to_datetime removes _meta.index.name but does not remove the partition index names #4904
Description
This problem was noticed when trying to call to_parquet and getting a columns do not match the metadata ValueError. It is similar to #3003 but I believe that behavior was actually fixed for assign (?)
Example:
in.csv:
0,1900-01-01T00:00:00.000000000
1,2019-01-07T00:00:00.000000000
2,2012-04-18T00:00:00.000000000
3,2004-07-01T00:00:00.000000000
4,2010-05-12T00:00:00.000000000
import dask.dataframe as ddf
this_ddf = ddf.from_pandas(pd.read_csv('in.csv', index_col=0),npartitions=1)
this_ddf.dates = ddf.to_datetime(this_ddf.dates, format='%Y-%m-%dT%H:%M:%S')
print(this_ddf._meta.index.name, this_ddf.head().index.name)
The printed value is: (None, 'ix'). Expected result: ('ix', 'ix')