Fix for "Expected instance of _PDFDict, but got instance of PDFNull"

[ImmaB]

What are you working on?

I have a pdf from a client, that I unfortunately can't share. Every pdf viewer I tried could open it but in pdf-lib I get the error "Expected instance of _PDFDict, but got instance of PDFNull". I found that in pdf-lib the page tree was missing while here for example it was not missing.

For those that have the same (apparently very rear) issue: My solution was to build and insert the page tree myself:

function fixMissingPageTree(doc: PDFDocument) {
    const pagesName = PDFName.of('Pages');
    const pages = doc.catalog.lookupMaybe(pagesName, PDFDict);
    if (pages) return;
    const indirectObjects = doc.context['indirectObjects'] as Map<PDFRef, PDFObject>;
    const pageLeafs = indirectObjects.entries()
        .filter(([_, obj]) => obj instanceof PDFPageLeaf)
        .toArray() as [PDFRef, PDFPageLeaf][];
    const structParentsName = PDFName.of('StructParents');
    const getIndex = (pageLeaf: PDFPageLeaf) => pageLeaf.lookup(structParentsName, PDFNumber)?.asNumber();
    pageLeafs.sort((a, b) => getIndex(a[1]) - getIndex(b[1]));
    const kids = PDFArray.withContext(doc.context);
    for (const pageLeaf of pageLeafs)
        kids.push(pageLeaf[0]);
    const pageTree = PDFPageTree.withContext(doc.context);
    pageTree.set(PDFName.of('Type'), pagesName);
    pageTree.set(PDFName.of('Kids'), kids);
    pageTree.set(PDFName.of('Count'), PDFNumber.of(pageLeafs.length));
    doc.catalog.set(pagesName, pageTree);
}

Additional Notes

No response

[Sharcoux]

Maybe you could try a PR to solve the problem more globally?

[adnsistemas]

I haven't looked into the code, but seems to be a problem with the way the page tree dict is represented in the PDF.
I know you can't share the PDF, but if you share PDF Dicts that makes up the page tree, I can look into it.
Blindly looking for the problem is not appealing at all..

[sedev01]

@adnsistemas I got a similar pdf which generate a very similar issue. Expected instance of PDFDict, but got instance of undefined

Let me know if it would be better to open a separate issue. This is a gouv issued pdf.

l1-fill-24e.pdf

Updated : turns out it's unrelated

[adnsistemas]

Thanks, I'll look into it and see if they seem to be the same problem or not.

[sedev01]

@adnsistemas I managed to open/use/modify the pdf using a fork of PDFSecurity. I'll post once I have more details. I feel my post is unrelated to original issue.

[adnsistemas]

@sedev01 The PDF you provided is encrypted, I guess that's the problem.
In respect to the original problem, I think the problem lies in the 'Parent' property of page leaves, not being properly registered. But without something to test it, or someone having the problem to test. There is not point in making changes.