Keywords that end a code stream are ignored.

[qued]

Bug report

Certain PDFs contain code where a stream of operations ends with a keyword, without any of the END_KEYWORD characters following it. This causes the parsing to end without the keyword being parsed, even though otherwise the operation stream is valid. For example, in the attached PDF, a code stream consists of b'543 0 0 738 0 0 cm\n /Im1 Do'.

The expected behavior is that if a code stream terminates while parsing a keyword, if the keyword is valid it will execute. Readers such as Adobe Reader, Chrome, and Firefox are able to render the images in the example PDF linked below.

To reproduce

from pdfminer.high_level import extract_pages

elements = [el for page in extract_pages("IRS-form-1987.pdf") for el in page]

...will produce a list with two empty lists (one for each page), when each page contains a single image as an XObject.

Attached file used to reproduce: IRS-form-1987.pdf

[pietermarsman]

I can reproduce this.

from pdfminer.high_level import extract_pages

for page in extract_pages("IRS-form-1987.pdf"):
    for el in page:
        print(page, el)

Prints nothing.

[cole-dda]

The fix issue other problem。

Because parser use buf，sush as, stream = b'/IM Do'
but now buf=b'/IM D'
parse keyword => 'D'
use old method ,continue fill buf,again parse => 'Do'

code may be: