Export to PDF breaks Combining Diacritical Marks

[Evgenii]

APub version 1.10.5.1342. I have a project with heavy use of combining marks, like in tañkatā̆ or mātā́. The font supports them and they are rendered correctly within APub itself (Fig. 1). However after the export to PDF the anchoring of the marks is incorrect (Fig. 2). Different fonts produce the same result. Word-exported PDF is rendered correctly. A sample file is attached (fonts which should render it correctly are Crimson Pro, Gentium or Vollkorn, for example), as well as resulted PDF and PDFlib log. This is not a new problem specific to this version (here I reported it 3 years ago). However this time I can't use the work around of exporting fonts as curves (which works), as I need text to be selectable in the output. How do people handle this problem when working with languages with diacritics, like Vietnamese?

UPD: After playing around some more, it seems that the conversion code puts the diacritics before the glyph it is supposed to get attached to instead of after. 

(Fig. 1, APub)

(Fig. 2, PDF)

export-bug.afpub export-bug.pdf

pdflib.log

[EmT]

Hi @Evgenii

This is a known issue, there is not yet a fix available. It has been passed over to our developers for further investigation.   

[BDanker]

I have exactly the same problem exporting Vietnamese text with accents. 

[kenmcd]

3 hours ago, BDanker said:

I have exactly the same problem exporting Vietnamese text with accents. 

What font are you using?

It may be possible to work-around the issue by removing the composites from the font. Or converting a TTF font to an OTF font (which also removes the composites).

And this happens when you Export to PDF?

And does it also happen when you Print?

[BDanker]

Thank you, LibreTaining, for your post. I have tested using Google fonts Merriweather, Manuale, and Montserrat (all TTF) and another TTF named New Century, plus an Adobe OTF font Minion Pro. They all may use composites for accented characters. I can check this in Fontlab.

Yes, this happens when I export to PDF from Publisher but to a lesser degree when I print from Publisher. I assume this will not happen if I use InDesign (I have not checked that out yet). 

[kenmcd]

1 hour ago, BDanker said:

Thank you, LibreTaining, for your post. I have tested using Google fonts Merriweather, Manuale, and Montserrat (all TTF) and another TTF named New Century, plus an Adobe OTF font Minion Pro. They all may use composites for accented characters. I can check this in Fontlab.

Yes, this happens when I export to PDF from Publisher but to a lesser degree when I print from Publisher. I assume this will not happen if I use InDesign (I have not checked that out yet). 

All the TTF fonts have composites, but the Minion Pro OTF does  not.
So I would be very surprised if Minion Pro shows this issue.

Some PostScript printers have problems when printing TTF fonts which have composites - when the printer is on-the-fly converting the TrueType outlines to PostScript outlines they do not handle the composites properly. Some have issues with all composites, some have issues with multiple stacked components, and some only have issues with nested components.
In that case just flatten the composites and the issue goes away.

OpenType-PS fonts, OTF, have no composites - so they do have this printer issue.

But having an issue when Exporting to PDF is a different issue.
If APub is showing an issue with Minion Pro on Export to PDF - that is really strange.

I have all the fonts except New Century.
Could you please provide a test document which shows this issue when you Export to PDF so I can test it myself?
You may not be able to attach files to the forum yet (not enough posts) so you could upload it somewhere such as WorkUpload.com or wherever.

I can also remove the composites from the fonts to test if that fixes it.

[BDanker]

Thank you again for  your willingness to help. I have attached an APub file and an exported PDF file that you can work with. The top three lines are in Montserrat, Merriweather, and Manuale, respectively, all of which are Google fonts. All the lines are in bold type. The font I really need to work correctly is Montserrat. I did not include the New Century font. The fourth line is in Minion Pro. On my computer, in the exported PDF the A-tilde and A-grave do not display correctly in Minion Pro. In the first line (Montserrat) the A-circumflex-acute does not display correctly even in APub (at least this is what I see). Thanks again.

Font Sample.afpub Font Sample.pdf

[kenmcd]

45 minutes ago, BDanker said:

Thank you again for  your willingness to help. I have attached an APub file and an exported PDF file that you can work with. The top three lines are in Montserrat, Merriweather, and Manuale, respectively, all of which are Google fonts. All the lines are in bold type. The font I really need to work correctly is Montserrat. I did not include the New Century font. The fourth line is in Minion Pro. On my computer, in the exported PDF the A-tilde and A-grave do not display correctly in Minion Pro. In the first line (Montserrat) the A-circumflex-acute does not display correctly even in APub (at least this is what I see). Thanks again.

Font Sample.afpub 35.36 kB · 0 downloads Font Sample.pdf 454.13 kB · 0 downloads

How are you typing this text?
The base characters are followed by a separate combining mark.
For example in the MinionPro line the à should be Atilde (U+00C3), but it is instead two characters - the A (U+0041) followed by a combining tilde (U+0303).
APub should be able to combine these (if they have proper anchors), but why not have the correct combined character to begin with and avoid the issue.
The second line is the combined character Atilde (U+00C3).

The other fonts issues are variations of the same issue.
Separate combining accent characters are used instead of the combined character.
In the Minion Pro font there is no anchor on the A to align the separate tilde.

So the work-around is to use the correct single combined characters.

This does not appear to be an issue with composite single characters as I initially thought.

[BDanker]

Thank you for your accurate diagnosis of the problem. I had no idea that the Vietnamese characters with diacriticals were typed incorrectly (I inherited the document that had the problem; I did not type the original text). When I replaced the combining accent characters with the Unicode characters, everything worked as expected. Thank you again for taking up this issue.

[sfriedberg]

For what it's worth, I have a technical document which uses IPA combining diacritics U+031F and U+0320 to place plus and minus signs below various letters.  Using Cambria italic, this seems to export correctly to PDF with AffPub 1.10.5.1342 on Windows 10.  However, this is a simpler case than the original (only one combining diacritic at a time), and I have not formally validated the PDF, only viewed it and printed it.

test.pdf

[kenmcd]

58 minutes ago, sfriedberg said:

For what it's worth, I have a technical document which uses IPA combining diacritics U+031F and U+0320 to place plus and minus signs below various letters.  Using Cambria italic, this seems to export correctly to PDF with AffPub 1.10.5.1342 on Windows 10.  However, this is a simpler case than the original (only one combining diacritic at a time), and I have not formally validated the PDF, only viewed it and printed it.

test.pdf 84.02 kB · 1 download

Cambria has the required anchors on both the base characters and the combining diacritics - so they line up nicely (your doc looks great). Cambria is really well made.
So this does show that APub can properly align the separate combining characters when the required anchors are actually there inside the font.

Minion Pro does not have any of the needed anchors, so I do not expect it to work.

Hmmm ... I just checked Montserrat and it does appear to have the required anchors on Acircumflex (U+00C2) and acutecomb (U+0301), but that is not working. So it appears that combo is not working properly in APub.
Wonder why some work and some do not.
Perhaps that is the "known issue" mentioned above.