pandas 1.5.0 compatibility issues

[martinfleis]

pandas 1.5 released yesterday broke our CI and is emitting a ton of new deprecation warnings.

https://github.com/pysal/momepy/actions/runs/3091088424/jobs/5000737392#step:6:187

[jGaboardi]

Woah boy...

[jGaboardi]

Not 100% sure why this is complaining so much. The message indicates the problem is with using .iloc, but we are using .loc there. I had thought using .loc was always safe for setting values in place?

elements.py:281: FutureWarning: In a future version, `df.iloc[:, i] = newvals` will attempt to set the 
values inplace instead of always setting a new array. To retain the old behavior, use either 
`df[df.columns[i]] = newvals` or, if columns are non-unique, `df.isetitem(i, newvals)`

[martinfleis]

The warning is a bit confusing but the change seem to apply to both loc and iloc - https://pandas.pydata.org/docs/whatsnew/v1.5.0.html#inplace-operation-when-setting-values-with-loc-and-iloc

[martinfleis]

The Percentiles broken in the SDSS 2022 tutorial are caused by pandas 1.5.0. With 1.4.4 as the only difference between the envs (numpy same in both), the issue is not present.

[jGaboardi]

Probably in numpy.nanpercentile, correct? Or is it specifically in pandas?

[martinfleis]

Update - there is a mismatch in assigned unique_id in buildings and tessellation. The building and its tessellation cell should have the same id. They do not right after tessellation is created:

So the issue seems to be within Tessellation. Digging deeper.

[martinfleis]

One note aside - we have a bug in Percentiles as well. This should be the same as line 898

momepy/momepy/diversity.py

Line 925 in c574bcf

results_list.append(np.nan)

[martinfleis]

I've found where it gets messed up.

momepy/momepy/elements.py

Lines 318 to 320 in c574bcf

	morphological_tessellation = gpd.clip(
	morphological_tessellation, gpd.GeoSeries(limit, crs=gdf.crs)
	)

clip seems to mess the order of geometries vs attributes. Still not sure where in clip that happens but it is a bug outside of momepy.

[martinfleis]

It happens here https://github.com/geopandas/geopandas/blob/af50debbbbeab62f3601518f814c9cb271b1859c/geopandas/tools/clip.py#L66-L68 There's absolutely no reason why this should happen. It must be a regression in pandas... This is a proper rabbit hole.

[jGaboardi]

Excellent digging.