Font won't Subset

Karellen · 04-18-2023, 08:33 PM

I have an ebook which uses Han characters, but were not being displayed.

I searched for a Han font and ended up finding and using this one and now it displays correctly on the libra2... https://github.com/adobe-fonts/sourc...SC-Regular.otf

Its quite a big file and I only needed 11 characters.

When I try to run Subset Embedded Font nothing happens...

Code:

Project Hail

Subsetting embedded fonts

Reduced total font size to 0.0% of original

Is there something with this font file that prevents subsetting?

DNSB · 04-18-2023, 09:20 PM

I checked the font using Sigil and it's SubsetFont plugin and it seemed to work without issue. The 12 characters I used gave me an output font file around 48KB down from 23088KB so about 0.2%.

The calibre editor showed a massive size reduction but the save button was greyed out which would seem to indicate that no changes were made.

Karellen · 04-18-2023, 10:06 PM

Thanks for testing @DNSB.

Sounds like you had some reduction in Calibre, whereas nothing happens for me.

I tried restarting in Debug Mode, but looks like that mode doesn't extend to the Editor.

kovidgoyal · 04-18-2023, 10:32 PM

To debug a subsetting issue i need both the full book since subsetting is a function both o fthe font and text in the book.

DNSB · 04-18-2023, 10:55 PM

I've attached the epub I generated with the .otf font from Karellen's download link embedded that I used for testing.

DNSB · 04-18-2023, 10:57 PM

Quote:

Originally Posted by Karellen

Thanks for testing @DNSB.

Sounds like you had some reduction in Calibre, whereas nothing happens for me.

I tried restarting in Debug Mode, but looks like that mode doesn't extend to the Editor.

calibre seemed to show that the font was being subsetted but did not make any changes in the font file and would not save. I added and removed a space in the text to allow me to save it but the font file had no change in it's size.

Karellen · 04-19-2023, 12:56 AM

Book sent via PM

Quoth · 04-19-2023, 05:06 AM

Some font files not only won't subset, but even cause conversion to fail if "subset fonts" is selected. I solve this by picking a similar appearance font and check that it has any extended non-Latin-Roman characters needed.

I think this is a more common problem with some non-alphabetic fonts, but I've had it years ago on a font with just regular Latin-Roman characters plus áéíóú ÁÉÍÓÚ. It was a font obtained maybe in 1994 on Windows 3.x, so hardly surprising.

kovidgoyal · 04-19-2023, 08:15 AM

That's a CID keyed font subsetting of these is not supported. I will fix the code to emit a better error message about such fonts.

Karellen · 04-19-2023, 07:06 PM

Thank you for checking @kovidgoyal

Too difficult to implement?

Thanks @DNSB for checking also

kovidgoyal · 04-19-2023, 09:51 PM

CID keyed fonts are not common enough for me to bother, but patches are most welcome.

Karellen · 04-19-2023, 10:30 PM

Quote:

Originally Posted by kovidgoyal

CID keyed fonts are not common enough for me to bother, but patches are most welcome.

Ok, thank you.

Martinoptic · 04-20-2023, 04:06 AM

From a quick web search you might be able to convert you cid keyed font with font forge?

Karellen · 04-20-2023, 04:37 AM

Quote:

Originally Posted by Martinoptic

From a quick web search you might be able to convert you cid keyed font with font forge?

I installed font forge, but its quite a learning curve to figure it out, especially as I am a dummy with the intricacies of font design, so didn't understand what all the settings and options meant. There is an option to Generate Fonts to OTF or TTF, but the conversion failed with "Self Intersecting, Wrong Direction, Missing Points at Extrema" and some others, but I have no idea what they mean.

I thought it would be a simple matter of deleting the characters I do not want, but even that seemed difficult, especially as there must be a thousand characters in the file, and doing it one by one is too time consuming to attempt. A simple "invert selection" would have made it easier.

Quoth · 04-20-2023, 11:58 AM

I think, but I'm not sure, that CID is maybe partly hierarchical to save storage for complex glyphs sharing sub-elements. It's not impossible that you could build a regular font table using Font Forge, but the way CID fonts work is so different that I doubt it. Hinting, for example, can apply to a row of symbols/glyphs.

Regular fonts, especially alphabetic or related, have a unique sequential entry for each character. The CID font concept seems to have been designed around elements of a distinct glyph. Some OCR for Asian languages even build a custom CID font for that document.

I can't see how you can even substitute a regular font instead of a CID font in an existing document, because they work so differently. This is also why a CID font won't subset in the same way as a regular font.

04-18-2023, 08:33 PM	#1
Karellen Wizard Posts: 1,611 Karma: 9500498 Join Date: Sep 2021 Location: Australia Device: Kobo Libra 2	Font won't Subset I have an ebook which uses Han characters, but were not being displayed. I searched for a Han font and ended up finding and using this one and now it displays correctly on the libra2... https://github.com/adobe-fonts/sourc...SC-Regular.otf Its quite a big file and I only needed 11 characters. When I try to run Subset Embedded Font nothing happens... Code: Project Hail Subsetting embedded fonts Reduced total font size to 0.0% of original Is there something with this font file that prevents subsetting?

