Skip NameObject when building outline

[michaeleveringham]

Fixes #193, uses @hannal's solution to resolve read/merge issues with wkhtmltopdf pdfs by skipping NameObject entities. Related to #778.

[michaeleveringham]

@MartinThoma Think I got it now. Thanks again for the help.

[michaeleveringham]

Never mind, need to update some tests it seems.

[michaeleveringham]

I tried numerous "bad" PDFs but couldn't get this error to raise. Will have to look into creating a PDF with garbage anchors and uploading to the samples repo.

[MartinThoma]

@LamerLink Could it be that your PR just solved an issue that test_unexpected_destination has? Maybe we could simply adjust the test to:

def test_unexpected_destination():
    url = "https://corpora.tika.apache.org/base/docs/govdocs1/913/913678.pdf"
    name = "tika-913678.pdf"
    reader = PdfReader(BytesIO(get_pdf_from_url(url, name=name)))
    merger = PdfMerger()
    merger.append(reader)

in this PR?

[michaeleveringham]

@MartinThoma Good call, updated. I thought about adding an assert len(reader.outlines) == 0 since there'll be no valid destinations added to the outline in this case, but I decided not too since it's a little ubiquitous fir this case. Not to mention some normal PDFs can have no outlines, especially if they're pure images.

Looks like all checks have passed!

[codecov]

Codecov Report

Merging #1068 (9d34e86) into main (c529884) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #1068   +/-   ##
=======================================
  Coverage   92.26%   92.26%           
=======================================
  Files          24       24           
  Lines        4794     4794           
  Branches      990      990           
=======================================
  Hits         4423     4423           
  Misses        230      230           
  Partials      141      141           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c529884...9d34e86. Read the comment docs.

[MartinThoma]

@LamerLink Thank you for your effort. There was another PR #1076 which also solved the same issue, but that one is clearer to me why it solves the issue. As the issue is now (hopefully) gone, this PR seems no longer necessary.

Please let me know if I got something wrong - I can always re-open a PR

[michaeleveringham]

@MartinThoma No problem at all, I read over that PR and it makes sense to me! Thanks for all the effort.

[MartinThoma]

I've just noticed that your pr increases coverage quite a bit. That is unexpected to me. I want to understand that better

[MartinThoma]

I'm not sure why it originally showed that the test coverage was increased by that much. If we would do it now with #1158, the coverage would decrease slightly (which is what I expected)