AssertionError: Invalid interval about pdfio.jl

[sambitdash]

The font enconding mapping to unicode used in PDF has issues. Please share the file to investigate.

The encoding cmaps have ranges lo:hi defined in them. It seems for some reason in the mapping file you have high value lesser than low value. Hence, this assertion error.

https://github.com/sambitdash/Rectangle.jl/blob/54f36a07257b17b8bc8e1f4698aef20df90d632f/src/interval.jl#L5

[bdeonovic]

A few comments:

1. I printed one page of the pdf (print to pdf on windows) which was causing the error so I could post it here as an example. However, when I tried to run the extract on this 1 page example the extract worked.
2. The extract doesn't correctly extract the text. The first sentence should be:

U ovoj je knjizi riječ pretežito o hobitima i iz nje će čitatelj doznati štošta o njihovu
značaju i nešto malo o njihovoj povijesti.

but the extract function seems to have a problem with the accent marks. I get this:

U ovoj je knjizi rijee
zna
Crvene knjige o Zapadnoj pokrajini koji su ve objelodanjeni pod naslovom Hobit.

which doesn't have accent marks and skips a bunch of text.

Thoughts?

test.pdf

[sambitdash]

I would believe you have some issues related to the font encoding in the file. If I open the file in Adobe Reader and select and copy the text I see exactly below. Which is close to what you are observing. This happens when the font toUnicode c-maps are not properly transferred. The extract text works on the same principle of copying and pasting text from a PDF file.

1 O hobitima
U ovoj je knjizi rije e
zna
Crvene knjige o Zapadnoj pokrajini koji su ve objelodanjeni pod naslovom Hobit.

[sambitdash]

I will need to investigate the original file with the C-Map to realize why the file does not get transmitted properly. Please share it here, if possible. If there are security concerns you can mail me at: sambitdash at gmail

[bdeonovic]

email sent

[sambitdash]

@bdeonovic Sorry for my delay in looking into the file. The CMap file in the PDF is not aligned to the spec. Figure-6 in the attached spec.

5014.CIDFont_Spec.pdf

That's the reason some readers behave differently. While I will try to repair the cmap for a special case, this is not the correct approach. Code space ranges are rectangular regions in the byte plane and not numbers.

/Registry (BKABIP+TT5+0) /Ordering (T42UV) /Supplement 0 >> def
/CMapName /BKABIP+TT5+0 def
/CMapType 2 def
1 begincodespacerange <00fb> <0108> endcodespacerange

2 beginbfchar
<00ff> <0111>
<0108> <0110>
endbfchar
2 beginbfrange
<00fb> <00fc> <0106>
<00fd> <00fe> <010C>
endbfrange
endcmap CMapName currentdict /CMap defineresource pop end end

is the CMap. As per the CMap spec the codespace range should have 2 elements.

2 begincodespacerange 
   <00fb> <00ff>
   <0100> <0108> 
endcodespacerange

[sambitdash]

A few comments:

1. I printed one page of the pdf (print to pdf on windows) which was causing the error so I could post it here as an example. However, when I tried to run the extract on this 1 page example the extract worked.
2. The extract doesn't correctly extract the text. The first sentence should be:

U ovoj je knjizi riječ pretežito o hobitima i iz nje će čitatelj doznati štošta o njihovu značaju i nešto malo o njihovoj povijesti.

but the extract function seems to have a problem with the accent marks. I get this:

U ovoj je knjizi rijee zna Crvene knjige o Zapadnoj pokrajini koji su ve objelodanjeni pod naslovom Hobit.

which doesn't have accent marks and skips a bunch of text.

Thoughts?

test.pdf

On Page-6 of the document you shared, I get:

     U ovoj je knjizi riječ pretežito o hobitima i iz nje će čitatelj doznati štošta o njihovu 
     značaju i nešto malo o njihovoj povijesti. 

This is what you are expecting. While I have introduced a workaround in the code, this is not the code as per spec.

[sambitdash]

9ed161f fixes this now.