write_frame fails to write NaN values (io.sql)

[jhubele]

At least connections via pyodbc or ceODBC do not accept NaN values in INSERT statements.

If I replace the NaN with None it is working as expected, i.e. these are NULLs in the data base:

frame2 = frame.where((pd.notnull(frame)), None)
data = [tuple(x) for x in frame2.values]
cur.executemany(insert_query, data)

It is pretty expensive to copy the data frame. There is probably a better solution.

I have tested this only with SQL Server for which I have written a small extension to sql.py.

[jreback]

closing as duped by #2754