7

Consider a file, a.dat, with contents:

address 1, address 2, address 3, num1, num2, num3
address 1, address 2, address 3, 1.0, 2.0, 3
address 1, address 2, "address 3, address4", 1.0, 2.0, 3

I am trying to import with numpy.genfromtxt. However the function sees an additional column in row 3. I get a similar error with pandas.read_csv:

np.genfromtxt('a.dat',delimiter=',',dtype=None,skiprows=1)

ValueError: Some errors were detected !
    Line #3 (got 7 columns instead of 6)

and

pandas read_csv sort of works - but it gives me an unaligned data structure:

pd.read_csv('a.dat')

pandas.parser.CParserError: Error tokenizing data. C error: Expected 6 fields in line 3, saw 7

I'm trying to find an input parameter to compensate for this. I don't mind if I end up with a numpy ndarray or pandas dataframe.

Is there a parameter that I can set within genfromtxt and/or read_csv that will let me ignore the comma within the speech marks?

I note that read_csv includes a quotechar='"' parameter, defined thus:

quotechar : string (length 1) The character used to denote the start and end of a quoted item. Quoted items can include the delimiter and it will be ignored.

This reads to me like read_csv should work for my case by default - yet it doesn't.

I can see that I could pre-process the file to strip out the commas - I'd like to avoid that if possible but would welcome suggestions if this is the only way.

atomh33ls
  • 23,172
  • 18
  • 91
  • 149

2 Answers2

16

Just managed to find this:

The key parameter that I was missing is skipinitialspace=True - this "deals with the spaces after the comma-delimiter"

a=pd.read_csv('a.dat',quotechar='"',skipinitialspace=True)

   address 1  address 2            address 3  num1  num2  num3
0  address 1  address 2            address 3     1     2     3
1  address 1  address 2  address 3, address4     1     2     3

This works :-)

Community
  • 1
  • 1
atomh33ls
  • 23,172
  • 18
  • 91
  • 149
2

Python's built-in csv module can deal with this kind of data.

with open("a.dat") as f:
    reader = csv.reader(f, skipinitialspace=True)
    header = next(reader)
    dtype = numpy.dtype(zip(header, ['S20', 'S20', 'S20', 'f8', 'f8', 'f8']))
    data = numpy.fromiter(itertools.imap(tuple, reader), dtype=dtype)
Sven Marnach
  • 483,142
  • 107
  • 864
  • 776