Fix datashader resolution collapse on cropped coordinate offsets#669
Merged
Conversation
After a ``bounding_box_query`` on a translated SpatialData, ``render_points(method="datashader")`` and ``render_shapes(method="datashader")`` silently collapsed the rasterization resolution. Three entangled defects in the datashader canvas pipeline made the data range ``min(0, world_min)`` to ``world_max`` instead of the actual element bbox, then masked the broken canvas extent with a ``Scale``-only rgba transform that happens to align at world 0. Atomic fix: - ``_datashader_canvas_from_dataframe`` and ``_get_extent_and_range_for_datashader_canvas`` use the real element extent (no ``min(0, ...)`` clamp). - ``_create_image_from_datashader_result`` accepts ``x_min``/``y_min`` and emits ``Sequence([Scale, Translation])`` so the rgba lands at the element's actual world position. - ``_render_ds_image`` and ``_render_ds_outlines`` thread ``x_min``/``y_min`` through from the caller. - ``_compute_datashader_canvas_params`` no longer merges with ``ax.get_xlim()`` (which was read pre-correction and returned intrinsic-pixel extent for an ``imshow``-rendered image). - The unused negative-only ``extent=`` shift branch in ``_ax_show_and_transform`` is removed. Also fix the related rasterize / empty-input bugs surfaced by the investigation: - ``_rasterize_if_necessary`` now computes ``target_unit_to_pixels`` against the world extent instead of intrinsic source dims; for Scale != 1 this changes pixel resolution by exactly the scale factor. - ``_datashader_canvas_from_dataframe`` short-circuits on empty input (NaN ``min()`` used to crash with ``cannot convert float NaN to integer``). - ``_render_points`` / ``_render_shapes`` skip the datashader pipeline when the element is empty. Tests added (all non-visual): - bbox_query points canvas factor matches no-offset baseline - bbox_query points end-to-end render does not crash - bbox_query shapes end-to-end render does not crash - empty-input canvas helper does not crash - rasterize target_unit_to_pixels uses world units under Scale=0.5
timtreis
changed the title
…ization Datashader shape outlines: pixel-level differences from the canvas extent fix; visually the outlines are essentially identical to the previous baselines (minor antialiasing only). raccoon_scale: sharper rasterization output now that _rasterize_if_necessary computes target_unit_to_pixels in world units instead of intrinsic pixels. Baselines pulled from CI run 26113431879.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #669 +/- ##
==========================================
- Coverage 77.61% 77.56% -0.06%
==========================================
Files 11 11
Lines 3610 3610
Branches 849 849
==========================================
- Hits 2802 2800 -2
- Misses 484 485 +1
- Partials 324 325 +1
🚀 New features to boost your workflow:
|
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
2 participants
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
Fixes #668. After a
bounding_box_queryon a translatedSpatialData,render_points(method=\"datashader\")andrender_shapes(method=\"datashader\")silently collapsed the rasterization resolution — points collapsed into a handful of right-edge pixels.Three entangled defects in the datashader pipeline forced the canvas to span
[0, world_max]instead of[world_min, world_max], then masked the broken canvas extent with aScale-only rgba transform that happens to align at world 0. Plus related resolution / empty-input issues surfaced by the investigation.What changed
Datashader canvas pipeline (Bugs 1–5 from #668):
_datashader_canvas_from_dataframe/_get_extent_and_range_for_datashader_canvasuse the real element extent (nomin(0, ...)clamp)._create_image_from_datashader_resultacceptsx_min/y_minand emitsSequence([Scale, Translation])so the rgba lands at the element's actual world position._render_ds_image/_render_ds_outlinesthreadx_min/y_minthrough._compute_datashader_canvas_paramsno longer merges withax.get_xlim()(which was read pre-correction and reflected intrinsic-pixel extent after animshowdraw).extent=shift branch in_ax_show_and_transformis removed.Rasterize / empty input (Bugs 6 + 8 from #668):
_rasterize_if_necessarynow computestarget_unit_to_pixelsagainst the world extent instead of intrinsic source dims; forScale != 1this changes the resulting pixel count by exactly the scale factor._datashader_canvas_from_dataframeshort-circuits on empty input (NaNmin()used to crash withcannot convert float NaN to integer)._render_points/_render_shapesskip the datashader pipeline when the element is empty.