Skip to content

pandas: handle non-contiguous memory#42

Merged
adsharma merged 2 commits into
mainfrom
use_arrow_scan_for_pandas
Jul 11, 2026
Merged

pandas: handle non-contiguous memory#42
adsharma merged 2 commits into
mainfrom
use_arrow_scan_for_pandas

Conversation

@adsharma

@adsharma adsharma commented Jul 11, 2026

Copy link
Copy Markdown
Contributor

adsharma added 2 commits July 11, 2026 12:35
Use the numpy backend to ensure all numpy arrays extracted from
pandas columns are C-contiguous before scanning. Non-contiguous arrays
can arise from DataFrames backed by views into larger 2D arrays (e.g.
pd.DataFrame(arr2d), transpose, concat, slicing).  Previously the scan
used memcpy with sizeof(T) stride which reads wrong data on
non-contiguous layouts.
- Honour GEN env var (default to Ninja)
- Wipe stale build directory on generator mismatch
- Only clean build when BUILD_CLEAN=1
- Use ccache when available
- Honour CMAKE_BUILD_PARALLEL_LEVEL / PARALLEL for parallelism
- Handle artifact location flexibility
@adsharma adsharma force-pushed the use_arrow_scan_for_pandas branch from 6f46106 to 2dde501 Compare July 11, 2026 19:36
@adsharma adsharma changed the title pandas: use arrow scan regardless of types pandas: handle non-contiguous memory Jul 11, 2026
@adsharma adsharma merged commit 0d0d18e into main Jul 11, 2026
2 checks passed
@adsharma adsharma deleted the use_arrow_scan_for_pandas branch July 11, 2026 20:03
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Bug: Linux Ladybug 0.18 COPY FROM pandas DataFrame turns valid string primary keys into empty values

1 participant