cudf-polars: preserve zero-column row counts by madsbk · Pull Request #23234 · NVIDIA/cudf

madsbk · 2026-07-13T14:32:23Z

Description

Follow-up to #22765. Propagate the row count of a zero-column table through the cudf-polars DataFrame container and Projection node, so a frame with no columns and N rows preserves its shape as (N, 0) instead of collapsing to
(0, 0).

coderabbitai · 2026-07-14T12:33:07Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 417bfe91-7be8-439d-90a3-de5a6a92174a

📥 Commits

Reviewing files that changed from the base of the PR and between f90c293 and 4589c64.

📒 Files selected for processing (8)

python/cudf_polars/cudf_polars/containers/dataframe.py
python/cudf_polars/cudf_polars/dsl/ir.py
python/cudf_polars/cudf_polars/testing/inject_gpu_engine.py
python/cudf_polars/tests/containers/test_dataframe.py
python/cudf_polars/tests/streaming/test_select.py
python/cudf_polars/tests/test_dataframescan.py
python/cudf_polars/tests/test_groupby.py
python/cudf_polars/tests/test_scan.py

💤 Files with no reviewable changes (1)

python/cudf_polars/cudf_polars/testing/inject_gpu_engine.py

📝 Walkthrough

Summary by CodeRabbit

Bug Fixes
- Preserved row counts for dataframes with no columns across copying, selection, projection, sorting, scanning, and conversion operations.
- Ensured adding no columns to a zero-column dataframe retains its existing height.
- Improved consistency when transforming empty or zero-width dataframes.
Tests
- Expanded coverage for zero-column dataframe row-count preservation.
- Updated GPU and streaming test expectations for current behavior.

Walkthrough

The DataFrame container now stores zero-column row counts through plc.Table, propagates them across constructors and transformations, and preserves them during empty projections. Tests and engine expectation tables are updated for the resulting behavior changes.

Changes

Zero-column row-count handling

Layer / File(s)	Summary
DataFrame row-count propagation `python/cudf_polars/cudf_polars/containers/dataframe.py`, `python/cudf_polars/tests/containers/test_dataframe.py`	`DataFrame` passes row counts to `plc.Table` and propagates them through copying, construction, sorting, column operations, and zero-column tests.
Projection zero-column preservation `python/cudf_polars/cudf_polars/dsl/ir.py`, `python/cudf_polars/tests/streaming/test_select.py`, `python/cudf_polars/tests/test_dataframescan.py`, `python/cudf_polars/tests/test_groupby.py`, `python/cudf_polars/tests/test_scan.py`	Empty projections retain the input row count, and related tests remove conditional streaming-engine xfail handling.
Engine expectation updates `python/cudf_polars/cudf_polars/testing/inject_gpu_engine.py`	GPU and streaming expected-failure entries are revised for updated aggregation, array, window, join, IO, and zero-column behaviors.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Suggested reviewers: matt711, wence-, mroeschke

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% 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 clearly summarizes the main change: preserving row counts for zero-column cudf-polars frames.
Description check	✅ Passed	The description directly matches the changeset by explaining zero-column row-count preservation in DataFrame and Projection.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

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

mroeschke

Just a suggestion for a cleanup (can be a follow up) otherwise LGTM

mroeschke · 2026-07-14T21:27:58Z

            ),
            stream=stream,
-            num_rows=num_rows,
+            num_rows=table.num_rows() if num_rows is None else num_rows,


suggestion: Now that pylibcudf.Table accepts a num_rows argument, would be good to see if the we can remove num_rows from this function and have the caller always pass in a table with the intended row count

Good point, done: ab5995a

…om:madsbk/cudf into cudf-polars-zero-column-with-multiple-rows

…-column-with-multiple-rows

wence-

One question about the comment in with_columns

wence- · 2026-07-15T10:40:22Z

+        # When the result has columns, the row count is derived from them (and the
+        # non-broadcast path deliberately allows mismatched lengths, so we must not
+        # force a num_rows here).
+        return type(self)(
+            merged.values(),
+            stream=stream,
+            num_rows=self.num_rows if not merged else None,
+        )


I think it is invalid to pass new columns that have a different length from the existing table. So what is this comment saying?

Not always, updated the comment:

# Only pass num_rows for a zero-column result. For results with columns, it # must remain None because HStack(should_broadcast=False) intentionally # produces mismatched column lengths that its Select parent reconciles later. return type(self)( merged.values(), stream=stream, num_rows=self.num_rows if not merged else None, )

…-column-with-multiple-rows

madsbk · 2026-07-16T18:14:55Z

/merge

…ction DataFrame.from_table lost its num_rows parameter in NVIDIA#23234 (row counts are now inferred from the pylibcudf table, which carries them even for zero-column tables), but the duplicated-output path added in NVIDIA#23114 still passed num_rows=0, breaking mypy on every PR and raising TypeError at runtime on that path. empty_like already produces a 0-row table (including for zero-column inputs), so the argument was redundant.

