Pandas API change leads to error in stare_dissolve()

[mbauer288]

In STAREPandas/starepandas/staredataframe.py

Changes to the Pandas API result in an STAREDataFrame unexpectedly being converted to a Pandas DataFrame which leads to failure when that DataFrame is treated as if it were a STAREDataFrame.

Suggest that we overload the Pandas set_index() function with a STAREDataFrame class set_index() function so we don't run into this issue when STAREDataFrames indexing is altered.

def stare_dissolve(self, by=None, num_workers=1, geom=False, aggfunc="first", **kwargs):
    ...
    sdf = STAREDataFrame(dissolved, columns=[by, self._sid_column_name])
    # sdf is a STAREDataFrame

    # Changes to the Pandas API result in sdf being changed to a Pandas DataFrame
    sdf.set_index(by, inplace=True)

    # This fails because sdf isn't a STAREDataFrame 
    sdf.set_sids(self._sid_column_name, inplace=True)

FIX

def stare_dissolve(self, by=None, num_workers=1, geom=False, aggfunc="first", **kwargs):
    ...
    sdf = STAREDataFrame(dissolved, columns=[by, self._sid_column_name])

    # Return sdf to a STAREDataFrame, required new_sdf being created and then recast.
    new_sdf = sdf.set_index(by, inplace=False)
    sdf = new_sdf
    sdf.__class__ = STAREDataFrame

    sdf.set_sids(self._sid_column_name, inplace=True)

[mbauer288]

okay found more similar issues. Basically anytime we call a pandas operation on a STAREPandas dataframe the result is a Pandas dataframe. We need to look at the source code of GeoPandas to see how they use overloading of Pandas so this doesn't happen to their dataframes. Even simply reordering or renaming the columns of a STAREPandas dataframe makes it a Pandas dataframe.