Document and Migrate host_span vs span public APIs by joannamooon · Pull Request #22588 · NVIDIA/cudf

joannamooon · 2026-05-19T19:27:46Z

Description

Adds a "When to use host_span vs std::span" subsection to DEVELOPER_GUIDE.md
Migrates all paired cudf:: / cudf::detail:: public APIs that used host_span purely as a host buffer view to std::span (null mask, groupby, concatenate, interop, scalar, dictionaries, rolling, parquet, hybrid scan)
Updates the corresponding Cython bindings (span.pxd, hybrid_scan.pxd/.pyx) to expose std::span via a new std_span wrapper
Internal-only implementation files (compute_groupby, compute_single_pass_aggs, extract_single_pass_aggs, etc.) were updated as a necessary cascade to keep the build consistent after the public API signatures changed — not as an intentional scope expansion

Checklist

I am familiar with the Contributing Guidelines.
New or existing tests cover these changes.
The documentation is up to date with these changes.

Clarified when to prefer `std::span` for host-side data and when to use `cudf::host_span`, including specific scenarios for each. This enhances the documentation for developers regarding memory management in the library.

copy-pr-bot · 2026-05-19T19:27:49Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

coderabbitai · 2026-05-22T04:10:27Z

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

@coderabbitai resume to resume automatic reviews.
@coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

▶️ Resume reviews
🔍 Trigger review

📝 Walkthrough

Walkthrough

Converts many host-side non-owning parameters from cudf::host_span to std::span, adds <span> includes, updates call-site adapters, updates developer guide, and adjusts Cython bindings for hybrid-scan and Parquet IO.

Changes

std::span migration and doc update

Layer / File(s)	Summary
Span migration (headers, sources, bindings, docs) `cpp/include/`, `cpp/src/`, `cpp/doxygen/developer_guide/DEVELOPER_GUIDE.md`, `python/pylibcudf/*`	Replaces `cudf::host_span` with `std::span` across public/detail headers and C++ sources (groupby, copying, IO/parquet, hybrid-scan, rolling, null-mask, dictionary, scalar, roaring, etc.), adds `#include <span>` where required, adapts call sites by constructing `host_span` adapters when downstream still expects `host_span`, updates developer guide subsection about when to use `host_span` vs `std::span`, and updates Cython `.pxd/.pyx` to use `std_span` interop.

Estimated code review effort:
🎯 4 (Complex) | ⏱️ ~60 minutes
Possibly related PRs:
- rapidsai/cudf#22587 — Updates Python parquet readers and footer plumbing; intersects with this PR’s Parquet footer std::span changes.
Suggested reviewers:
- shrshi
- mroeschke
- wence-

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

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

joannamooon · 2026-06-01T17:28:54Z

rapidsmpf links against libcudf and is pulled into CI prebuilt from nightlies. It's being compiled against the old host_span signatures.

I created a new PR in rapidsmpf: rapidsai/rapidsmpf#1077
This PR update cudf call sites.

After merging rapidsmpf#1077, the next published rapidsmpf nightly should clear the failures in the current ci.

…td-span

PointKernel

Looks good and thank you @joannamooon for updating rapidsmpf accordingly.

PointKernel · 2026-06-02T17:19:55Z

Those cuIO changes look legitimate to me. Just FYI, @mhaseeb123 is also working in the same files

PointKernel · 2026-06-02T18:24:18Z

/ok to test 0c4a9d4

vyasr · 2026-06-03T01:09:19Z

@joannamooon I'm in the middle of a large refactor that will move all the parts of rapidsmpf that call libcudf into a new library in this repository. All rapidsmpf development is temporarily frozen as a result of that. Once that is done, though, you should have a much easier time getting unblocked here.

…td-span # Conflicts: # cpp/src/io/parquet/experimental/hybrid_scan_impl.cpp # cpp/src/io/parquet/experimental/hybrid_scan_impl.hpp # cpp/src/io/parquet/io_utils/parquet_io_utils.cpp # cpp/src/rolling/grouped_rolling.cu

vyasr · 2026-06-16T20:44:45Z

/ok to test 40581d5

…std::span

vyasr · 2026-06-23T23:59:32Z

/ok to test 210fb82

vyasr · 2026-06-24T02:00:05Z

/ok to test 058fb95

vyasr · 2026-06-24T02:31:37Z

/ok to test a10bd5c

PointKernel · 2026-06-27T00:12:35Z

/ok to test c1834b7

PointKernel · 2026-06-30T16:06:07Z

/ok to test b3949e9

PointKernel · 2026-06-30T18:26:16Z

@joannamooon We don't need to keep the branch continuously updated with the latest upstream main. I'll relaunch the CI, and if it passes, I'll merge it as-is, even if it's a few commits behind main.

