Missing real example

[deeplook]

I'm trying to evaluate this package by writing a little example, but I'm running into some errors below. I try to write a small small example, to read a simple CSV file with lat/lon columns, add a geometry column and turn this into a Dask-GeoDateFrame, but I'm running into errors like these:

• AttributeError: 'Series' object has no attribute 'map_partitions'
• AttributeError: 'DataFrame' object has no attribute 'geometry'

import dask_geopandas
import dask.dataframe as dd
import pandas as pd
import geopandas as gd

df = pd.read_csv('airport_volume_airport_locations.csv')

# ok
gdf = gd.GeoDataFrame(
    df, geometry=gd.points_from_xy(df.Airport1Latitude, df.Airport1Longitude)
)

ddf = dask_geopandas.from_geopandas(df, npartitions=4)
# raises AttributeError: 'DataFrame' object has no attribute 'set_geometry'
ddf.set_geometry(
    dask_geopandas.points_from_xy(ddf, "Airport1Latitude", "Airport1Longitude")
)

[jorisvandenbossche]

@deeplook thanks for taking a look and giving feedback! (and yes, examples are still very scarce ..)

The dask_geopandas.from_geopandas expects a GeoDataFrame, but you are passing it a DataFrame (df, and not gdf). We should probably already check that and directly raise an error, instead of letting it result in an error in a later operation.
So you could pass gdf instead, but this of course doesn't do the xy-to-points conversion with dask.

For what you are trying to do (create a dask_geopandas.GeoDataFrame from a dask.DataFrame (and not from a geopandas.GeoDataFrame)), you can do something like:

ddf = dd.from_pandas(df, npartitions=4)
ddf["geometry"] = dask_geopandas.points_from_xy(ddf, "Airport1Latitude", "Airport1Longitude")
gddf = dask_geopandas.from_dask_dataframe(ddf)

The reason is that this is a bit more complicated than the set_geometry which you tried (and works for plain GeoPandas), is because GeoPandas monkey-patches the pandas DataFrame with a set_geometry method so that you can convert a pandas DataFrame which already has a geometry column to a GeoDataFrame using that method.
But we didn't (yet) do the same in dask_geopandas (we probably could for consistency with geopandas).

[deeplook]

Thanks! I'll surely come back with other things... ;)