PyPDF2 throws exception during extract_text()

[lenemeth]

I'm working on a script that is parsing PDF invoices and I'm getting exception during pdf reading. This happens only with a specific type of PDF coming from a tapwater utility service provider company. However, all PDFs from them are failed to be parsed with the same error.

Environment

Windows 10

c:\>python --version
Python 3.11.1

c:\>pip show pyPdf2
Name: PyPDF2
Version: 3.0.1
Summary: A pure-python PDF library capable of splitting, merging, cropping, and transforming PDF files
Home-page:
Author:
Author-email: Mathieu Fenniak <biziqe@mathieu.fenniak.net>
License:
Location: C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages
Requires:
Required-by:

Code + PDF

from PyPDF2 import PdfReader
reader = PdfReader(filePath)

for page in reader.pages:
     text = page.extract_text()

I can share the PDF in email as it contains personal data (invoice). Let me know where to send it

Traceback

Traceback (most recent call last):
  File "C:\Users\lenemeth\Documents\AOB\doc\Erőmű 8\albérlet\szamla_parser\EstateManager.py", line 63, in <module>
    em.parse_invoices()
  File "C:\Users\lenemeth\Documents\AOB\doc\Erőmű 8\albérlet\szamla_parser\EstateManager.py", line 22, in parse_invoices
    self.ip.parse_invoices(self.config['input_data']['invoices']['directory_path'])
  File "C:\Users\lenemeth\Documents\AOB\doc\Erőmű 8\albérlet\szamla_parser\InvoiceParser.py", line 47, in parse_invoices
    self.extract_pdf(os.path.join(directory, file))
  File "C:\Users\lenemeth\Documents\AOB\doc\Erőmű 8\albérlet\szamla_parser\InvoiceParser.py", line 63, in extract_pdf
    text = page.extract_text()
           ^^^^^^^^^^^^^^^^^^^
  File "C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages\PyPDF2\_page.py", line 1851, in extract_text
    return self._extract_text(
           ^^^^^^^^^^^^^^^^^^^
  File "C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages\PyPDF2\_page.py", line 1342, in _extract_text
    cmaps[f] = build_char_map(f, space_width, obj)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages\PyPDF2\_cmap.py", line 28, in build_char_map
    map_dict, space_code, int_entry = parse_to_unicode(ft, space_code)
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages\PyPDF2\_cmap.py", line 196, in parse_to_unicode
    process_rg, process_char, multiline_rg = process_cm_line(
                                             ^^^^^^^^^^^^^^^^
  File "C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages\PyPDF2\_cmap.py", line 264, in process_cm_line
    multiline_rg = parse_bfrange(l, map_dict, int_entry, multiline_rg)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\lenemeth\AppData\Local\Programs\Python\Python311\Lib\site-packages\PyPDF2\_cmap.py", line 278, in parse_bfrange
    nbi = max(len(lst[0]), len(lst[1]))
                               ~~~^^^
IndexError: list index out of range

[pubpub-zz]

At first glance, Looks like a duplicate of #1091
A PR and a fix is proposed can you try it

[lenemeth]

Thanks! I've tried this one and it seems to be working. However now there is an another issue: the returned text charset seems to be messed up a bit as Hungarian letters (iso-8859-2 / "Latin-2") are unreadable:

I got this: sz♥mlakibocs♥t♦hoz t♣rt☺n☻ regisztr♥ci♦
Should look like this: számlakibocsátóhoz történő regisztráció

Not sure if it's because of this particular PDF type but the rest of the invoices using similar alphapet looks fine :)

[pubpub-zz]

@lenemeth can you provide your pdf please for review

[lenemeth]

@lenemeth can you provide your pdf please for review

@pubpub-zz please provide an email address so that I can send it. It contains personal data (invoice) so I don't want to publicly share it. Thanks for your understanding.

[MartinThoma]

@lenemeth I know that @pubpub-zz values privacy and I could imagine that he wants to keep his email address private. If you want, you can send it to me and I can forward it: info@martin-thoma.de

[lenemeth]

@MartinThoma sent via email. Please share with @pubpub-zz privately.

[MartinThoma]

I did. Thanks for sharing :-)

[pubpub-zz]

@lenemeth,
thanks for your contribution the extraction was buggy with unicode cmap where ranges were set on multiple lines.

Can you check that the PR is now good for you. I will add a test for coverage

[pubpub-zz]

test file for test coverage
iss1533.pdf

[lenemeth]

@pubpub-zz I've checked with all of my invoice types and works well. Thanks for the correction!

[MartinThoma]

Thank you for confirming that it works and thank you for sharing the PDF for investigation. We will close this issue once the PR is merged :-) I guess we will have a fixed version on PyPI on Sunday.

@pubpub-zz Thank you so much for taking care of this again 🙏

[gzeng11]

I have tried to use PyPDF2 to chat with PDF with OpenAI and Langchian. For any PDF files which cannot be copied, it will throw "IndexError: list index out of range. "

If I run the following code:

from PyPDF2 import PdfReader

reader = PdfReader(filePath)

for page in reader.pages:
text = page.extract_text()
print(text)

For this type of PDF files, it will print nothing.

Thanks.

Guoping

[MartinThoma]

PyPDF2 is deprecated. Use pypdf.