PointKernel · 2026-06-30T18:26:31Z

/ok to test c9a88c0

PointKernel · 2026-06-30T20:28:09Z

/merge

joannamooon added 2 commits May 19, 2026 13:00

Refactor null mask functions to use std::span instead of host_span

6dbda88

github-actions Bot assigned joannamooon May 19, 2026

github-actions Bot added the libcudf Affects libcudf (C++/CUDA) code. label May 19, 2026

joannamooon and others added 10 commits May 19, 2026 15:27

Merge branch 'main' into docs/host-span-vs-std-span

7873397

migrate to span copy and null_mask

d071374

migrate concatenate

e1c3b0d

migrate groupby

e51d8a8

migrate interop

c294e31

migrate scalar

ff8ab76

migrate dictionaries

d0071b3

migrate rolling, bitmap

5ee1bbb

migrate parquet

79b376e

migrate hybrid scan, migrate cython files

45656ce

github-actions Bot added Python Affects Python cuDF API. pylibcudf Issues specific to the pylibcudf package labels May 20, 2026

github-project-automation Bot added this to cuDF Python May 20, 2026

linter

59c58ac

GPUtester moved this to In Progress in cuDF Python May 20, 2026

joannamooon and others added 3 commits May 20, 2026 15:37

Merge branch 'main' into docs/host-span-vs-std-span

b8f2edd

fix compile errors

79f14fd

Merge branch 'rapidsai:main' into docs/host-span-vs-std-span

7f143d4

joannamooon marked this pull request as ready for review May 22, 2026 04:01

joannamooon requested review from a team as code owners May 22, 2026 04:01

joannamooon requested review from mroeschke, shrshi and wence- May 22, 2026 04:01

joannamooon mentioned this pull request Jun 1, 2026

Use std::span for cudf::split/cudf::concatenate calls rapidsai/rapidsmpf#1077

Open

Merge remote-tracking branch 'upstream/main' into docs/host-span-vs-s…

4f5840e

…td-span

PointKernel approved these changes Jun 2, 2026

View reviewed changes

Merge branch 'main' into docs/host-span-vs-std-span

0c4a9d4

github-actions Bot assigned vyasr Jun 16, 2026

migrate hybrid scan row_group_indices params from cudf::host_span to …

210fb82

…std::span

joannamooon and others added 2 commits June 23, 2026 21:12

fix template mismatch

3aa4fa3

Merge branch 'main' into docs/host-span-vs-std-span

058fb95

Fix copyright headers to match canonical format

a10bd5c

Merge branch 'main' into docs/host-span-vs-std-span

c1834b7

joannamooon and others added 2 commits June 30, 2026 10:01

fix template mismatch

a8bb9cc

Merge branch 'main' into docs/host-span-vs-std-span

b3949e9

Merge branch 'main' into docs/host-span-vs-std-span

c9a88c0

rapids-bot Bot merged commit aa8cfca into NVIDIA:main Jun 30, 2026
137 checks passed

github-project-automation Bot moved this from In Progress to Done in cuDF Python Jun 30, 2026

This was referenced Jul 1, 2026

Refactor host_span to use cuda::std::span internally and remove span_base #23072

Merged

Support explicit row counts for zero-column tables #22765

Merged

Uh oh!

Conversation

joannamooon commented May 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Checklist

Uh oh!

copy-pr-bot Bot commented May 19, 2026

Uh oh!

coderabbitai Bot commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Reviews paused

Walkthrough

Changes

Uh oh!

joannamooon commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

PointKernel left a comment

Choose a reason for hiding this comment

Uh oh!

PointKernel Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

PointKernel commented Jun 2, 2026

Uh oh!

vyasr commented Jun 3, 2026

Uh oh!

vyasr commented Jun 16, 2026

Uh oh!

vyasr commented Jun 23, 2026

Uh oh!

vyasr commented Jun 24, 2026

Uh oh!

vyasr commented Jun 24, 2026

Uh oh!

PointKernel commented Jun 27, 2026

Uh oh!

PointKernel commented Jun 30, 2026

Uh oh!

PointKernel commented Jun 30, 2026

Uh oh!

PointKernel commented Jun 30, 2026

Uh oh!

PointKernel commented Jun 30, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

joannamooon commented May 19, 2026 •

edited

Loading

coderabbitai Bot commented May 22, 2026 •

edited

Loading

joannamooon commented Jun 1, 2026 •

edited

Loading