read_csv not correctly parsing dates when parse_dates is string and index_col not set

[cancan101]

With:

csv="""A,B,C
1,2,2003-11-1
"""

These all work as expected:

In [40]: pd.read_csv(StringIO(csv), parse_dates="C",index_col="C").index[0]

Out[40]: Timestamp('2003-11-01 00:00:00', tz=None)

In [41]: pd.read_csv(StringIO(csv), parse_dates=["C"],index_col="C").index[0]

Out[41]: Timestamp('2003-11-01 00:00:00', tz=None)

In [42]: pd.read_csv(StringIO(csv), parse_dates=["C"]).C[0]

Out[42]: Timestamp('2003-11-01 00:00:00', tz=None)

but this does not parse the string:

In [39]: pd.read_csv(StringIO(csv), parse_dates="C",).C[0]

Out[39]: '2003-11-1'

[jreback]

I don't think parse_dates can be a string, so not sure this should work at all (I think your first example works because 'C' is True). will mark as a bug though.

[gfyoung]

@jreback : I think this issue can be closed. The "buggy" example fails as expected because parse_dates="C" is the same as parse_dates=True, which will try to convert the index attribute of the resulting DataFrame as datetime objects per the 0.18.0 documentation. By not specifying index to be the C column, it correctly does not parse it as a datetime.

[jreback]

parse_dates : boolean or list of ints or names or list of lists or dict, default False

    * boolean. If True -> try parsing the index.
    * list of ints or names. e.g. If [1, 2, 3] -> try parsing columns 1, 2, 3
      each as a separate date column.
    * list of lists. e.g.  If [[1, 3]] -> combine columns 1 and 3 and parse as
        a single date column.
    * dict, e.g. {'foo' : [1, 3]} -> parse columns 1, 3 as date and call result
      'foo'

So I think this should then have a nice meesage if a non-boolean scalar is passed as its not valid