Closed sebastic closed 1 year ago
Looks like Geos 3.12 changed two things that break these tests.
They added parentheses in MultiPoint
, which is technically correct. Very surprised this wasn't already the case. Easy enough to adjust that part. The change also broke a Shapely test. https://github.com/shapely/shapely/pull/1820 They now have a test that checks based on version, but not sure how we go about that without a bunch of regex replacement which probably defeats the purpose of testing. Might be worth just testing against static strings vs Shapely / Geos. Then we don't break again when things like this happen. It is nice to try and match up to what they do though, so there is some purpose to keeping it.
And I guess they flipped from 0
to NaN
when adding a third dimension? Not sure what is up with that, still need to dig through where that change is and see what the logic is. I think this may be a Geos issue though, as NaN
shouldn't be valid in WKT.
Figuring out the best way to test everything is the hard part.
we could do the same as they do (skip if geos version is too low) (https://github.com/shapely/shapely/pull/1820/files#diff-dd888b20a0fd5d6ecbff7de3e85aa28070ebc9a80be24711eeb57f4eaf69d2c3R326-R329)
I think it's nice to have a validation using shapely
And I guess they flipped from 0 to NaN when adding a third dimension? Not sure what is up with that, still need to dig through where that change is and see what the logic is. I think this may be a Geos issue though, as NaN shouldn't be valid in WKT.
maybe linked to https://github.com/shapely/shapely/issues/1524#issuecomment-1239340672 🤔
The Debian package build fails with GEOS 3.12: