Tim Sweña (Swast)

This commit performs a **documentation update** and **maintenance** task by consistently renaming all references from "Blob column" to "ObjectRef column" across the `bigframes` package. This change affects **internal comments** in modules like `bigframes.dataframe` and `bigframes.operations`, as well as **user-facing docstrings** in `bigframes.ml.llm` (e.g., `MultimodalEmbeddingGenerator` and `GeminiTextGenerator`) and `bigframes.session` (e.g., `from_glob_path` and `read_gbq_object_table`). The primary goal is to align the terminology with **official BigQuery documentation**, enhancing clarity and consistency for both developers and users. Relevant external links to the official documentation are now included in user-facing docstrings to provide further context.

This commit introduces a **new capability** by exposing the `DataFrame.bigquery` accessor, enabling a **unified API** for BigQuery-specific operations across both **BigFrames and pandas DataFrames**. It involves **refactoring** the `bigframes.extensions` module to establish core abstract base classes for `BigQuery` and `AI` accessors, which are then implemented by BigFrames-specific accessors and inherited by pandas accessors. This allows users to seamlessly access functionalities like `sql_scalar` and `ai.forecast` directly from `df.bigquery`, enhancing the consistency and usability of the **BigFrames library**. The change primarily impacts the **`bigframes.dataframe`** and **`bigframes.extensions`** subsystems, supported by updated documentation and new unit tests.

This commit **enhances the `bigframes.extensions.pandas` module** by **integrating API logging** for its accessor classes. It **refactors** the `AIAccessor` and `BigQueryDataFrameAccessor` by renaming them to `PandasAIAccessor` and `PandasBigQueryDataFrameAccessor` respectively, and then applies a logging decorator to both. This change ensures that interactions with these **Pandas accessors** are properly recorded in the API logs, improving observability and debugging capabilities. Additionally, the commit updates the module's `__init__.py` to correctly expose the renamed accessors and adjusts relevant documentation and unit tests.

This commit introduces a **new capability** by adding the `forecast()` method to the **`df.bigquery.ai` pandas DataFrame accessor**, enabling users to perform **time series forecasting** directly on BigFrames DataFrames. This method proxies to the underlying `bigframes.bigquery.ai.forecast` function, which was updated to support an `output_historical_time_series` parameter for controlling historical data output. The **BigFrames API** is enhanced, allowing data scientists to seamlessly integrate BigQuery AI forecasting into their workflows. Unit tests were also added to ensure the reliability of this new feature.

This commit performs **code cleanup** by **removing an unused Ibis Python to JavaScript UDF compiler** from the `bigframes` package. Specifically, it targets the **BigQuery backend**'s vendored Ibis code, deleting the `udf/` directory and references to `PythonToJavaScriptTranslator` and `_get_udf_source` from `third_party/bigframes_vendored/ibis/backends/bigquery/__init__.py`. This **maintenance** task streamlines the codebase by eliminating components that were not supported or utilized in production. The change improves overall code hygiene and reduces unnecessary complexity within the **Ibis integration** for BigQuery.

This commit performs **documentation maintenance** by significantly enhancing the **BigFrames documentation** for improved clarity and search engine optimization. It primarily **retouches the main documentation index page** (`docs/index.rst`) with updated content, an SEO meta description, and reorganized sections to boost discoverability. Additionally, module docstrings for `bigframes.bigquery.ai`, `bigframes.pandas`, and `bigframes.bigquery` are updated to provide more comprehensive overviews of their respective functionalities and APIs. This work ensures the project's documentation is more accessible and informative for users.

This commit **releases version `v2.38.0` of the `bigframes` library**, incorporating several **new features** and **bug fixes**. Key additions include the `bigframes.bigquery.rand()` function, a `df.bigquery` pandas accessor, `str` and `dt` accessors for `pd.col` Expression objects, and the `bigquery.ml.get_insights()` function. It also **resolves issues** such as handling unsupported types in `describe()` and preventing automatic `anywidget` usage in `%bqsql` magics. This **maintenance** update ensures users of the **BigQuery DataFrames** library receive the latest enhancements and improved stability.

This commit introduces a **new capability** by adding a **`df.bigquery` pandas accessor** to the `bigframes` library, specifically providing a `sql_scalar` method. This allows users to execute **BigQuery SQL functions directly on pandas DataFrames**, returning scalar results and enhancing the interoperability between pandas and BigQuery within the `bigframes` ecosystem. The work involves significant changes to the **`bigframes` core logic** for SQL operations, new **documentation** including a user guide and example notebook, and comprehensive **unit tests** to validate the accessor's functionality and registration. This feature simplifies the process of integrating BigQuery operations into pandas-based data analysis workflows.

This commit **fixes** an issue within the **BigQuery DataFrames** library where the `%bqsql` magic was automatically using `anywidget` for displaying query results. Specifically, the `_cell_magic` function in `bigframes/_magics.py` has been modified to remove this default integration. This **behavior change** ensures that dataframes returned by `%bqsql` queries will now revert to the standard display mechanism, preventing unintended rendering and resolving an internal issue. The impact is limited to how results from the `%bqsql` magic are presented to the user.

