ENH: Make top-level Pandas functions serializable

[mrocklin]

import pandas as pd
import pickle
pickle.dumps(pd.read_csv)

AttributeError: Can't pickle local object '_make_parser_function.<locals>.parser_f'

[mroeschke]

Looks like read_csv is serializable in 1.1.0, as well as all callable object in the top level namespace

In [20]: import pickle

In [21]: import pandas as pd

In [22]: pd.__version__
Out[22]: '1.1.0'

In [23]: pickle.dumps(pd.read_csv)
Out[23]: b'\x80\x03cpandas.io.parsers\nread_csv\nq\x00.'

In [24]: for name, obj in pd.__dict__.items():
    ...:     if callable(obj):
    ...:         try:
    ...:             pickle.dumps(obj)
    ...:         except:
    ...:             print(f'{name}' is not pickleable)
    ...:

In [25]:

[mroeschke]

#34976 probably made read_csv serializable.

I suppose we could add unit tests to ensure all these objects are serializable

[ikedaosushi]

May I take it?

[mrocklin]

Oh that's great news. Thank you @mroeschke for checking, and my apologies for testing against an older version.

@ikedaosushi I think that you do not need permission to submit a pull request. I do not know Pandas' development policies, but I think that it is ok for you to contribute.

[mroeschke]

No worries @mrocklin

Yes, @ikedaosushi, happy to have you take this issue.