uql-orm 0.1.3 → 0.2.0

package/CHANGELOG.md

@@ -1,24 +1,38 @@
-# Change Log
-
-All notable changes to this project will be documented in this file.
-See [Conventional Commits](https://conventionalcommits.org) for commit guidelines.
-
-## [0.1.3](https://github.com/rogerpadilla/uql/compare/uql-orm@0.1.2...uql-orm@0.1.3) (2026-03-08)
-
-
-### Performance Improvements
-
-* Reduce allocations and simplify logic in utility functions and SQL query generation. ([5bc1ed5](https://github.com/rogerpadilla/uql/commit/5bc1ed5d85b91ba4af96075a91c93c923a204e76))
+# Changelog
 
+All notable changes to this project will be documented in this file. Please add new changes to the top.
 
+date format is [yyyy-mm-dd]
 
+## [0.2.0] - 2026-03-08
+### New Features
+- **Transaction Isolation Levels**: `beginTransaction()` and `transaction()` now accept an optional `TransactionOptions` object with an `isolationLevel` property. Supports all standard SQL isolation levels: `read uncommitted`, `read committed`, `repeatable read`, and `serializable`.
+  - **PostgreSQL**: Uses inline syntax (`BEGIN TRANSACTION ISOLATION LEVEL ...`).
+  - **MySQL / MariaDB**: Uses the `SET TRANSACTION ISOLATION LEVEL` + `START TRANSACTION` two-statement pattern.
+  - **SQLite / LibSQL / MongoDB**: Isolation level is silently ignored (these databases do not support configurable isolation levels).
+  ```ts
+  await querier.beginTransaction({ isolationLevel: 'serializable' });
+  // or with the callback API
+  const result = await querier.transaction(async () => {
+    return querier.findMany(User, {});
+  }, { isolationLevel: 'read committed' });
+  ```
+- **Config-Driven Dialect Strategy**: Added `isolationLevelStrategy` to `DialectConfig` (`'inline'` | `'set-before'` | `'none'`), enabling declarative per-dialect SQL generation without dialect-name branching.
 
+## [0.1.5] - 2026-03-08
+### Type Safety
+- **Eliminated `any` Types**: Replaced `any` with proper types across decorators (`serialized.ts`, `log.ts`, `transactional.ts`), Express middleware (`querierMiddleware.ts`), MongoDB dialect pipeline types, SQLite querier pool, and migrator. Remaining `any` usages are documented and justified (generic variance, `Reflect.getMetadata`).
+- **Typed `raw()` Return**: `raw()` now returns `QueryRaw` instead of `any`, enabling IDE autocompletion and compile-time validation.
 
-# Changelog
+### Bug Fixes
+- **Fixed `IsolationLevel` Typo**: Corrected `'repeteable read'` → `'repeatable read'` in the `IsolationLevel` type.
 
-All notable changes to this project will be documented in this file. Please add new changes to the top.
+### Security
+- **`raw()` Safety Documentation**: Added JSDoc warning that `raw()` bypasses SQL parameterization, with guidance to use `$where` operators for user-supplied data.
 
-date format is [yyyy-mm-dd]
+## [0.1.4] - 2026-03-08
+### Bug Fixes
+- **Fixed Virtual Field Alias in Relations**: `getRawValue` was missing a dot separator in prefixed aliases and had a stale dot→underscore replacement from the old convention. Added tests to prevent regressions.
 
 ## [0.1.3] - 2026-03-08
 ### Code Quality
@@ -26,7 +40,7 @@ date format is [yyyy-mm-dd]
 
 ## [0.1.1] - 2026-03-08
 ### Bug Fixes
-- **Fixed Row Parsing for Underscore Columns**: Columns containing underscores (e.g., `user_id`) were incorrectly unflattened into nested objects (`{ user: { id: value } }`). SQL JOIN aliases now use quoted dot-notation (e.g., `` `profile.pk` `` instead of `` `profile_pk` ``), eliminating the ambiguity. Dot-delimited aliases are safe because they are always quoted identifiers.
+- **Fixed Row Parsing for Underscore Columns**: Columns containing underscores (e.g., `user_id`) were incorrectly unflattened into nested objects (`{ user: { id: value } }`). SQL JOIN aliases now use quoted dot-notation (e.g., `` `profile.pk` `` instead of `` `profile_pk` ``), eliminating the ambiguity. Dot-delimited aliases are safe because they are always quoted identifiers. Updated tests to prevent regressions.
 
 ### Performance
 - **Faster SQL Query Generation**: Optimized the internal SQL generation pipeline to reduce overhead on every query. Identifier escaping now reuses pre-compiled regex patterns instead of creating new ones per call. Relation detection short-circuits without intermediate array allocations. The query context tracks SQL length incrementally, avoiding repeated string joins. These changes reduce per-query CPU and memory cost, improving throughput for high-volume workloads.

package/dist/browser/uql-browser.min.js

@@ -774,7 +774,7 @@ class ItemAdjustment extends (_BaseEntity12 = BaseEntity) {
 }
 
 const holder = globalThis;
-const metaKey = 'uql-orm/entity/decorator';
+const metaKey = Symbol.for('uql-orm/entity/decorator');
 const metas = holder[metaKey] ?? new Map();
 holder[metaKey] = metas;
 function getEntities() {