Skip exhibit pages with no valid files instead of crashing

[rajeswari1301]

Fixes a bug where a user's document download kept failing. when every page for one file failed to load, it tried to build a PDF out of nothing and crashed the whole download instead of just skipping that file.

Three small fixes at each level where this could happen:

1. ALExhibit.as_pdf() - filters out missing or broken pages before building that exhibit's PDF. If nothing valid is left, logs it and returns None instead of crashing.
2. ALExhibitList.as_pdf() - now skips any exhibit that came back empty from step 1, and returns None if every exhibit ends up empty.
3. ALExhibitDocument.as_pdf() - handles step 2 returning None, so it doesn't crash trying to combine it with the table of contents.

None of these results get cached, so if something comes back on a retry (like it did in our actual case, which was a one-time blip), it's not stuck as skipped for the rest of the session.

Still need to test this against something close to the actual broken case before merging.

[nonprofittechy]

This looks good - wait to merge until you do your tests.

Main question is what happens if a None is able to escape--is it safe all the way up? I would try some adversarial bundles to test this.

Also: I see why this happened - we probably patched ALDocument but not ALExhibitList. Are there other descendants of ALDocument that we need to repair/guard in the same way? It would make sense to do those at the same time.

[Copilot]

Pull request overview

This PR prevents exhibit PDF generation from crashing when an exhibit ends up with zero loadable pages by filtering invalid pages, skipping empty exhibits, and propagating “no output” (None) up to the top-level exhibit-document assembly.

Changes:

• ALExhibit.as_pdf(): filters out missing/broken pages before concatenation; returns None when no valid pages remain.
• ALExhibitList.as_pdf(): skips exhibits that return None; returns None when all exhibits are empty.
• ALExhibitDocument.as_pdf(): handles an empty exhibit list PDF result to avoid crashing when combining with the table of contents; return types updated to Optional[DAFile].

Suppressed comments (3)

docassemble/AssemblyLine/al_document.py:3118

• ALExhibitList.as_pdf() now returns Optional[DAFile], but the docstring doesn't mention that it can return None when there are no valid exhibits to include. Please document the None case explicitly.

        """
        Compiles all exhibits in the list into a single PDF.

docassemble/AssemblyLine/al_document.py:3476

• as_docx() delegates to as_pdf(), which can now return None. The docstring should mention the None possibility so callers don't assume a DAFile is always returned.

        """
        Despite the name, renders the document as a PDF. Provided for signature compatibility.

docassemble/AssemblyLine/al_document.py:3415

• ALExhibitDocument.as_pdf() now returns Optional[DAFile], but the docstring doesn't mention that it can return None (e.g., when there are no exhibits and no table of contents to return). Please document the None case.

        """
        Render the document as a PDF.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[nonprofittechy]

Copilot's findings all also look worth addressing prior to merging, esp. updating the docstring. ideally we could also scan for now-unsafe use in our other repos

[rajeswari1301]

This looks good - wait to merge until you do your tests.

Main question is what happens if a None is able to escape--is it safe all the way up? I would try some adversarial bundles to test this.

Also: I see why this happened - we probably patched ALDocument but not ALExhibitList. Are there other descendants of ALDocument that we need to repair/guard in the same way? It would make sense to do those at the same time.

Good question, no not fully yet. Found a few more spots in ALDocumentBundle that assume as_pdf/as_docx never returns None, as_zip, as_pdf_list, as_docx_list, get_cacheable_documents, and a couple branches inside as_pdf itself. I'll add those to this PR too