Explode collect

[jwass]

A couple of additional tools here:
explode() expands multi-part geometries into multiple rows, and into
their single part geometries. This returns a GeoSeries with a MultiIndex.

collect() - take multiple single geometries and combine them into their
Multi* counterparts.

explode() is a new member of GeoPandasBase, but could also be a separate function in tools like collect() is. These are similar to PostGIS ST_Dump and ST_Collect functions.

To use MultiIndex you can use .loc on the first level to get back:

>>> singles = df.explode()
0  0     POLYGON ((970217.0223999023 145643.3322143555,...
   1     POLYGON ((969488.1658325195 149753.5946044922,...
   2     POLYGON ((939997.0946044922 173013.5794067383,...
   3     POLYGON ((961436.3049926758 175473.0296020508,...
1  0     POLYGON ((1021176.479003906 151374.7969970703,...
   1     POLYGON ((1020482.590393066 157430.9501953125,...
   2     POLYGON ((1006493.460205078 157737.2102050781,...
...

>>> singles.loc[0]
0    POLYGON ((970217.0223999023 145643.3322143555,...
1    POLYGON ((969488.1658325195 149753.5946044922,...
2    POLYGON ((939997.0946044922 173013.5794067383,...
3    POLYGON ((961436.3049926758 175473.0296020508,...
dtype: object

Using the NYC boros examples you can do some neat things quickly:

Take the polygon with the greatest area from each multipolygon, preserving the index:

>>> df.explode().groupby(level=0).apply(lambda x: x.loc[x.area.argmax()])
0    POLYGON ((961436.3049926758 175473.0296020508,...
1    POLYGON ((996887.8187866211 208559.3403930664,...
2    POLYGON ((1033946.682983398 231157.9963989258,...
3    POLYGON ((1004601.953430176 259027.5151977539,...
4    POLYGON ((1019370.870605469 268815.8876342773,...
dtype: object

Remove any polygons with small perimeters from each multipolygon, recombine to MultiPolygon:

>>> df.explode().groupby(level=0).apply(lambda x: collect(x[x.exterior.length >= 3000]))
0    (POLYGON ((969488.1658325195 149753.5946044922...
1    (POLYGON ((1021176.479003906 151374.7969970703...
2    (POLYGON ((1029606.076599121 156073.8142089844...
3    (POLYGON ((972081.7882080078 190733.4674072266...
4    (POLYGON ((1015023.713439941 230286.7592163086...
dtype: object

[coveralls]

Coverage increased (+0.37%) when pulling 01c7d4c on jwass:explode_collect into 5219026 on geopandas:master.

[kjordahl]

I like this a lot. Nice use of MultiIndex!

It needs merge conflict with master resolved after #148.

[coveralls]

Coverage increased (+0.25%) when pulling fb57a6c on jwass:explode_collect into 0a82279 on geopandas:master.

[kjordahl]

Looking good, thanks.