PDF slow load

[alpha-mipl]

I need your help!
When I try to load a PDF with version 20.5.0-alpha.1, it loads in under 1.02 seconds, but with the latest version after 20, it takes up to 3 seconds. Even the basic PDF.js viewer can load this PDF in under 1 second.

Demo PDF file
Here you can check the PDF

demo.pdf

[alpha-mipl]

I found that it works fine up to version 21.0.0-alpha.6, but after that, PDFs load slowly.

[alpha-mipl]

I found the issue
that in viewer-4.7.728.mjs, there are two setDocument methods:

One at line 35,004
Another at line 30,450
Both setDocument methods are being called when loading a PDF.

If I add a return in the top of the second setDocument method, the PDF loads much faster and works perfectly.

[stephanrauh]

Hm. As far as I can tell, you've disabled the setDocument method of the PdfScriptingManager:
https://github.com/stephanrauh/pdf.js/blob/15ec795de8ad7172ef6a815a24203b2336ca00f6/web/pdf_scripting_manager.js#L83

Can you confirm that?

I've add the return statement, but I don't see the difference you're describing. Loading the document still takes 3-4 seconds. Subjectively, I suspect it's faster with the return statement, but it's a far cry from the single second you're expecting. In other word: at the moment, I can't reproduce your solution, so I'm a bit puzzled.

BTW: I guess it's obvious that deactivating the scripting manager isn't the solution. It may be a good performance optimization for those who don't need JavaScript in PDF files, so adding an option to disable scripting makes a lot of sense to me. However, you describe it as a regression, so I'd like to find the root cause.

[alpha-mipl]

I noticed that when I added a return statement, the PDF loading became much faster. However, when I remove this return statement, I face slow PDF loading issues again.
The issue seems to be that during PDF loading, page 1 is being called twice, which appears to be blocking the main thread.

[alpha-mipl]

However, after adding the return statement, the PDFThumbnailViewer stopped working. I think this is because the setDocument function is being used for the thumbnails as well.

[alpha-mipl]

I tested this with different PDFs and noticed a significant performance difference of about 10 seconds. For example, I have another PDF that's 350MB in size - it usually takes 20.59 seconds to load, but with the return statement added, it only takes 10 seconds.

I think the slow performance is happening because setDocument is being called simultaneously for both the thumbnail and the main PDF.