This commit introduces a **new capability** by adding the `bigframes.bigquery.rand()` function to the **BigFrames library**, fulfilling a customer request. This function wraps the BigQuery `RAND()` SQL function, enabling users to generate pseudo-random `FLOAT64` values within their BigFrames workflows. A significant **refactoring** was also performed on the core `SqlScalarOp` class to include an `is_deterministic` attribute, which is set to `False` for `rand()`, ensuring proper handling of non-deterministic operations. The new function is exposed in the `bigframes.bigquery` module and is thoroughly validated with both unit and system tests.

This commit provides a **documentation fix** by correcting a spelling error in the docstring for the `ExperimentOptions` class. Specifically, it updates the word 'configration' to 'configuration' within the `packages/bigframes/bigframes/_config/experiment_options.py` file. This **maintenance** task enhances the accuracy and readability of the **BigFrames configuration module's documentation**, particularly for users interacting with experimental options. There is no functional impact on the `bigframes` library's behavior, only an improvement to its descriptive text.

This commit prepares the **`pandas-gbq` library for its `v0.34.0` release**, initiated by the Librarian CLI. It's a **maintenance** task that performs a **version bump**, updating the internal version string in `pandas_gbq/version.py` and the Librarian state file, `packages/pandas-gbq/.librarian/state.yaml`. Merging this pull request will automatically trigger the release process for `pandas-gbq`, making new features like BigLake table support and compatibility with pandas 3.0 available to users.

This commit **introduces support for BigLake tables** within the `pandas_gbq.sample` function, significantly expanding its data source capabilities. It achieves this by adding new **core modules** such as `pandas_gbq.core.biglake` for metadata retrieval and `pandas_gbq.core.resource_references` for parsing table identifiers. The **`pandas_gbq.core.sample` module** is extensively enhanced to accommodate BigLake tables, involving refactored internal sampling methods and adjusted byte estimation logic. This **new capability** allows users to seamlessly sample data from both BigQuery and BigLake tables, broadening the utility of `pandas-gbq` for diverse data analytics workflows.

This commit provides a **dependency compatibility fix** for the **`pandas-gbq` library** to support the upcoming `pandas 3.0` prerelease. It updates system test data in `packages/pandas-gbq/tests/system/test_to_gbq.py` to use `pandas.NA` for missing values instead of `None`, reflecting changes in `pandas 3.0`. Additionally, a datetime value is adjusted to ensure proper compatibility with `pandas 3.0` and BigQuery's CSV loader. This **maintenance** work ensures the `to_gbq` functionality remains robust and accurate when interacting with BigQuery using the new pandas version.

This commit performs a **release preparation chore** for the **`bigquery-magics`** package, specifically for version `v0.12.1`. It updates the project version from `0.12.0` to `0.12.1` in the `.librarian/state.yaml` configuration file and the `__version__` string within `bigquery_magics/version.py`. This automated update, initiated by the Librarian CLI, enables the official release of `v0.12.1`, which incorporates a **bug fix** for a usability improvement to the graph visualizer.

This commit performs a **maintenance chore** by updating the Docker image reference for the `librarian` tool. Specifically, it modifies the `image` field within the `packages/bigquery-magics/.librarian/state.yaml` configuration file to a new SHA256 hash. This ensures that the **`bigquery-magics`** package's `librarian` integration utilizes the latest specified image, potentially incorporating upstream updates or fixes to the `librarian` tool itself without altering the package's core functionality.

This commit **improves documentation** for the **`sql_compiler` experimental option** within the `bigframes` package. It adds a comprehensive docstring and a practical code sample to the `bpd.options.experiments.sql_compiler` property, located in `packages/bigframes/bigframes/_config/experiment_options.py`. This enhancement provides clearer guidance and examples for users configuring the SQL compiler, making the experimental feature more accessible and understandable. The change is purely a **documentation update**, with no impact on runtime behavior or functionality.

This commit performs a **dependency update** for the **`bigframes` package**, specifically updating the reference to the `python-librarian-generator` Docker image. It modifies the `packages/bigframes/.librarian/state.yaml` file to point to a newer `sha256` hash for this image. This **tooling update** ensures that the `bigframes` development environment leverages the latest version of the librarian code generation or management tool. The change helps maintain consistency and incorporates any new features, bug fixes, or performance improvements from the updated generator into the `bigframes` build process.

This commit **prepares for the release** of **bigquery-magics** version `0.12.0` by updating the internal version configuration in `.librarian/state.yaml` and the `__version__` string in `bigquery_magics/version.py`. This **maintenance** task is part of an automated release process, which, upon merging, will trigger the official publication of the new `0.12.0` package. The new release introduces **features** like schema view support and the removal of the `bqsql` magic, alongside a **bug fix** addressing graph visualization conflicts within the `bigquery-magics` library.

This commit performs a **maintenance chore** by updating the reference to the `python-librarian-generator` Docker image. It specifically modifies the `sha256` hash within the `.librarian/state.yaml` configuration file for the **`bigquery-magics`** package. This update ensures that the `bigquery-magics` module leverages the **latest version of its code generation tool**, which may include new capabilities or critical fixes for its internal processes.