james-willis opened a new issue, #3120:
URL: https://github.com/apache/sedona/issues/3120
## Expected behavior
`RS_AsRaster` with `allTouched = true`, and `LINESTRING` rasterization, aim
to match GDAL (`rasterio.features.rasterize(all_touched=True)`) on which pixels
are burned.
## Actual behavior
After the exact-traversal fix (#3118), the two engines agree on all general
input (a 100-case randomized fuzz over anisotropic grids matches GDAL exactly).
They can still differ by a single pixel when a geometry vertex lands *exactly*
on a grid line.
Example on a 6x6 unit grid:
```sql
SELECT RS_BandAsArray(RS_AsRaster(
ST_GeomFromWKT('POLYGON ((1.5 1.5, 3.8 3.0, 4.5 4.4, 3.4 3.5, 1.5 1.5))'),
RS_MakeEmptyRaster(1, 'B', 6, 6, 0.0, 6.0, 1.0, -1.0, 0.0, 0.0, 0),
'B', true, 1.0, 0.0, false), 1);
```
The vertex `(1.5 1.5)` sits exactly on a pixel corner. Sedona burns the
corner-touched cell (the geometry does touch it, at that point); GDAL omits it.
## Notes
This only occurs at exact grid-aligned vertices, never on general-position
input, and it is a *superset* of GDAL — Sedona's result is consistent with the
literal documented rule ("all pixels touched by the geometry"). GDAL's own
behaviour at these degeneracies is not internally consistent (it includes some
corner touches and omits others depending on edge direction), so there is no
single reference to match. Filing for visibility; low priority and not
currently planned.
Discovered while fixing #3118.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
