Reason or Problem
Epic #2341 calls out that the highest release risk for the GeoTIFF surface is pixels matching while attrs, coords, or dims silently disagree across backends. Today, the eager (open_geotiff) and dask (read_geotiff_dask) entry points both promise stable behavior, but no single regression test asserts full raster equivalence — pixels + dims + coords + the seven release-attr keys — across the two paths on the same files.
Proposal
Add xrspatial/geotiff/tests/test_release_gate_eager_dask_parity_2341.py.
For each file in a representative-corpus list (integer dtype + nodata, float dtype + NaN nodata, MinIsWhite, masked_nodata-lifecycle), open it eagerly and via read_geotiff_dask, materialize the dask result, and assert all of:
.values bit-exact (NaN-aware).dims equal.coords element-wise equal under xr.testing.assert_identical-style strictnessattrs['transform'], attrs['crs'], attrs['crs_wkt'], attrs['nodata'], attrs['masked_nodata'], attrs['georef_status'], attrs['raster_type'] all equal
The assertions are inlined as a small helper inside the test module (no new shared helper file — keeps the PR self-contained; consolidation across the PR-2..PR-5 sibling tests can be a follow-up). Add a row to docs/source/reference/release_gate_geotiff.rst citing the new test file under "Local GeoTIFF read and write".
Stakeholders and Impacts
Touches: one new test file under xrspatial/geotiff/tests/, one row added to the release-gate doc page. No production code changes. Runtime: bounded to the existing corpus fixtures already used by test_backend_pixel_parity_matrix_1813.py.
Out of scope (sibling PRs of epic #2341):
- Windowed-read shifted-transform parity (PR 2).
- Overview / sidecar metadata survival (PR 3).
- Stable-codec round-trip (PR 4).
- Negative tests for ambiguous metadata (PR 5).
Drawbacks
Slight overlap with test_backend_full_parity_2211.py, which already checks attrs finalization parity. This PR's test adds the eager-vs-dask cross-backend axis on top, and pins it under a release_gate-themed name so the release checklist row points at one test, not three.
Alternatives
- Extend
test_backend_full_parity_2211.py instead of adding a new file. Rejected: the release-gate doc row needs a stable test path to cite; reusing an existing file couples the row to unrelated future edits.
- Make the assertions a shared helper module up front. Rejected: that requires sequential ordering across the five sibling PRs; inlining keeps PRs 1-5 independent.
Additional Notes or Context
Parent epic: #2341. Sibling issues to be created for PRs 2-5.
Reason or Problem
Epic #2341 calls out that the highest release risk for the GeoTIFF surface is pixels matching while attrs, coords, or dims silently disagree across backends. Today, the eager (
open_geotiff) and dask (read_geotiff_dask) entry points both promise stable behavior, but no single regression test asserts full raster equivalence — pixels + dims + coords + the seven release-attr keys — across the two paths on the same files.Proposal
Add
xrspatial/geotiff/tests/test_release_gate_eager_dask_parity_2341.py.For each file in a representative-corpus list (integer dtype + nodata, float dtype + NaN nodata, MinIsWhite, masked_nodata-lifecycle), open it eagerly and via
read_geotiff_dask, materialize the dask result, and assert all of:.valuesbit-exact (NaN-aware).dimsequal.coordselement-wise equal underxr.testing.assert_identical-style strictnessattrs['transform'],attrs['crs'],attrs['crs_wkt'],attrs['nodata'],attrs['masked_nodata'],attrs['georef_status'],attrs['raster_type']all equalThe assertions are inlined as a small helper inside the test module (no new shared helper file — keeps the PR self-contained; consolidation across the PR-2..PR-5 sibling tests can be a follow-up). Add a row to
docs/source/reference/release_gate_geotiff.rstciting the new test file under "Local GeoTIFF read and write".Stakeholders and Impacts
Touches: one new test file under
xrspatial/geotiff/tests/, one row added to the release-gate doc page. No production code changes. Runtime: bounded to the existing corpus fixtures already used bytest_backend_pixel_parity_matrix_1813.py.Out of scope (sibling PRs of epic #2341):
Drawbacks
Slight overlap with
test_backend_full_parity_2211.py, which already checks attrs finalization parity. This PR's test adds the eager-vs-dask cross-backend axis on top, and pins it under arelease_gate-themed name so the release checklist row points at one test, not three.Alternatives
test_backend_full_parity_2211.pyinstead of adding a new file. Rejected: the release-gate doc row needs a stable test path to cite; reusing an existing file couples the row to unrelated future edits.Additional Notes or Context
Parent epic: #2341. Sibling issues to be created for PRs 2-5.