04-19-2023, 05:06 AM	#8
Quoth Still reading Posts: 14,025 Karma: 105092227 Join Date: Jun 2017 Location: Ireland Device: All 4 Kinds: epub eink, Kindle, android eink, NxtPaper	Some font files not only won't subset, but even cause conversion to fail if "subset fonts" is selected. I solve this by picking a similar appearance font and check that it has any extended non-Latin-Roman characters needed. I think this is a more common problem with some non-alphabetic fonts, but I've had it years ago on a font with just regular Latin-Roman characters plus áéíóú ÁÉÍÓÚ. It was a font obtained maybe in 1994 on Windows 3.x, so hardly surprising. Last edited by Quoth; 04-19-2023 at 05:11 AM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Font won't embed	NIRC	Conversion	2	06-19-2020 08:38 PM
Hexadecimal support for calibre-debug --subset-font ?	un_pogaz	Calibre	9	11-23-2018 03:00 AM
FontShrinker - tool to subset a font	Toxaris	Workshop	91	02-08-2018 06:54 PM
Embedded font-subset sometimes fails	GrannyGrump	Sigil	3	10-20-2012 09:47 AM
HELP! Courier Font Won't Work	TedPark	Sony Reader	0	07-06-2008 07:34 PM

04-18-2023, 09:20 PM	#2
DNSB Bibliophagist Posts: 46,183 Karma: 168983734 Join Date: Jul 2010 Location: Vancouver Device: Kobo Sage, Libra Colour, Lenovo M8 FHD, Paperwhite 4, Tolino epos	I checked the font using Sigil and it's SubsetFont plugin and it seemed to work without issue. The 12 characters I used gave me an output font file around 48KB down from 23088KB so about 0.2%. The calibre editor showed a massive size reduction but the save button was greyed out which would seem to indicate that no changes were made.

04-18-2023, 10:06 PM	#3
Karellen Wizard Posts: 1,611 Karma: 9500498 Join Date: Sep 2021 Location: Australia Device: Kobo Libra 2	Thanks for testing @DNSB. Sounds like you had some reduction in Calibre, whereas nothing happens for me. I tried restarting in Debug Mode, but looks like that mode doesn't extend to the Editor.

04-18-2023, 10:32 PM	#4
kovidgoyal creator of calibre Posts: 45,345 Karma: 27182818 Join Date: Oct 2006 Location: Mumbai, India Device: Various	To debug a subsetting issue i need both the full book since subsetting is a function both o fthe font and text in the book.

04-19-2023, 12:56 AM	#7
Karellen Wizard Posts: 1,611 Karma: 9500498 Join Date: Sep 2021 Location: Australia Device: Kobo Libra 2	Book sent via PM

04-19-2023, 08:15 AM	#9
kovidgoyal creator of calibre Posts: 45,345 Karma: 27182818 Join Date: Oct 2006 Location: Mumbai, India Device: Various	That's a CID keyed font subsetting of these is not supported. I will fix the code to emit a better error message about such fonts.

04-19-2023, 07:06 PM	#10
Karellen Wizard Posts: 1,611 Karma: 9500498 Join Date: Sep 2021 Location: Australia Device: Kobo Libra 2	Thank you for checking @kovidgoyal Too difficult to implement? Thanks @DNSB for checking also

04-19-2023, 09:51 PM	#11
kovidgoyal creator of calibre Posts: 45,345 Karma: 27182818 Join Date: Oct 2006 Location: Mumbai, India Device: Various	CID keyed fonts are not common enough for me to bother, but patches are most welcome.

04-20-2023, 04:06 AM	#13
Martinoptic Bibliophist Posts: 7,020 Karma: 7173892 Join Date: Dec 2021 Location: England Device: none	From a quick web search you might be able to convert you cid keyed font with font forge?

04-20-2023, 11:58 AM	#15
Quoth Still reading Posts: 14,025 Karma: 105092227 Join Date: Jun 2017 Location: Ireland Device: All 4 Kinds: epub eink, Kindle, android eink, NxtPaper	I think, but I'm not sure, that CID is maybe partly hierarchical to save storage for complex glyphs sharing sub-elements. It's not impossible that you could build a regular font table using Font Forge, but the way CID fonts work is so different that I doubt it. Hinting, for example, can apply to a row of symbols/glyphs. Regular fonts, especially alphabetic or related, have a unique sequential entry for each character. The CID font concept seems to have been designed around elements of a distinct glyph. Some OCR for Asian languages even build a custom CID font for that document. I can't see how you can even substitute a regular font instead of a CID font in an existing document, because they work so differently. This is also why a CID font won't subset in the same way as a regular font.