Pieter Roggemans
Pieter Roggemans
> GDAL ignores setting the feature index to read (via `OGR_L_SetNextByIndex`) when using arrow, so that isn't a drop-in fix. > > If `max_features` is set, we can iterate over...
I think it would be practical to be able to overrule the default values for those kind of parameters in an environment variable. Par example for `use_arrow`: in `read_dataframe()`, change...
Shapely with GEOS 3.13 also seems to crash on the wkt: I've had a quick look at the shapely code, but don't immediately see where it can go wrong specifically...
This has been fixed in shapely.
> It does work using Snap Rounding, with a fairly large scale factor (10^7). Not sure if that's available in Shapely. I don't immediately find anything regarding snaprounding in the...
OK, I further automated the discovery of the extra cases... each .csv contains some WKTs that when unioned trigger an error. [error_6.csv](https://github.com/user-attachments/files/17217946/error_6.csv) [error_0.csv](https://github.com/user-attachments/files/17217949/error_0.csv) [error_1.csv](https://github.com/user-attachments/files/17217951/error_1.csv) [error_2.csv](https://github.com/user-attachments/files/17217953/error_2.csv) [error_3.csv](https://github.com/user-attachments/files/17217955/error_3.csv) [error_4.csv](https://github.com/user-attachments/files/17217957/error_4.csv) [error_5.csv](https://github.com/user-attachments/files/17217959/error_5.csv)
> It looks like the Shapely [overlay](https://shapely.readthedocs.io/en/stable/reference/shapely.unary_union.html#shapely.unary_union) functions accept a `gridSize` parameter. Ah, oops, ok... grid_size is synonymous for snap-rounding... I should have known, sounds logical (blushing).
I was actually allready applying `gridsize=0.01` (the data is projected, so 0.01 meter is still plenty precise), but I'm kind of using geopandas, and in geopandas the `gridsize` parameter isn't...
Can you share the GDB triggering this error?
> Yes. It's because merging loses information, and takes slightly more time. So the design decision was to leave the results unmerged, and use a separate call to merge them...