Text being incorrectly parsed in table

[enrac5]

I have a very odd issue with the attached file:
test.pdf

Basically, the text in the second column is being read incorrectly and I'm not sure why. This is basically what I'm doing:

import pdfplumber

pdf_path = '/[path to attached]/test.pdf'
pdf = pdfplumber.open(pdf_path, repair=True)
for page in pdf.pages:
    tables = page.extract_tables()
    for table in tables:
        for row in table:
            if row[1] is not None:
                print(row[1])

The output should be 13+13 but I get 113+13. This is just a small part of a 78 page document (a PDF printed from MS-Word). (Puts on Leia's clothes, "Help me @jsvine, you're my only hope")

[jsvine]

Hi @enrac5, it looks like the PDF has two 1s written on top of one another:

print(page..extract_text(layout=True))

Produces:

...
1  113 +13 MT #1 FADES IN OVER  1 52+11  58+10  5+15 LUCY TO KOS
...

If you use page.dedupe_chars(), this seems to fix it:

print(page.dedupe_chars().extract_text(layout=True))

...
1  13 +13 MT #1 FADES IN OVER   1 52+11  58+10  5+15 LUCY TO KOS
...

[jsvine]

Strange. I run the exact same code as you but get the correct output.

Ran that on Python 3.10.4 on MacOS, though neither should not have any effect on the processing. Try upgrading pdfplumber and running print(pdfplumber.__version__) to confirm?

[enrac5]

Which version of MacOS are you on?

[jsvine]

For an operation like this one, the operating system should not have any effect on the results. (Both theoretically, and from practical experience; I haven't encountered any PDF where an operation such as dedupe_chars performs differently on different OSes.) Different versions of Ghostscript (used for repair=True), however, may have an effect, so I'd focus on the output of without repair=True.

Is it possible you're processing a slightly different version of the PDF than the one shared in this issue?

[enrac5]

I'm using the same PDF, though this is a stripped down example of another one I cannot share.

[enrac5]

On my WSL instance, this works amazingly, however on my MacOS install, it doesn't, that's the flummoxing part.