geotiff: forward max_pixels, window, band on GPU stripped fallback (#1732)#1738
Merged
Conversation
Contributor
Author
|
@copilot resolve the merge conflicts in this pull request |
Contributor
Resolved in df44959. The only conflict was in |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR fixes read_geotiff_gpu behavior for stripped (non-tiled) TIFFs so its CPU-fallback path honors the same max_pixels, window, and band arguments as the tiled GPU path, and adds regression tests for issue #1732.
Changes:
- Forward
max_pixels,window, andbandtoread_to_array()in the stripped-TIFF GPU fallback. - Replace post-decode
_gpu_apply_window_bandslicing in the stripped fallback with coord-only computation (since the array is now pre-sliced byread_to_array). - Add new GPU regression tests covering
max_pixelsenforcement, windowed reads, band selection, and window+band composition.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
xrspatial/geotiff/__init__.py |
Forward missing kwargs to read_to_array in stripped fallback and adjust coordinate generation accordingly. |
xrspatial/geotiff/tests/test_gpu_stripped_forwarding_1732.py |
New regression tests ensuring stripped fallback honors max_pixels, window, and band. |
Comments suppressed due to low confidence (2)
xrspatial/geotiff/init.py:2531
- The comment says
read_to_arrayonly updatesgeo_info.transformfor orientations 5–8, butread_to_arraycurrently updates the transform for orientations 2–4 as well (whengeo_info.has_georefis true). This is misleading for future maintainers; please update or remove the outdated note (and the reference to a sibling PR) so it matches the actual behavior.
through to the next stage (CPU mmap re-decode for the first
failure, full CPU decode + GPU transfer for the second). This
preserves backward-compatible behaviour while making GPU
regressions visible.
- ``'strict'``: re-raise the original exception from either stage
xrspatial/geotiff/init.py:2578
- The windowed coord computation here checks
t is None, butGeoInfo.transformis always aGeoTransform(even whenhas_georef=False), so this branch is effectively dead and non-georeferenced windowed reads will incorrectly compute float coords from the default identity transform. Consider keying offgeo_info.has_georef(like_geo_to_coords) and returning integer pixel coords for non-georeferenced files, plus dropping the unreachablet is Nonebranch.
on_gpu_failure = gpu
elif not new_passed:
on_gpu_failure = 'auto'
gpu = on_gpu_failure
if gpu not in ('auto', 'strict'):
raise ValueError(
f"on_gpu_failure must be 'auto' or 'strict', got {gpu!r}")
try:
import cupy
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| win = (8, 16, 40, 80) # 32x64 window | ||
| out = read_geotiff_gpu(p, window=win) | ||
| assert out.shape == (32, 64) | ||
| np.testing.assert_array_equal(out.data.get(), data[8:40, 16:80]) |
Contributor
Author
There was a problem hiding this comment.
Addressed in f087212: added explicit assertions on out.coords['y'], out.coords['x'], and out.attrs['transform'] so regressions in the coord-only path are caught.
…1732) The stripped-TIFF branch of read_geotiff_gpu called read_to_array(source, overview_level=overview_level), dropping the caller's max_pixels, window, and band kwargs. Three consequences: - max_pixels safety cap bypassed (default 1B pixel limit applied instead of the smaller value the caller set). - Windowed reads decoded the entire image before slicing on the GPU. - Single-band reads on multi-band sources decoded every band. Forward all three. The post-decode _gpu_apply_window_band call is replaced with coord-only computation since read_to_array now produces the pre-sliced array. Adds regression coverage in test_gpu_stripped_forwarding_1732.py.
The windowed test only checked array values and shape. Since this PR swapped the post-decode `_gpu_apply_window_band` call for an inline coord-only computation, a regression in that new branch would not show up. Give the source fixture explicit coords so the file has a real georef, then assert `out.coords['y']`, `out.coords['x']`, and `out.attrs['transform']` against both the CPU eager path and the expected windowed tuple. The CPU comparison pins backend parity; the literal tuple catches drift on either side. Per Copilot review comment 3229433881.
4 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
read_geotiff_gpu's stripped-TIFF fallback atxrspatial/geotiff/__init__.py:2533calledread_to_array(source, overview_level=overview_level)and threw awaymax_pixels,window, andband.read_to_array. The post-decode_gpu_apply_window_bandis replaced with coord-only computation since the array is now pre-sliced.Closes #1732.
Test plan
test_gpu_stripped_forwarding_1732.pycover themax_pixelscap, windowed reads, band selection, and the window+band composition.test_gpu_stripped_multiband.py,test_orientation_gpu.py, predictor + nodata coverage).