Knut Olav Løite

This commit delivers a **bug fix** for the **`django-google-spanner`** library, resolving an issue where the `db_returning` attribute of `AutoFields` was being incorrectly overridden globally. It modifies the `autofield_init` function in `django_spanner/__init__.py` to **conditionally set `db_returning` and `validators`** only when Spanner's random ID generation is active. This ensures that Spanner-specific `AutoField` behavior is correctly applied without inadvertently affecting other database backends. The change significantly improves the **robustness and compatibility** of the Spanner integration with Django's ORM, with updated unit tests verifying the conditional logic.

This commit introduces a **new feature** to the **Spanner client library**, specifically enabling the direct **scanning of string values into `NullUUID`** fields. It also **enhances the robustness of `Scan` methods** for various `Null` types by ensuring they correctly handle `nil` pointer inputs, preventing potential panics during data unmarshaling. This **improvement** primarily affects the **data marshaling and unmarshaling logic** within the `spanner` package, making data retrieval more flexible and resilient for users working with nullable UUIDs and other nullable types. Extensive new tests were added to validate both the `NullUUID` string scanning and the improved `nil` handling for `Null` types.

This commit **fixes failing SQLAlchemy 2.0 conformance tests** for the `sqlalchemy-spanner` package by resolving a dependency conflict with OpenTelemetry. It **disables OpenTelemetry tracing within the test environment**, specifically for the `compliance_test_20` session, to prevent deprecation warnings that caused test failures. This **maintenance** work involves removing OpenTelemetry dependency installation from `noxfile.py` and **monkeypatching `trace_call`** in `conftest.py` to bypass tracing. The `_opentelemetry_tracing.py` module is also refactored for robustness, ensuring the **test suite passes cleanly with SQLAlchemy 2.0**.

This commit introduces a **new capability** for the **`spanner` package** by implementing the `sql.Scanner` and `driver.Valuer` interfaces for the `PGNumeric` type. This enhancement allows `PGNumeric` values to be seamlessly read from and written to a database using the standard **`database/sql` driver**, significantly improving its usability. The core change is in `spanner/value.go`, where the `Scan` and `Value` methods are defined for `PGNumeric`. Corresponding tests in `spanner/value_test.go` validate the correct behavior of the `PGNumeric.Scan` method. This makes handling Spanner's numeric data types more straightforward for developers interacting with the database via Go's standard library.

This commit provides a **bug fix** for the **Spanner** client library, specifically addressing an issue with decoding **PostgreSQL JSONB arrays**. Previously, arrays of PostgreSQL JSONB values failed to correctly decode into a slice of `PGJsonB` structs due to a faulty check of the array element's type code annotation. The fix corrects the variable used for this type annotation check within the `decodeValue` function in `spanner/value.go`. This ensures accurate deserialization of complex PostgreSQL JSONB data types, improving data integrity and reliability for applications using the Spanner client, with new tests in `spanner/value_test.go` confirming the correct behavior.

This commit introduces **GormDataType methods** for the `PGNumeric` and `PGJsonB` types within the **Spanner client library**. This **feature enhancement** ensures that **GORM** correctly interprets these PostgreSQL-specific types, preventing them from being treated as models. By defining their default data types via the `GormDataType` method, it guarantees that columns are automatically created with the appropriate type when these structs are used as fields in GORM models. This is a critical **compatibility improvement** for the `go-gorm-spanner` adapter, enabling proper schema migration and data handling for PostgreSQL types with GORM and Cloud Spanner.

This commit introduces a **performance optimization** to the **Spanner client library's session pool management**. It modifies the `spanner/client.go` logic to **prevent the creation of default regular sessions** when multiplexed sessions are fully enabled and no custom minimum session count (`MinOpened`) is explicitly set. Consequently, the **session pool will now default to zero sessions** under these conditions, reducing unnecessary resource allocation and maintenance overhead. This change primarily affects the **initialization and behavior of the Spanner session pool**, leading to more efficient resource utilization when multiplexing is active.

This commit introduces a **bug fix** for the **Spanner client library**, specifically by **disabling configuration logging** at startup by default. This change modifies the `enableLogClientOptions` function within `spanner/client.go` to ensure that logging of client options is suppressed unless explicitly enabled. Users can now control this behavior by setting the `GOOGLE_CLOUD_SPANNER_ENABLE_LOG_CLIENT_OPTIONS` or `GOOGLE_CLOUD_SPANNER_DISABLE_LOG_CLIENT_OPTIONS` environment variables, providing a less verbose default experience.

This commit delivers a **bug fix** for the **Spanner client library**, addressing an issue where the `transaction_tag` was not consistently applied. It modifies the `beginTransaction` function within the **transaction management** subsystem to ensure the `transaction_tag` is correctly set on the `BeginTransactionRequest` when utilizing **multiplexed sessions**. This change is critical for maintaining accurate observability and tracing of Spanner operations, as it guarantees that transaction tags are propagated as expected across all session types. The fix is validated by updated test cases in `spanner/client_test.go` and `spanner/transaction_test.go`, which now verify proper tag propagation.

This commit **optimizes the Spanner integration test suite** by configuring the `TestIntegration_DML` function within `spanner/integration_test.go` to run **in parallel**. This change primarily affects the **Spanner module's testing infrastructure**, specifically improving the efficiency of its Data Manipulation Language (DML) integration tests. Additionally, it performs a **minor cleanup** by removing an unnecessary debug logging statement from the same test, leading to cleaner test output. The overall impact is **faster test execution** and **improved test maintainability** for the **Spanner component**.

