pandas.read_csv: how to skip comment lines

I think I misunderstand the intention of read_csv. If I have a file 'j' like

# notes
a,b,c
# more notes
1,2,3

How can I pandas.read_csv this file, skipping any '#' commented lines? I see in the help 'comment' of lines is not supported but it indicates an empty line should be returned. I see an error

df = pandas.read_csv('j', comment='#')

CParserError: Error tokenizing data. C error: Expected 1 fields in line 2, saw 3

I'm currently on

In [15]: pandas.__version__
Out[15]: '0.12.0rc1'

On version'0.12.0-199-g4c8ad82':

In [43]: df = pandas.read_csv('j', comment='#', header=None)

CParserError: Error tokenizing data. C error: Expected 1 fields in line 2, saw 3

So I believe in the latest releases of pandas (version 0.16.0), you could throw in the comment='#' parameter into pd.read_csv and this should skip commented out lines.

These github issues shows that you can do this:

https://github.com/pydata/pandas/issues/10548
https://github.com/pydata/pandas/issues/4623

See the documentation on read_csv: http://pandas.pydata.org/pandas-docs/stable/generated/pandas.read_csv.html

One workaround is to specify skiprows to ignore the first few entries:

In [11]: s = '# notes\na,b,c\n# more notes\n1,2,3'

In [12]: pd.read_csv(StringIO(s), sep=',', comment='#', skiprows=1)
Out[12]: 
    a   b   c
0 NaN NaN NaN
1   1   2   3

Otherwise read_csv gets a little confused:

In [13]: pd.read_csv(StringIO(s), sep=',', comment='#')
Out[13]: 
        Unnamed: 0
a   b            c
NaN NaN        NaN
1   2            3

This seems to be the case in 0.12.0, I've filed a bug report.

As Viktor points out you can use dropna to remove the NaN after the fact... (there is a recent open issue to have commented lines be ignored completely):

In [14]: pd.read_csv(StringIO(s2), comment='#', sep=',').dropna(how='all')
Out[14]: 
   a  b  c
1  1  2  3

Note: the default index will "give away" the fact there was missing data.

pandas.read_csv: how to skip comment lines

Related

Recent Posts