read_csv dateparser empty date's are assigned today's date

[jassinm]

Hi,

import StringIO
s = StringIO.StringIO("Date, test\n2012-01-01, 1\n,2")
pd.read_csv(s, parse_dates=["Date"])


    Date     test
0    2012-01-01 00:00:00     1
1    2012-11-15 00:00:00     2

second row should be None.

is this the default behaviour?
this is a quick fix for me

from dateutil.parser import parse as dateparser
def parse_date(d):
    if d.strip():
        return dateparser(d)

pd.read_csv(s, parse_dates=["Date"], date_parser=parse_date)


using pandas 0.9.1

[changhiskhan]

I believe this is fixed now on master:

In [2]: import pandas as pd

In [3]: paste
import StringIO
s = StringIO.StringIO("Date, test\n2012-01-01, 1\n,2")
pd.read_csv(s, parse_dates=["Date"])
## -- End pasted text --
Out[3]: 
         Date   test
0  2012-01-01      1
1         nan      2

Can you try updating?

[wesm]

Note however in a pathological case:

In [12]: pd.read_csv(s, parse_dates=["Date"], na_filter=False)
Out[12]: 
                  Date   test
0  2012-01-01 00:00:00      1
1  2012-11-18 00:00:00      2

I'm as yet unconvinced about being beholden to dateutil's wonkiness, but maybe not so bad.