Kerning issues after merging pages

[pferreir]

Hello. First of all, congratulations for this library. I'm impressed at how accurate the DOM layout implementation is!

I am using weasyprint in a web app and trying to produce a multi-page PDF. The only way I found to do that (short of using some separate lib to manually join weasyprint-produced PDFs) was a code fragment I found in your docs:

all_pages = [p for doc in documents for p in doc.pages]
documents[0].copy(all_pages).write_pdf('combined.pdf')

This works fine, except that then I get very serious kerning/space issues in the resulting PDF. Some letters end up really too far apart and break words apart.

Is this a known issue and what can I do to avoid it?

Thanks!

[liZe]

Hello!

First of all, congratulations for this library. I'm impressed at how accurate the DOM layout implementation is!

💛

This works fine, except that then I get very serious kerning/space issues in the resulting PDF.

This problem really looks like #1489. Do you use the latest 53.4 version?

[pferreir]

You are absolutely right! Apologies, I should have upgraded to the latest version. I'll close the ticket.

[SchrodingersGat]

@liZe I have found a regression here - documented below. Please LMK if you want me to start a new issue rather than resurrecting this thread.

This issue has returned in v56.1 - I have confirmed by switching between 56.1 and 53.4 (without any other changes).

53.4	56.1

And the code used to merge these PDF files together:

pages = []

for output in outputs:
    doc = output.get_document()
    for page in doc.pages:
        pages.append(page)

pdf = outputs[0].get_document().copy(pages).write_pdf()

Some further testing reveals that:

• 55.0 : BAD
• 54.3: GOOD

[liZe]

That’s maybe related to #1662.

[liZe]

That’s maybe related to #1662.

Yes, that’s definitely a duplicate of #1662. Mixing pages from different documents (as in your issue) or adding elements into the tree after the rendering (as in #1662) are features that are not explicitly prevented by WeasyPrint, but they were not supposed to work and to be supported 😀.

Let’s continue the discussion in #1662.