Ligature display

[roger64]

Hi

I decided to try some headings in lower case with a fancy font. I am facing a ligature display problem with the most common of them in French, fi and fl. (see screenshot)

Up to now, this is what happens:
- Sigil displays the text, ADE does it too, but with a standard fi or fl (screenshot)
- Kobo just displays a blank space...
- Prince (PDF) is the stricter of them: it gives me a dire warning and crashes the file. Though it is obvious now, it took me a while to find the right culprit after reading this text:

*** Mercury runtime: caught segmentation violation ***
cause: address not mapped to object
address involved: 0x7a51000
This may have been caused by a stack overflow, due to unbounded recursion.
exiting from signal handler

My question is: which kind of character should I write in the xhtml file (Sigil) within the EPUB to get a fine ligature display or at least to obtain what I did from Sigil and ADE, and avoid a crash?

[Doitsu]

As you probably know, ligatures need to be defined in the font as glyphs and in layout tables. AFAIK, ADE supports the fi and fl ligatures if the font supports it. Some fonts also support the use of ZWNJ and ZWJ between two letters to enable/disable optional ligatures. (You could try inserting a ZWJ between f and i to test if that makes any difference.)
Also, to exclude font issues as the cause of your problems, you may want to repeat your tests with a commercial font.

[roger64]

@Doitsu

Thank you for your reply. Today I learnt something.

It's indeed a commercial font.
I included ‍ in the ligatures to see what happens:
- Sigil is impressed and displays a nice ligature.
- Kobo is not and still leaves a blank...
- Prince stops whining and now accepts to print the file. However, it would be very OK if I did not see this:

Same result using: &-#8205;

[exaltedwombat]

You're asking more of epub that it's able to deliver, particularly if you want portability between different readers.

[Jellby]

If you want ligatures:

1. Don't change anything in the text, just write f + i, as usual.
2. If you are embedding and subsetting a font, make sure the ligatures and their definitions are not removed.
3. Hope that the reading software supports ligatures.

[roger64]

@Jellby

Yes I am embedding the font and I realize now that the subset is probably crippled as far as ligatures are concerned. Thanks all for your advices. I will keep this font only for dropcaps and upppercase to keep Kobo happy...

[ShellShock]

Kobo will display ligatures aslong as the font has the necessary glyphs--many do not, but Charis SIL does.

[yucca]

It’s not clear how you have tried to use ligatures, and which ligatures using which font. Primarily, ligatures are used via OpenType features in fonts that support them, and I would not expect e-book readers to be able to deal with this (you would use font-feature-settings in CSS). Inserting a zero-width joiner is a character-level request for applying ligature behavior when available, but probably ignored by most e-book software.

There is however a clumsy way, using ligature characters such as U+FB01 LATIN SMALL LIGATURE FI, which you can enter in XHTML as fi (or as such, character fi). There is a small set of such characters, but they work reliably in modern software provided that the font being used contains that character, in its correct position. (Many fonts contain such ligatures but possibly in wrong code positions.)

[roger64]

Originally Posted by yucca

There is however a clumsy way, using ligature characters such as U+FB01 LATIN SMALL LIGATURE FI, which you can enter in XHTML as fi (or as such, character fi). There is a small set of such characters, but they work reliably in modern software provided that the font being used contains that character, in its correct position. (Many fonts contain such ligatures but possibly in wrong code positions.)

@yucca

Thanks for your explanations. I tried this "clumsy way", forcing the use of the required glyph (if present) using fi and it does indeed work for my fancy font with Sigil, ADE 1.7 and... yes, Kobo Glo!

I understand that it is probably not advisable to use it for body text, since on most modern readers, the ability of changing fonts is a common feature and that some fonts may not have these ligatures. So the missing glyphs would probably be replaced by blanks or ? flags...

But for my limited purpose (lowercase title in a non-replaceable font) it seems to work fine. Where could I find a list of these ligatures?

I can think also of another use. Sometimes, in a book we find a small quote (only one... ) in, say, Greeks characters. I realize that these characters could also be summoned this way (screenshot). If they are missing in the font, provided we wrote something like:
body (font-family:"My_special_font"; serif), they would also probably be displayed with the serif replacement font.

[yucca]

Originally Posted by roger64

Where could I find a list of these ligatures?

There’s just a handful of ligatures of Latin letters defined as characters. Not counting ij and oe “ligatures” (which are really separate letters, though of ligature origin), the list (extracted from the Unicode character database) is

U+FB00 LATIN SMALL LIGATURE FF
U+FB01 LATIN SMALL LIGATURE FI
U+FB02 LATIN SMALL LIGATURE FL
U+FB03 LATIN SMALL LIGATURE FFI
U+FB04 LATIN SMALL LIGATURE FFL
U+FB05 LATIN SMALL LIGATURE LONG S
U+FB06 LATIN SMALL LIGATURE ST

Sometimes, in a book we find a small quote (only one... ) in, say, Greeks characters. I realize that these characters could also be summoned this way (screenshot). If they are missing in the font, provided we wrote something like:
body (font-family:"My_special_font"; serif), they would also probably be displayed with the serif replacement font.

