Get the name of the geometry column(s) of a GeoDataFrame

Question

Sometime, the name of the geometry column of a GeoDataFrame is simply not 'geometry' but a custom name which in some situations cannot be known in advance because it may have been defined dynamically prior to when we need to access it.

So, is there something more simple/elegant/pythonic than the following snippet to get the name of the geometry columns in a GeoPandas' GeoDataFrame?

gdf.dtypes[gdf.dtypes=="geometry"].index.values.tolist()[0]

'custom_name'

Notice the [0] because the list is having only one element, but if your GeoDataFrame is having more than one geometry column, then you may want to fine tune that to suit your needs.

This topic is not helping much as I'd like to stay within GeoPandas.

Answer 1

In geopandas 0.14.2, there is an easier way to find the active geometry column. (I'm not sure what version this started in).

gdf.geometry.name

You can also use the attribute gdf.geometry to always get the current geometry column data.

Only one geometry column can be "active" at any given moment, though you can have as many columns containing the dtype == 'geometry' as you want.

See more details in the docs.

Answer 2

The best result I have had so far is this one:

list(gdf.select_dtypes('geometry'))[0]

And I think it's OK but if there was a more direct method I would be glad to hear about it.

In the case of a GeoDataFrame with multiple geometry columns, as there should normally be one active geometry column, one can get this particular one as follows (see at the bottom for the rare case of a GeoDataFrame with multiple active geometry columns):

import geopandas as gpd
from shapely.geometry import Point

d = {
  'col1': ['name1', 'name2'],
  'fancy': [Point(1, 2), Point(2, 1)],
  'custom': [Point(3, 4), Point(5, 6)],
}

gdf = geopandas.GeoDataFrame(d)


gdf 
>: 
    col1        fancy       custom
0  name1  POINT (1 2)  POINT (3 4)
1  name2  POINT (2 1)  POINT (5 6)

Currently, none of the column is seen a a geometry column:

gdf.dtypes
>: 
col1      object
fancy     object
custom    object
dtype: object

At this stage, if we use the method shown above to extract the geometry column name on the GeoDataFrame, it throws an error:

gdf.dtypes
>: 
col1      object
fancy     object
custom    object
dtype: object

# running this naturally raises an error:
list(gdf.select_dtypes('geometry'))[0]

Traceback (most recent call last):

  File "/tmp/ipykernel_369762/3000512511.py", line 1, in <cell line: 1>
    list(gdf.select_dtypes('geometry'))[0]

IndexError: list index out of range

Because the list is actually empty:

gdf.select_dtypes('geometry')

>: 
Empty GeoDataFrame
Columns: []
Index: [0, 1]

That's why we need to actually "set" the geometry column, i.e. we mark it as active:

gdf.set_geometry('fancy', inplace=True, crs=4326,)

gdf.dtypes
>: 
col1        object
fancy     geometry
custom      object
dtype: object

So mow the previous code returns the active geometry column:

list(gdf.select_dtypes('geometry'))[0]
>: 'fancy'

But fortunately, you can set more than one active geometry column (although this is not the most common case):

gdf.set_geometry('custom', inplace=True, crs=4326,)

gdf.dtypes
>: 
col1        object
fancy     geometry
custom    geometry
dtype: object

Then, because the list of geometry columns has more than one element now, you will have to manually figure out which element you want:

list(gdf.select_dtypes('geometry'))
>: ['fancy', 'custom']