…utput path (#23303) `DataFrame.from_table` lost its `num_rows` parameter in #23234 (row counts are now inferred from the pylibcudf table, which carries them even for zero-column tables), but the duplicated-output path added in #23114 still passes `num_rows=0`. The two PRs merged around the same time, so this surfaced only after both landed: mypy now fails on every PR's `check-style` job (`engine/core.py:840: Unexpected keyword argument "num_rows"`), and the path would raise `TypeError` at runtime. `plc.copying.empty_like` already produces a 0-row table (including for zero-column inputs, verified), so dropping the argument preserves the intended "freshly-allocated empty same-schema frame" semantics exactly. Authors: - GALI PREM SAGAR (https://github.com/galipremsagar) Approvers: - Matthew Roeschke (https://github.com/mroeschke) - Lawrence Mitchell (https://github.com/wence-) URL: #23303

…utput path (NVIDIA#23303) `DataFrame.from_table` lost its `num_rows` parameter in NVIDIA#23234 (row counts are now inferred from the pylibcudf table, which carries them even for zero-column tables), but the duplicated-output path added in NVIDIA#23114 still passes `num_rows=0`. The two PRs merged around the same time, so this surfaced only after both landed: mypy now fails on every PR's `check-style` job (`engine/core.py:840: Unexpected keyword argument "num_rows"`), and the path would raise `TypeError` at runtime. `plc.copying.empty_like` already produces a 0-row table (including for zero-column inputs, verified), so dropping the argument preserves the intended "freshly-allocated empty same-schema frame" semantics exactly. Authors: - GALI PREM SAGAR (https://github.com/galipremsagar) Approvers: - Matthew Roeschke (https://github.com/mroeschke) - Lawrence Mitchell (https://github.com/wence-) URL: NVIDIA#23303

madsbk self-assigned this Jul 13, 2026

madsbk added improvement Improvement / enhancement to an existing function non-breaking Non-breaking change labels Jul 13, 2026

github-actions Bot added Python Affects Python cuDF API. cudf-polars Issues specific to cudf-polars labels Jul 13, 2026

github-project-automation Bot added this to cuDF Python Jul 13, 2026

cudf-polars: preserve zero-column row counts

4589c64

madsbk force-pushed the cudf-polars-zero-column-with-multiple-rows branch from f51d8e7 to 4589c64 Compare July 13, 2026 14:40

GPUtester moved this to In Progress in cuDF Python Jul 13, 2026

cudf-polars: preserve zero-column row counts

8b6a497

madsbk marked this pull request as ready for review July 14, 2026 12:28

madsbk requested a review from a team as a code owner July 14, 2026 12:28

madsbk requested a review from mroeschke July 14, 2026 12:28

NVIDIA deleted a comment from copy-pr-bot Bot Jul 14, 2026

mroeschke approved these changes Jul 14, 2026

View reviewed changes

madsbk added 3 commits July 15, 2026 10:11

Merge branch 'cudf-polars-zero-column-with-multiple-rows' of github.c…

b67c449

…om:madsbk/cudf into cudf-polars-zero-column-with-multiple-rows

Merge branch 'main' of github.com:rapidsai/cudf into cudf-polars-zero…

958cb32

…-column-with-multiple-rows

num_rows cleanup

ab5995a

wence- approved these changes Jul 15, 2026

View reviewed changes

madsbk added 2 commits July 15, 2026 13:07

docs

4027205

Merge branch 'main' of github.com:rapidsai/cudf into cudf-polars-zero…

1faba5a

…-column-with-multiple-rows

wence- approved these changes Jul 15, 2026

View reviewed changes

Merge branch 'main' into cudf-polars-zero-column-with-multiple-rows

ca8bbe5

rapids-bot Bot merged commit 9d9fec8 into NVIDIA:main Jul 16, 2026
291 of 298 checks passed

github-project-automation Bot moved this from In Progress to Done in cuDF Python Jul 16, 2026

madsbk deleted the cudf-polars-zero-column-with-multiple-rows branch July 16, 2026 18:17

galipremsagar mentioned this pull request Jul 16, 2026

Fix stale num_rows argument breaking mypy in cudf-polars duplicated-output path #23303

Merged

3 tasks

TomAugspurger mentioned this pull request Jul 20, 2026

Avoid leaving multi-rank execution in a bad state in cudf-polars tests #23340

Merged

3 tasks

josephine-wolf-oberholtzer added this to cuDF Python Aug 12, 2026

josephine-wolf-oberholtzer moved this to Done in cuDF Python Aug 12, 2026

Navigation Menu

Uh oh!

cudf-polars: preserve zero-column row counts - #23234

cudf-polars: preserve zero-column row counts#23234
rapids-bot[bot] merged 8 commits into
NVIDIA:mainfrom
madsbk:cudf-polars-zero-column-with-multiple-rows

madsbk commented Jul 13, 2026

Uh oh!

coderabbitai Bot commented Jul 14, 2026

Summary by CodeRabbit

Walkthrough

Changes

❌ Failed checks (1 warning)

Uh oh!

mroeschke left a comment

Uh oh!

mroeschke Jul 14, 2026

Uh oh!

madsbk Jul 15, 2026

Uh oh!

wence- left a comment

Uh oh!

wence- Jul 15, 2026

Uh oh!

madsbk Jul 15, 2026

Uh oh!

madsbk commented Jul 16, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Uh oh!

Conversation

madsbk commented Jul 13, 2026

Description

Uh oh!

coderabbitai Bot commented Jul 14, 2026

Summary by CodeRabbit

Walkthrough

Changes

❌ Failed checks (1 warning)

Uh oh!

mroeschke left a comment

Choose a reason for hiding this comment

Uh oh!

mroeschke Jul 14, 2026

Choose a reason for hiding this comment

Uh oh!

madsbk Jul 15, 2026

Choose a reason for hiding this comment

Uh oh!

wence- left a comment

Choose a reason for hiding this comment

Uh oh!

wence- Jul 15, 2026

Choose a reason for hiding this comment

Uh oh!

madsbk Jul 15, 2026

Choose a reason for hiding this comment

Uh oh!

madsbk commented Jul 16, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants