PDFObjRef is not iterable

[pietermarsman]

Bug report

Copy of #471 (by @imochoa)

Sadly, I cannot upload the problematic PDFs due a non-disclosure agreement. I can however point out the issue and share my fix.

When trying to instantiate a PDFCIDFont object at:

pdfminer.six/pdfminer/pdfinterp.py

Line 193 in 0b44f77

font = PDFCIDFont(self, spec)

Where

self
<pdfminer.pdfinterp.PDFResourceManager at 0x7f0c80493048>

spec
{'Type': /'Font',
 'Subtype': /'CIDFontType2',
 'BaseFont': /'ISOCPEUR',
 'FontDescriptor': <PDFObjRef:5>,
 'CIDSystemInfo': <PDFObjRef:41>,
 'W': <PDFObjRef:13>,
 'Encoding': /'Identity-H'}

The execution ends up at PDFStream.decode with:

self
<PDFStream(7): raw=8963, {'Length1': 28048, 'Length': 8961, 'Filter': /'FlateDecode', 'DecodeParms': <PDFObjRef:39>}>

The origin of the bug is that in this case, get_filters(self)

pdfminer.six/pdfminer/pdftypes.py

Line 258 in 0b44f77

filters = self.get_filters()

returns:

 In[8]: filters
Out[8]: [(/'FlateDecode', <PDFObjRef:39>)]

As you can see, the second element of the first and only Tuple is a PDFObjRef, which is then saved to params and fails a little later down the line when trying to evaluate 'Predictor' in params:

pdfminer.six/pdfminer/pdftypes.py

Line 297 in 0b44f77

if params and 'Predictor' in params:

I noticed that the default value of params is an empty dictionary, which effectively skips that check. So I extended the check to only continue if params was a dictionary:
https://github.com/imochoa/pdfminer.six/blob/2d996c9ae26c8a336711178a8afe3091e1140970/pdfminer/pdftypes.py#L297

I think the underlying error is the fact that get_filters(self) is returning a PDFObjRef instead of a dictionary. I've tried to find the exact origin, but I'm not too familiar with the project and couldn't pinpoint the exact issue. The furthest I got was that PDFParser was that the problematic <PDFObjRef:39> value was being set at the dictionary in:

pdfminer.six/pdfminer/pdfparser.py

Line 122 in 0b44f77

obj = PDFStream(dic, data, self.doc.decipher)

as:

dic['DecodeParms']
<PDFObjRef:39>

Since I couldn't prevent it from coming up, refining the check seemed like the next best option and it works well on the 2 problematic PDFs I have.

[imochoa]

Hello again :)

I found a PDF I can share where the issue is happening, in the text box at the bottom right where it says "PET BLACK"

pdfminer_testpart.pdf

[EvaSDK]

Jumping in as I had the same issue on some PII document that I would not be able to share.
The minimal code sample to reproduce is the following:

from pdfminer.high_level import extract_text
extract_text("./pdfminer_testpart.pdf")

It should return:

'8\n\n7\n\n6\n\n5\n\n4\n\n3\n\n2\n\n1\n\n150,00\n\n30,00\n\n(cid:72) 0,05 A\n\n0\n0\n,\n0\n2\n\n0\n0\n,\n8\n\n(cid:69) 0,05\n\n0\n0\n,\n0\n5\n\nA\n\nF\n\nE\n\nD\n\n20,00\n\n16,00\n\n+\n0,05\n15,00 - 0,00\n\nC\n\n0\n0\n,\n0\n4\n\n0\n0\n,\n0\n2\n\nR 1 8 , 0 0\n\nM12x1.75 - 6H\n\n0\n0\n,\n5\n4\n\nB\n\nA\n\n0\n0\n,\n6\n1\n(cid:142)\n\n0\n0\n,\n6\n1\n\n+\n0,50\n15,00 - 0,00\n\n60,00 (cid:66)0,02\n\n100,00 (cid:66)0,05\n\n132,00\n\n9\nH\n0\n1\n(cid:142)\n\n9\nH\n0\n1\n(cid:142)\n\n(cid:68) 0,1 A\n\n+\n0,00\n70,00 - 0,02\n\n50,00\n\n(cid:76) 0,1\n\n(cid:76) 0,1\n\n0\n0\n,\n5\n3\n\nF\n\nE\n\nD\n\nC\n\nB\n\nAllgemeintoleranzen\n\nMATERIAL\n\nDIN ISO 2768 - mK\n\nPET BLACK\n\nFINISH\n\nEloxieren (natur)\n\nRa 1,6\n\nDate\n29.03.2021\n\nName\nLucas Giering\n\nDrawn\n\nChecked\n\nStandard\n\nArretierungshilfe\n\nA\n\n1 \n\nA2\n\n8\n\n7\n\n6\n\n5\n\n4\n\nState\n\nChanges\n\nDate\n\nName\n\n3\n\n2\n\n1\n\n\x0c'

[pietermarsman]

Closed by #906