addition to support multi page invoices

[noxqs]

Thanks for your work on this project, you made my Monday enjoyable.
I would like to return the favour and contribute my code changes over the day based on the missing feature to support multi page invoices. The start / end for lines is only calculated once which makes it impossible to find all line items.

Hence in parser/lines.py I suggest to make a small modification:

  # start = re.search(settings["start"], content)
  # end = re.search(settings["end"], content)

  starts = list(re.finditer(settings["start"], content))
  ends = list(re.finditer(settings["end"], content))

  assert len(starts) == len(ends), "start end miss match"
  if not starts or not ends:
      logger.warning("no lines found - start %s, end %s", start, end)
      return

  new_content=""
  for start,end in zip(starts,ends):
      new_content += content[start.end(): end.start()]

  content = new_content
  # content = content[start.end() : end.start()]

This worked for me.

There is some cleaning up to do when you get the lines back, not sure if this belongs in your library but I used:

        for line in res['lines']:   # res is the output of extract_data
            add = {lk: lv for lk, lv in line.items() if lv not in ['', 0.0, None]}

            if line['index'] != '':    # first item on each line to be parsed
                if item != {}:
                    print(",".join(["%s:%s" % (str(k).strip(), str(v).strip()) for k, v in item.items()]))
                item=add
            else:
                for kk, vv in add.items():
                    if kk in item:
                        item[kk] += " " + vv
                    else:
                        item[kk] = vv

        if item != {}:
            print(",".join(["%s:%s" % (str(k).strip(), str(v).strip()) for k, v in item.items()]))

There are so many combinations possible with regEx (for ex. start can be cond1 | cond2).
Anyway this is quite a useful little gem of a library,

[m3nu]

Thanks for the improvement. Wanna make a pull request for it to make it official and make sure all tests are still working?

[noxqs]

Sure ! Thanks

[rmilecki]

starts = list(re.finditer(settings["start"], content))
ends = list(re.finditer(settings["end"], content))

I think we need some better logic.

Consider invoice with something like:

Lines begin
Lines begin
LINE 1
LINE 2
LINE 3
Line end
Line end

I think that your change would result in pasing all three lines twice.

We should probably make looking for starts and ends iterative. You should probably look for a next start after the last found end.

[bosd]

Would be interesting to see if PR #378 (comment) would accomplish this.

Technically one could call the lines parser multiple times.
It will be slow, as it go trough every line in the pdf multple times.
But it could get the job done.

Sorry, bit in a hurry so for now I can't to provide a more detailed answer and tests.

[rmilecki]

Implemented