Is your feature request related to a problem? Please describe.

When using the bigquery %%bigquery cell magic, there is no ability to return the result as a GeoPandas GeoDataFrame when the expected results include GEOGRAPHY columns. This leads to repeated post-processing when performing geospatial analysis.

Describe the solution you'd like

An ideal solution would be to add an argument (e.g. --as_geodataframe) so that the magic itself could implicitly perform the conversion.

# this example would use the default behavior of GeoDataFrame and convert the column named 'geometry'
%%bigquery df --as_geodataframe

SELECT
  col1,
  col2,
  geometry,
  ...
  ...
FROM ...

This could be done by leveraging the existing QueryJob.to_geodataframe method.

The same argument could be used to selected specific geometry column:

%%bigquery df --as_geodataframe geography_col

SELECT
  col1,
  col2,
  geography_col,
  ...
  ...
FROM ...

Optional: The argument could also be a dictionary if more options need to be passed into QueryJob.to_geodataframe

Describe alternatives you've considered

Currently using a workaround function to post-process the resulting pandas DataFrame using both shapely and geopandas as follows:

import shapely
import geopandas as gpd

def pandas_to_geopandas(df, col='geometry'):
  # example geometry column
  col = 'geometry_column'

  # use shapely library to parse WKT strings into shapely Geometry based objects
  df[col] = df[col].dropna().apply(shapely.from_wkt)

  # convert to a geopandas Dataframe
  return gpd.GeoDataFrame(df, geometry=col, crs='EPSG:4326')

However, using a magic argument as proposed previously would prevent the need to call this function. This could reduce the number of cells in notebooks that focus on geospatial analysis.

Additional context

References:

• BigQuery column types
  • GEOGRAPHY column type
• BigQuery python API:
  • QueryJob.to_geodataframe
• geopandas
• shapely