New to using GeoPandas instead of Pandas. I am trying to export a GeoPandas dataframe to a PostgreSQL database (with PostGIS extension).
The import is from the same database as I try to export to using:
gpd.GeoDataFrame.from_postgis(sql, con, geom_col='geom')
The CRS seems to be imported as well as srid=25832. When I print gdf.crs it yields as expected:
{'init': 'epsg:25832'}
A sample of the geodataframe looks like this, where the last column geom (the geometry) seems to be the problem child:
row_num model ... mean_res geom
0 1 644862.0 ... 37.40605 POINT (573187.125 6131561.456)
1 2 644862.0 ... 36.36506 POINT (573189.374 6131560.369)
2 3 644862.0 ... 35.42172 POINT (573191.623 6131559.282)
3 4 644862.0 ... 34.90574 POINT (573193.873 6131558.194)
4 5 644862.0 ... 34.36104 POINT (573196.122 6131557.107)
.. ... ... ... ... ...
116 117 644862.0 ... 41.61498 POINT (573448.099 6131435.305)
117 118 644862.0 ... 41.26093 POINT (573450.350 6131434.217)
118 119 644862.0 ... 41.39089 POINT (573452.600 6131433.129)
119 120 644862.0 ... 41.49015 POINT (573454.851 6131432.041)
120 121 644862.0 ... 41.56080 POINT (573457.101 6131430.954)
My problem is when I try to export the dataframe using:
df.to_sql(tablename, con=engine, schema=schemaname, if_exists='replace', index=False)
The engine is from the SQLAlchemy lib using psycopg2, which is working for a "normal" Pandas dataframe (i.e. without geometry)
I get the error:
"Table could not be edited: databasename.schemaname.tablename
'GeometryDtype' object has no attribute 'base'"
Best Answer
GeoPandas will support postGIS from the next release. At this moment, you can try using this branch https://github.com/geopandas/geopandas/pull/1248 which brings that support or just get the code from it.
Since geopandas 0.8.0 you will be able to do
df.to_postgis(tablename, con=engine, schema=schemaname, if_exists='replace', index=False).