PDF attachments are not PDF/A-3 compliant

[timoramsauer]

I compile a file with file attachments using anchors.
pdf_variant is set to pdf/a-3b.

Checking the pdf with veraPDF problems are found:

Furthermore, ModDate is missing even though it is required by spec.

The following attributes were added:

• EmbeddedFile/CreationDate (technically not required for compliance)
• EmbeddedFile/ModDate
• EmbeddedFile/MimeType
• Filespec/AFRelationship
• Catalog/AF

This pull request works for me as I only compile to pdf/a.
I understand that this might not be a general fit and I'm happy to adjust it.

However, I have some open questions:
Which attributes shall be part of all pdfs / just pdf/a and therefore likely reside in pdfa.py?

Where should the information come from?
Currently:

• EmbeddedFile/CreationDate: manually defined when creating Attachment, file attribute (for local files) or current timestamp
  • Alternative: DocumentMetadata.CreationDate
• EmbeddedFile/ModDate: same as CreationDate
• EmbeddedFile/MimeType: mimetypes.guess_type(url, strict=False),
  • Alternative1: for remote files the http mime_type may also be used
  • Alternative2: the html type attribute should likely be preferred
• Filespec/AFRelationship: manually defined when creating Attachment or currently 'Source' by default.
  • Alternative: Add an additional rel attribute. According to html spec, several rel types can be given (space seperated)
• Catalog/AF: All attachments are also linked to the whole document:
  • Alternative: Attachment may also be linked to pages or marked content Source This seems more accurate. However, it needs some more implementation work and I currently prefer to have my attachments also show up as attachments in my document viewer in the attachment section.

There is currently no option in this pull request to not include these attributes.

[liZe]

Sorry for the awfully loooooooong time to comment this PR.

We need some time and some real-life testers to improve PDF/A and PDF/UA. We have other related issues we’d like to solve too, we’ll probably take care of all of them at the same time, sometime in the future.

[kesara]

I've tested this by generating a PDF for an RFC and producing PDF/A-3B compliant documents. 🎉

verapdf output:

  <jobs>
    <job>
      <item size="91980">
        <name>/docs/rfc9527.pdf</name>
      </item>
      <validationReport jobEndStatus="normal" profileName="PDF/A-3B validation profile" statement="PDF file is compliant with Validation Profile requirements." isCompliant="true">
        <details passedRules="144" failedRules="0" passedChecks="42171" failedChecks="0"></details>
      </validationReport>
      <duration start="1706726655753" finish="1706726656329">00:00:00.576</duration>
    </job>
  </jobs>

[liZe]

@timoramsauer @kesara Thanks!

I’ve updated the PR to handle PDF/A specificities only when actually generating PDF/A files. I’ve also (hopefully!) followed the different rules for A-1b, A-2b and A-3b (A-4b is not tested yet by VeraPDF). Tests are welcome!

[timoramsauer]

This works for me. Thanls!

[kesara]

@liZe, For example in #2052, I get an error with the latest changes:

$ weasyprint --pdf-identifier foobar  --pdf-variant "pdf/a-3b" foobar.html foobar.pdf
Traceback (most recent call last):
  File "/Users/kesara/lab/WeasyPrint/venv/bin/weasyprint", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/Users/kesara/lab/WeasyPrint/venv/lib/python3.11/site-packages/weasyprint/__main__.py", line 183, in main
    html.write_pdf(output, **options)
  File "/Users/kesara/lab/WeasyPrint/venv/lib/python3.11/site-packages/weasyprint/__init__.py", line 259, in write_pdf
    self.render(font_config, counter_style, **options)
  File "/Users/kesara/lab/WeasyPrint/venv/lib/python3.11/site-packages/weasyprint/document.py", line 391, in write_pdf
    pdf = generate_pdf(self, target, zoom, **options)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kesara/lab/WeasyPrint/venv/lib/python3.11/site-packages/weasyprint/pdf/__init__.py", line 297, in generate_pdf
    variant_function(
  File "/Users/kesara/lab/WeasyPrint/venv/lib/python3.11/site-packages/weasyprint/pdf/pdfa.py", line 59, in pdfa
    relationships = {
                    ^
  File "/Users/kesara/lab/WeasyPrint/venv/lib/python3.11/site-packages/weasyprint/pdf/pdfa.py", line 61, in <dictcomp>
    for attachment in attachments if attachment.md5}
                                     ^^^^^^^^^^^^^^
AttributeError: 'tuple' object has no attribute 'md5'

[liZe]

Everything should be fixed now, thanks for the feedback!

(And don’t hesitate to add a comment if there’s anything wrong.)