ENH: support missing values in dissolve

[MAGALLANESJoseManuel]

In R, after using dissolve, I get missing values. If I use dissolve in geopandas, the missing values in the 'by=' are not considered, and all polygons have non-missing values...how can I keep them? (as in R)

[martinfleis]

Can you show an reproducible example illustrating the behaviour? I think I understand what do you mean, but just to be sure.

[MAGALLANESJoseManuel]

Let me give you this quickly...I have this:

As you see WINNEREP has missing values. Then I dissolve:

As I WINNEREP has two values (true/false) expect that dissolving produsces 3 geometries, one of them for missing values in WINEREP, but it does not:

In R, you get the missing after dissolving:
allZipREP=ms_dissolve(allZipBetter,field = "winnerREP")

How can I achieve the same as in R. Thanks in advance!

[jorisvandenbossche]

This is the pandas groupby behaviour, which drops missing values. In the next pandas release, however, this is a new feature to preserve missing values: pandas-dev/pandas#30584
So then we can also make this available in geopandas' dissolve.

[martinfleis]

@MAGALLANESJoseManuel As a workaround in the meantime, you can generate NaN-based geometry manually and append it to the result of dissolve.

nan_geom = allZipBetter[allZipBetter.winnerREP.isna()].unary_union

[mgmanalili]

I have the same issue. The column used in dissolve 'by' disappears. Is there any way to implement something like, keep=True?

[martinfleis]

@mgmanalili I don't think that is the issue described above. The column used in dissolve is used as an index in the dissolved GeoDataFrame, it does not disappear. If you want to have it as a regular column, you can call reset_index() after dissolving.

[martinfleis]

dropna keyword has been implemented. Closing.