Fixes
- Fixed Major Memory Leak in Relation Caching:
- The previous event handling system for cache notifications (
WeakEventManager) had a subtle flaw that preventedImmutableRelationandImmutableForeignKeyobjects from being garbage collected. This could lead to significant memory consumption over time. - Solution: The
WeakEventManagerhas been completely replaced with a new, highly performant, lock-freeCacheNotificationManager. This new system uses a custom array-swapping pattern withInterlocked.CompareExchangeto ensure thread-safety and memory-safety without the overhead of reflection or the risks of finalizers, completely curing the leak.
- The previous event handling system for cache notifications (
- Fixed Memory Leak in Index Cache:
- A bug was discovered where calling
TableCache.ClearCache()would not fully clear theIndexCache. The reverse mapping dictionary (primaryKeysToForeignKeys) was not being cleared, causing it to grow indefinitely. This has been fixed.
- A bug was discovered where calling
- Corrected Nullability for Generated Properties:
- Auto-incrementing primary key columns are now correctly generated with nullable C# types (e.g.,
int?) to reflect theirnullstate before an entity is inserted into the database. - Columns with a
DEFAULTvalue in the database are also now correctly generated as nullable, as the value is optional on insert.
- Auto-incrementing primary key columns are now correctly generated with nullable C# types (e.g.,
✨ Features & Enhancements
- Enhanced Schema-to-Model Type Mapping:
- The
MetadataFromMySqlFactorynow correctly maps database integer types to their corresponding C# types based on size andUNSIGNEDflags. This improves type safety and correctness when generating models from a MySQL/MariaDB database. The new mappings include:MySQL Type C# Type TINYINT UNSIGNEDbyteTINYINTsbyteSMALLINT UNSIGNEDushortSMALLINTshortINT UNSIGNEDuintINTintBIGINT UNSIGNEDulongBIGINTlong
- The
- New
--overwrite-typesCLI Flag:- The
datalinq create-modelscommand now accepts an--overwrite-typesflag. - Default Behavior: DataLinq preserves user-defined types in source code (like custom classes or
enums) even if the underlying database column is a primitive type. - New Behavior: When
--overwrite-typesis used, DataLinq will force C# property types to be updated based on the schema from the database. This is ideal for when you change a column type in the database (e.g., fromINTtoBIGINT) and want your C# model to automatically update. This override intelligently preserves user-definedenumtypes and other custom classes.
- The
- Improved CLI Filtering Logic:
- Unified the
TablesandViewsconfiguration lists indatalinq.jsoninto a single, more intuitiveIncludelist. If the list is empty or omitted, all tables and views are included. Otherwise, only the specified items are included.
- Unified the
🛠️ Internal Improvements & Testing
- Improved Test Isolation: Created new, dedicated test fixtures (
MySqlFilteringTestFixture,MySqlTypeMappingFixture) that create temporary databases. This ensures that tests for metadata parsing are fully isolated, faster, and more reliable. - Comprehensive Test Coverage: Added a full suite of unit tests for the new type mapping logic, CLI filtering behavior, and the
--overwrite-typesfeature to prevent future regressions.
Full Changelog: 0.5.3...0.5.4