| Inductiveload | |
| Feb 17 2021, 10:48 AM |
| F34109497: 2021-02-17_104517_584x90_screenshot.png | |
| Feb 17 2021, 10:48 AM |
| F34109481: 2021-02-17_104351_640x401_screenshot.png | |
| Feb 17 2021, 10:48 AM |
| F34109492: Domicile_Ordinance.pdf | |
| Feb 17 2021, 10:48 AM |
Downloading any text at enWS or zhWS (and presumably the other Wikisources) with CJK characters in it causes "boxes of doom" in the PDF export:
This works OK in EPUB and TXT exports:
Example: https://en.wikisource.org/wiki/Domicile_Ordinance
And the PDF produced from https://ws-export.wmcloud.org/?page=Domicile_Ordinance&lang=en&format=pdf:
It looks like we just don't have the right fonts installed. I just installed fonts-noto-cjk on the test site, and this seems to work: https://ws-export-test.wmcloud.org/?title=Domicile_Ordinance&lang=en&format=pdf-a4&fonts=noto-sans-cjk-sc (I'm not quite sure what the different CJK fonts are for).
@Samwilson that makes sense.
The Noto fonts are:
AFAIK they contain the same glyphs but with different variants:
e.g. https://commons.wikimedia.org/wiki/File:Source_Han_Sans_Version_Difference.svg
Thanks, that makes sense.
I've installed the above fonts package on production as well now, so the fix for this issue is to include fonts=noto-sans-cjk-sc or whichever.
Makes me wonder if we want to add a per-work way of setting a default font for export, because even once we've got a per-wiki way of doing that (T274561) it won't help works like this one.
it won't help works like this one.
Even if the font isn't default, at least for PDF, it'll still fall back to the any font that supports the glyphs. For example, https://ws-export.wmcloud.org/?page=Domicile_Ordinance&lang=en&format=pdf is now working.
I can't immediately tell *which* fallback font it's using, pdffont shows:
name type encoding emb sub uni object ID ------------------------------------ ----------------- ---------------- --- --- --- --------- LiberationSerif-Bold CID TrueType Identity-H yes no yes 27 0 LiberationSerif CID TrueType Identity-H yes no yes 28 0 [none] Type 3 Custom yes no yes 38 0 [none] Type 3 Custom yes no yes 39 0 [none] Type 3 Custom yes no yes 40 0 [none] Type 3 Custom yes no yes 41 0 [none] Type 3 Custom yes no yes 42 0 [none] Type 3 Custom yes no yes 43 0 [none] Type 3 Custom yes no yes 44 0 [none] Type 3 Custom yes no yes 45 0 LiberationSerif-Italic CID TrueType Identity-H yes no yes 46 0 [none] Type 3 Custom yes no yes 206 0 [none] Type 3 Custom yes no yes 207 0 [none] Type 3 Custom yes no yes 208 0 [none] Type 3 Custom yes no yes 204 0 [none] Type 3 Custom yes no yes 205 0
But I do now see CJK chars in the PDF.
For EPUB, even if the font isn't bundled, the reader itself can do the fallback, and I think most readers come with a decent set of fonts.
@Samwilson I would normally bring this up in estimation today, but since we'll be having the earlier meeting that you can't attend, I'm just pinging you directly. What is the status of this ticket? It seems a fix is in progress, so I just want to make sure this doesn't fall through the cracks. Maybe we can put it on the board if you have already done some work on it? Thanks!
Fantastic; thanks for the update, @Samwilson! In that case, I'll mark this ticket as Resolved.