Custom Fonts Copy & Paste Error

[Jonathan-Ganss]

Hi @gettalong,

We recently switched to HexaPDF to support embedding XML files (PDF/A compliance), and most things are working great. However, we’ve run into an issue with custom fonts where copying text from the generated PDF results in unreadable / gibberish output.

This doesn’t happen when using one of the 14 standard fonts (e.g., Helvetica, Times), and does not occur with the same custom fonts in PDFs generated by other tools.

Here’s a summary of our setup:

composer = HexaPDF::Composer.new(page_size: :A4, margin: [85, 32.5])
composer.document.task(:pdfa, level: '3u')

composer.document.config['font.map'] = {
'MuseoSans' => {
none: File.join('assets', 'fonts', 'MuseoSans300.ttf'),
bold_italic: File.join('assets', 'fonts', 'MuseoSans500.ttf'),
bold: File.join('assets', 'fonts', 'MuseoSans700.ttf')
},
'DejaVuSans' => {
none: File.join('assets', 'fonts', 'DejaVuSans.ttf')
}
}
composer.document.config['font.fallback'] = ['DejaVuSans']

What I’ve already tried:

• Setting subset: false when loading fonts (both via config and manually)
• Verified that /ToUnicode maps are present using both MuPDF and Poppler
• Confirmed the fonts themselves are valid (copy/paste works in PDFs generated by other tools)

Despite this, any text using a custom font still results in mostly unreadable characters when copied from the PDF.

Could this be related to how the Unicode CMap is being generated in HexaPDF?
Any tips or guidance on resolving this would be very appreciated!

Thanks in advance 🙏

[gettalong]

Which version of HexaPDF are you using? If not the latest, could you try with the latest?

Could you provide a minimal example script with all necessary resources so that I can reproduce the problem?

[Jonathan-Ganss]

I'm using the latest version of HexaPDF, so 1.2.0.
Furthermore, we're using Ruby 3.3.2 in our production environment and when I just tested locally in a new Rails project, I used Ruby 3.4.2.

This is the example file I used:

---

require 'hexapdf'

class App
  def create_pdf
    composer.text("Hello from HexaPDF. This is a test to verify the copy & paste issues with custom fonts!", font: 'MuseoSans')
    
    pdf_content = StringIO.new
    composer.write(pdf_content, validate: false, optimize: true)
    
    File.write("example.pdf", pdf_content.string)
  end

  def composer
    @composer ||= begin
      composer = HexaPDF::Composer.new(page_size: :A4, margin: [85, 32.5])
      composer.document.task(:pdfa, level: '3u')
      composer.document.config['font.map'] = {
        'MuseoSans' => {
          none: File.join('app', 'assets', 'MuseoSans-100.ttf')
        }
      }

      composer
    end
  end
end

App.new.create_pdf

---

Apparently, it's not possible to attach a TTF file here, but you can download the font I used from this GitHub repo:
https://github.com/enlacee/pbi-peru/blob/master/src/assets/fonts/museo-sans/MuseoSans-100/MuseoSans-100.ttf

When you open the saved pdf and copy & paste the text to a browser/notes file etc., you should see the gibberish text.
Thanks for your help!

[gettalong]

Thanks for providing everything!

Alas, I cannot reproduce the problem (example.pdf):

$ hexapdf ins example.pdf psd 1
save_graphics_state
  set_font_and_size /F1 10
  begin_text
    move_text 32.5 749.389764
    text> Hello from HexaPDF. This is a test to verify the copy & paste issues with custom fonts!
  end_text
restore_graphics_state

This shows the text just fine and I tested in Okular and other PDF viewers as well.

[Jonathan-Ganss]

Interesting 🤔
I’m actually seeing the same issue with your PDF that I’ve encountered with my own.

When I save the PDF and open it in a browser, copy & paste works correctly. However, when I save the PDF to my desktop and open it using the default viewer on my Mac, the pasted text turns into gibberish. I get something like:
"Hello from HexaPDF .Thi hi a seis so terhfv sTe yocv p caise hii&ei uhsT y&isom fowsin"

I asked a few colleagues to test this as well, and we’re all experiencing the same behavior.
We're all using Macs, so if you can't reproduce the issue, it might be related to the default PDF viewer on macOS...

[Jonathan-Ganss]

This appears to be working! I tested it with a colleague, and we were both able to copy individual words as well as the entire sentence without any issues.

[Jonathan-Ganss]

Thanks a lot for your help!

[gettalong]

Great! And thanks for testing it out!

Alas, this is not the best news. Leaving out that part of the code makes HexaPDF slower by 10-15% in the raw_text benchmark. I will see if I can find a variant that works for Preview and doesn't lead to worse performance.

[gettalong]

@Jonathan-Ganss Thanks for being patient!

I was running out of ideas to test and asked on StackOverflow (see https://stackoverflow.com/questions/79560847/problem-with-tounicode-map-only-on-macos-preview). Turns out that Apple Preview is the culprit.

With the information from the answer I was able to modify HexaPDF to create a file that still doesn't need escaping when serializing and works with macOS Preview: example.pdf. Could you please also test that file? Thanks!

[Jonathan-Ganss]

Seems to be a very complicated and profound problem 😅
But yes, this new file also works perfectly for me. Thank you!

[gettalong]

Thanks for testing it out! I will properly add the changes and then release a new HexaPDF soon.

[gettalong]

Okay, the fix is in. One rare use case still won't work on Preview. Will release a new version in the coming week.