Some help in refining this regex for CSV files

Tim Chase python.list at tim.thechases.com
Thu Dec 6 15:27:21 CET 2012


On 12/06/12 01:21, Oltmans wrote:
> Hi guys,
> 
> I've to deal with CSVs that look like following
> 
> CSV (with one header and 3 legit rows where each legit row has 3 columns)
> ----
> Some info
> Date: 12/6/2012
> Author: Some guy
> Total records: 100
> 
> header1, header2, header3
> one, two, three
> one, "Python is great, so are other languages, isn't ?", three
> one, two, 'some languages, are realyl beautiful\r\n, I really cannot deny \n this \t\t\t fact. \t\t\t\tthis fact alone is amazing'
> ----
> 
> So inside this CSV, there will always be bad lines like the top 4 (they could end up in the beginning, in the middle and even in the last). So above sample, csv has 3 legit lines and a header. I want to read those three lines and here is a regex that I came up with (which clearly isn't working)
> 
>     #print line
>     pattern = r"([^\t]+\t|,+)"
>     matches = re.match(pattern, line) 
> 
> Do you've any better ideas guys? I will really appreciate all help.

I agree with Mark that using the "csv" module will likely be your
easiest way to go.  Just consume the lines you don't want before
passing it to the csv.reader(), or parse them and discard invalid
items.  The first could be done something like

  import csv
  f = file("data.csv", "rb")
  while True:
      line = f.next().rstrip("\r\n")
      if not line: break
  r = csv.reader(f)
  for row in r:
      print repr(row)

The latter might be done something like

  f = file("data.csv", "rb")
  r = csv.reader(f)
  for row in r:
      if len(row) != 3: continue
      print repr(row)

However, I also noticed that your example file doesn't seem to fit a
true csv file definition, as you seem to switch quoting notations,
sometimes using single, sometimes using double quotes.

-tkc






More information about the Python-list mailing list