This commit implements a **bug fix** and **error handling improvement** within the **Spanner client library**'s **transaction management**. It addresses a scenario where using a `RowIterator` after its parent transaction has committed or rolled back would previously cause a **panic**. Now, the `query` method in `spanner/transaction.go` will gracefully return a `FailedPrecondition` error, specifically `errTransactionNoLongerActive`, instead of crashing the application. This change significantly enhances the **stability and predictability** of the **Spanner client**, allowing for more robust error handling by users, and is verified by new test cases in `spanner/transaction_test.go`.

This commit **enhances test coverage** for the **`spanner`** client library by adding new test cases to `spanner/row_test.go`. Specifically, it expands `TestSelectAll` to verify the function's behavior with nested named and unnamed structs, and includes checks for proper panic handling. This **test enhancement** proactively ensures the stability and correctness of the `SelectAll()` function, particularly in anticipation of changes from PR #13316. By improving the robustness of these tests, the overall reliability of the `spanner` module's data retrieval capabilities is strengthened.

This commit **fixes a build error** within the **SQLAlchemy Spanner samples** by modifying the `TicketSale` model in `packages/sqlalchemy-spanner/samples/model.py`. It **excludes the `last_update_time` column from `THEN RETURN` clauses** by setting `spanner_exclude_from_returning=True`. This **maintenance update** is necessary because columns assigned `PENDING_COMMIT_TIMESTAMP()` values cannot be included in `THEN RETURN` clauses, a constraint now correctly enforced by a recent Spanner Emulator fix. The change ensures the samples accurately reflect Spanner's behavior, preventing build failures and providing correct usage examples for the **SQLAlchemy Spanner integration**.

This commit **updates the default Python version to 3.12** across the project's build and testing infrastructure. It involves **build system maintenance** and **testing framework refactoring**, specifically updating the **CI/CD workflows** in `.github/workflows/test_suite.yml` to use Python 3.12 and removing deprecated test jobs. The `noxfile.py` is also updated to reflect Python 3.12 as the default, removing outdated test sessions like `compliance_test_13` and `migration_test_1310`, and adjusting SQLAlchemy version requirements for `migration_test`. Additionally, a **minor bug fix** is included in `migration_test_cleanup.py` to add `IF EXISTS` to `DROP TABLE` statements, improving the robustness of test cleanup by preventing errors if tables do not exist.

This commit provides a **bug fix** for the **Spanner client library**, specifically addressing an issue with transaction rollback reliability. It modifies the `ReadWriteTransaction` mechanism in `spanner/transaction.go` to use a **fresh background context** for `Rollback` calls. This ensures that rollback operations are not skipped or cancelled due to an expired original context, preventing **Spanner transactions** from leaving locks in place and improving resource cleanup. A new test case in `spanner/transaction_test.go` validates this critical change.

This commit **introduces a new sample and comprehensive tests** to demonstrate and verify the usage of Spanner's `PARSE_JSON` function within **SQLAlchemy applications**. A new sample file, `packages/sqlalchemy-spanner/samples/parse_json_sample.py`, provides a clear example of integrating `PARSE_JSON`, accessible via a new Nox session in `packages/sqlalchemy-spanner/samples/noxfile.py`. Additionally, **new test cases** in `packages/sqlalchemy-spanner/test/mockserver_tests/test_json.py` validate the function's behavior, including `wide_number_mode` and improved JSON insertion assertions. This **enhancement** significantly improves **developer guidance** and ensures the robust functionality of JSON parsing within the **SQLAlchemy Spanner dialect**, addressing a previously identified issue.

This commit introduces a **new capability** to the **Spanner client library**'s **transaction handling**, specifically for **statement-based transactions**. It enables the **delayed application of transaction options** by allowing a callback to be registered with `ReadWriteTransaction` via `NewReadWriteStmtBasedTransactionWithCallbackForOptions`. This callback provides updated transaction options when the transaction is truly initiated, such as on the first DML statement, aligning the `database/sql` driver's behavior with common SQL patterns. This **feature enhancement** improves flexibility and compatibility for users interacting with Spanner, allowing for more robust transaction initiation patterns.

This commit performs a **maintenance chore** by updating the **blunderbuss configuration** for the `sqlalchemy-spanner` package. It modifies the `.github/blunderbuss.yml` file to change the default issue assignee for automated issue management. Specifically, new issues will now be automatically assigned to 'olavloite' instead of 'harshachinta'. This change streamlines the project's automated workflow by updating the responsible party for issue triage.

This commit introduces a **bug fix** to the **Google Cloud Spanner client library** to correctly handle pre-commit tokens returned by the **Spanner Emulator**. Previously, the emulator's empty pre-commit tokens were misinterpreted, preventing necessary transaction retries and causing test failures due to lingering transactions. The **transaction and snapshot retry logic** in `google.cloud.spanner_v1/` is updated to properly account for this emulator behavior, specifically in `snapshot.py` and `transaction.py`. Additionally, the **CI/CD workflow for emulator integration tests** is updated to unpin the emulator version, enabling proper testing of multiplexed sessions and improving overall test reliability.

This commit introduces a **new capability** to the `google-cloud-spanner` DB-API connector, allowing users to **configure a custom logger** for database connections. The `connect` function in `google.cloud.spanner_dbapi.connection` now accepts a `logger` parameter via keyword arguments, which is then used by the underlying database instance. Additionally, this change includes **maintenance** to significantly reduce log verbosity in test environments. A default WARN-level logger is now employed for **mock server tests** to prevent excessive "Created multiplexed session." messages, and other test log spam is addressed, improving test readability.