Well, it should syntactically be body {font-family:"My_special_font", serif}. But there is no guarantee that the generic name serif maps to a font that contains the Greek characters you need, especially if the text is polytonic Greek. And you can enter the Greek characters as such (Sigil has no special tool for this, so you may need to use e.g. a Greek keyboard layout or use copy and paste), as named character references, or a numeric character references, as shown in the screenshot. But the problem is: will e-book readers be able to display them?

That’s really a different question, and an interesting one. You might need to use an embedded font to make sure that Greek characters will show OK.

[roger64]

Thanks for your handy list and your advices.

If I spoke about this random Greek quote, it's because I had a book with two short Greek words. I used embedded fonts in this book (subsets of) and did not want to add a special Greek font only for this purpose.

The plain: ,serif coupled with the ISO codes (html codes are OK too) did the trick (at least for Sigil, calibre, ADE, Kobo...). This seems to work.

[Turtle91]

Instead of embedding a font for short/limited quotes in an alternate language, you could just insert it as an image. It wouldn't change if the reader changed font size, and it's not elegant, but if it is a stand-alone quote it should work well.

[Tex2002ans]

Originally Posted by yucca

Well, it should syntactically be body {font-family:"My_special_font", serif}. But there is no guarantee that the generic name serif maps to a font that contains the Greek characters you need, especially if the text is polytonic Greek. And you can enter the Greek characters as such (Sigil has no special tool for this, so you may need to use e.g. a Greek keyboard layout or use copy and paste), as named character references, or a numeric character references, as shown in the screenshot.

I personally use this site to easily search for unicode characters (then I copy/paste them):

http://www.fileformat.info/info/unic...preview=entity

Or for the basic Greek characters I use Wikipedia:

https://en.wikipedia.org/wiki/Greek_letters

I recently stumbled upon this incredible Wikipedia page for Greek letters with accents (look further down the page for organized tables of characters):

https://en.wikipedia.org/wiki/Greek_diacritics

Takes a little while to copy/paste one by one, but if there are only a few that are needed, it is not that bad.

Originally Posted by yucca

But the problem is: will e-book readers be able to display them?

I did not have any luck getting these obscure Greek characters to display on my Nook (embedded font or no). I also heard it had zero luck displaying on the Nook App on Android (I assume because they do not have proper Unicode support?).

It read perfectly fine on ADE on the PC, Sigil, Mantano Reader (Android)... I assume because these have proper Unicode support.

Originally Posted by yucca

That’s really a different question, and an interesting one. You might need to use an embedded font to make sure that Greek characters will show OK.

Hmmm, perhaps someone with a little more embedded font experience might be able to help in my case.

I have attached the EPUB (no embedded font). All Greek words are marked with the CSS class "greek" to allow very easy font tweaking.

Would be great to figure out this Greek situation, would really be helpful for many books.

[Jellby]

Originally Posted by Tex2002ans

Takes a little while to copy/paste one by one, but if there are only a few that are needed, it is not that bad.

Why not use the character map to copy-paste whatever you want?

[yucca]

Originally Posted by Tex2002ans

I did not have any luck getting these obscure Greek characters to display on my Nook (embedded font or no). I also heard it had zero luck displaying on the Nook App on Android (I assume because they do not have proper Unicode support?).

I suspect it is a font issue rather than Unicode support problem. The readers probably interpret the character data properly, they just fail to display it. E.g., Adobe Digital Editions (on Win 7) shows the Greek characters in your epub OK, except that they are too slanted, but this is probably a feature of the font it is using. (Greek letters have often been designed in italic style.)

It read perfectly fine on ADE on the PC, Sigil, Mantano Reader (Android)... I assume because these have proper Unicode support.

Reader for PC shows simple Greek letters OK but shows the polytonic letters as question marks. This is typically a font problem, since many fonts lack the polytonic, or “Greek Extended”, characters (as they are mostly used in scholarly works only when presenting classical Greek text).

Hmmm, perhaps someone with a little more embedded font experience might be able to help in my case.

It seems that I have been too optimistic about font embedding. It works fine in web browsers nowadays, and I had tested the technique in Android Reader, too, and was happy about it. But neither Adobe Digital Editions nor Reader for PC seems to support font embedding.

I have attached the EPUB (no embedded font). All Greek words are marked with the CSS class "greek" to allow very easy font tweaking.

Yes, that makes font tweaking very easy (if you want just the Greek words in an embedded font): after other @ rules, you would add e.g.

@font-face {
font-family: FreeSerif;
src: url(/Fonts/FreeSerif.otf;
}

.greek {
font-family: FreeSerif;
}

and you would just need to add the FreeSerif.otf font in the Fonts folder. (I probably would not use FreeSerif, but it was easy to test with it.)

But this does not help much in some important readers, which seem to lack support to font embedding.

(In practice, if the technique worked widely enough, I would probably use a suitable free font for all text, since Greek text should be style-compatible with text in Latin letters. And then I would probably need to embed an italic and a bold and possibly even an italic bold typeface, too, making the file rather large, but probably not excessively large.)

As far as Sigil is considered, things are fine. It lets you add the code for font embedding (though it has no special tools for it, you just write CSS), and it supports font embedding when it shows the book in Book View.