Make pylibcudf.Table.columns() return a tuple

[madsbk]

Description

pylibcudf.Table.columns() now returns a tuple[Column, ...] instead of a list[Column], preventing callers from mutating the table's internal column collection through a borrowed reference.

This aligns with the convention already used in cuDF, where Frame._columns and Frame._column_names are exposed as tuples.

Additionally, this simplifies handling zero-column tables: #22765.

This PR also adds pylibcudf.Table.release(), which transfers ownership of the columns to the caller as a fresh mutable list[Column] and empties the table (mirroring libcudf's table::release()).

Breaking change

This is a breaking change to the public pylibcudf.Table.columns() return type. Pass-through into pylibcudf functions still works (those functions now accept any Sequence), but callers that mutate the returned value must either copy it (list(table.columns())) or use Table.release().

Also, pylibcudf.io.types.TableWithMetadata.columns delegates to Table.columns(), so it now returns a tuple[Column, ...] as well.

[wence-]

Looks good

[coderabbitai]

📝 Walkthrough

Walkthrough

pylibcudf.Table now stores columns as a tuple, returns a tuple from columns(), and adds release() to transfer ownership as a list while clearing internal state. Several pylibcudf APIs widen sequence inputs from list to Sequence/Iterable, and downstream cudf/cudf_polars call sites update to release() and tuple return types.

Changes

pylibcudf API and downstream adoption

Layer / File(s)	Summary
Table: tuple storage, columns() → tuple, new release() python/pylibcudf/pylibcudf/table.pxd, python/pylibcudf/pylibcudf/table.pyx, python/pylibcudf/pylibcudf/table.pyi	Table._columns is stored as a tuple; columns() returns tuple[Column, ...]; release() returns a list copy and clears _columns. Declarations and stubs are updated.
Column: children accept Iterable python/pylibcudf/pylibcudf/column.pyx, python/pylibcudf/pylibcudf/column.pyi	Column.__init__ and Column.from_rmm_buffer accept Iterable[Column] for children, materializing to list internally before validation.
concatenate, null_mask, transform: list → Sequence inputs python/pylibcudf/pylibcudf/concatenate.pxd, python/pylibcudf/pylibcudf/concatenate.pyx, python/pylibcudf/pylibcudf/concatenate.pyi, python/pylibcudf/pylibcudf/null_mask.pxd, python/pylibcudf/pylibcudf/null_mask.pyx, python/pylibcudf/pylibcudf/null_mask.pyi, python/pylibcudf/pylibcudf/transform.pxd, python/pylibcudf/pylibcudf/transform.pyx, python/pylibcudf/pylibcudf/transform.pyi	Primary sequence parameters for concatenate, bitmask_and, bitmask_or, and transform widen from list to Sequence across declarations, stubs, and docstrings.
TableWithMetadata.columns stub: list → tuple python/pylibcudf/pylibcudf/io/types.pyi, python/pylibcudf/pylibcudf/io/types.pyx	TableWithMetadata.columns returns tuple[Column, ...] in the stub and docstring.
cudf internal helpers: list → tuple return annotations python/cudf/cudf/core/_internals/copying.py, python/cudf/cudf/core/_internals/sorting.py, python/cudf/cudf/core/frame.py, python/cudf/cudf/core/indexed_frame.py, python/cudf_polars/cudf_polars/dsl/ir.py	Return annotations for gather, scatter, columns_split, sort_by_key, _drop_duplicates_columns, _drop_nulls_columns, split_with_dtypes, and Join._reorder_maps change from list to tuple; scatter returns () on empty input.
Callers adopt tbl.release() for parquet and string split python/cudf/cudf/io/parquet.py, python/cudf_polars/cudf_polars/dsl/ir.py, python/cudf_polars/cudf_polars/dsl/expressions/string.py, python/cudf_polars/cudf_polars/dsl/expressions/rolling.py	The cudf parquet low-memory path and cudf_polars parquet chunked scan switch from .tbl.columns() to .tbl.release(). StringFunction split handling uses release(), and GroupedWindow._build_groupby_requests uses gathered_cols.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• rapidsai/cudf#22905 — Both PRs touch python/cudf/cudf/core/_internals/copying.py’s scatter function, with changes to its type/annotation behavior.

Suggested labels

Python, improvement, breaking, cudf-polars, pylibcudf

Suggested reviewers

• vyasr
• mroeschke
• wence-
• Matt711

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 20.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: changing pylibcudf.Table.columns() to return a tuple.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description check	✅ Passed	The description matches the changeset, explaining the tuple return type, added release method, and related sequence-accepting updates.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[Matt711]

I also checked for uses throughout RAPIDS and didn't see any we needed to change outside cudf https://github.com/search?q=org%3Arapidsai+%22.columns%28%29%22&type=code&p=1

[vyasr]

/merge