Design of Common Ligatures

[Bhikkhu Pesala]

I'm still undecided on how best to design ligatures. I have opted for a design that doesn't alter the letter spacing. Here are some I designed for Verajja, my version of Bitstream Vera. The top row is plain text, the second row is Verajja ligatures, the third row is DejaVu ligatures. Which design do you like best?

Ligatures.png (22.1 KiB) Viewed 20094 times

[William]

I must be missing something, because I decided for which choice to vote yet cannot find out how to cast a vote.

Could you explain what one needs to do to vote please?

William Overington

[Bhikkhu Pesala]

Could you explain what one needs to do to vote please?

You have to send some cookies to Erwin.

[Erwin Denissen]

When you click "View Results", you are no longer able to vote. But when you open the topic again, you should be able to vote. Try this:
viewtopic.php?t=1376

[William]

Thank you both for replying.

I found that I could not find out how to cast a vote.

So, I posted a reply. The system asked me to log in. This I did, then I made my post. I then used the back button of the browser to go back to the index page and refreshed. I then clicked on this thread so as to check that my post had been accepted properly and the system then invited me to vote.

I suspect that one needs to be logged on when one accesses the thread in order to be offered the opportunity to vote. So, if someone is looking at the thread without being logged on, he or she may not be offered the chance to vote, which might well explain why the thread has so many views and only a few votes.

William

[Bhikkhu Pesala]

Mrs Eaves is an example of a font with some interesting ligatures.

[Dick Pape]

I finally got back to reviewing this post and wondered how your lig spacing is going?

It would seem your guideline of not altering letter spacing clearly wasn't used in the Mrs Eaves wonderful example. I think that would be preferable to trying to maintain word width as the gaps are very evident between the i and the n in some of the examples.

This is really a specialized font with nothing but unusual letter combinations, like TR, CC, THE (?), LA, VA and AV, etc. Must have been fun to put together and must be a challenge to use.

[Bhikkhu Pesala]

With a font like Mrs Eaves you definitely need Open Type features to use it. Although one might manage OK with Œ, it would be a pain inserting other ligatures from the Private Use area.

I have added a few more ligatures to my fonts: ft and st, and maintain the same spacing. I wrote macros for Open Office that makes inserting or removing them very quick.

I suspect that almost no-one uses them, but as a font design exercise it was interesting.

[sevry7]

With a font like Mrs Eaves you definitely need Open Type features to use it. Although one might manage OK with Œ, it would be a pain inserting other ligatures from the Private Use area.

As I said a year ago, that seemed the promise of Open Type, which was supposed to be supported in Vista, particularly with regard to these auto-substitution 'enhanced' features. I don't yet have Vista, so I don't know.

But there is a need, perhaps, for an extensive range of special abbreviations and ligature pairs or triples ("ssi" in necessity, for ex). Here's an example of a font I created with Font Creator in order to duplicate the text in a document from the 16th century. For example, the 'long s' has a form that cheats the next letter in close, but also that stands off against a tall letter. That's two separate Unicode cells, neither of which are the letter, "s". So throughout this text, the underlying 'ASCII' is filled with special Unicode callouts and other things in order to get the proper 'slugs' represented (it's easy as I have almost all the special cells using either Shift or Ctrl and the numeric keypad as customized button pad (in a way like some CAD systems)). "S" is used at the end of words, but in two alternate forms. Some are Unicode standard, ct, etc. The 'fi' and so on. But some are 'private area' as well. If the underlying text read as one would expect, and all these special characters were added by the OS, it would certainly be better when it comes time to search text. As it is, I make this possible only by using a series of regexp substitutions and loading the plain text into a separate window.

[William]

> Although one might manage OK with Œ, it would be a pain inserting other ligatures from the Private Use area.

Inserting other ligatures from the Private Use area does, however, in a way simulate the way that someone using a ligature in handsetting some metal type for letterpress printing in olden days would have needed to proceed. The need for a ligature would have needed to be realized and a piece of metal type for the ligature selected from the compartment in the type case which contained that sort.

> Some are Unicode standard, ct, etc. The 'fi' and so on. But some are 'private area' as well.

There is no regular Unicode standard for a ct ligature glyph. Unicode does have code points for the following seven ligature glyphs, at U+FB00 to U+FB06 inclusive.

ff
fi
fl
ffi
ffl
long s t
st

Some years ago I suggested some Private Use Area glyphs for some other ligatures, one of which is a ct ligature.

http://www.users.globalnet.co.uk/~ngo/golden.htm

I have since started to make fonts and have included many of them in some of my fonts. I have also used a few other Private Use Area code points for ligature glyphs in a few of my fonts, such as Chronicle Text, Quest text, Sonnet to a Renaissance Lady and Herb Garden, all of which have threads in the Gallery section of this forum.

> But there is a need, perhaps, for an extensive range of special abbreviations and ligature pairs or triples ("ssi" in necessity, for ex). Here's an example of a font I created with Font Creator in order to duplicate the text in a document from the 16th century.

The golden ligatures colection includes such items as a long s long s i ligature glyph. In the golden ligatures collection, the long s long s i ligature glyph is at U+E757.

There is another system MUFI, the Medieval Unicode Font Initiative, which has Private Use Area codepoints for many items, including some ligatures.

http://www.mufi.info/

In the http://www.mufi.info/specs/MUFI-Alphabetic-2-0.pdf document, at page 80, is LATIN SMALL LIGATURE LONG S LONG S I at U+EBA7.

I have just found LATIN SMALL LIGATURE CT at U+EEC5 in that document. The golden ligatures collection has a ct ligature at U+E707.

Could you possibly say please what mappings you used for the ct ligature and for the other ligatures that you used?

Having just found the ct ligature and various other ligatures which I have used in the MUFI document I am thinking that I should update my fonts so that the MUFI code points are used for the ligatures as well.

This matter of ligatures does get quite interesting as the official Unicode position is that no new code points will be encoded for ligatures, yet there is clearly a need to encode ligatures in plain text for use in systems which do not have OpenType capability to use ligatures in one of the Unicode ways: those ways being, using a ct ligature as an example, use ct in order to leave the matter to the font, use c ZWJ t in order to specify the use of a ligature (assuming that the font and application can together support that request), use c ZWNJ t in order to specify that a ligature should not be used (assuming that the font and application can together support that request).

The version 2.0 MUFI specifications, in pdf format, are available from the following web page.

http://www.mufi.info/specs/

http://www.mufi.info/specs/MUFI-Alphabetic-2-0.pdf
http://www.mufi.info/specs/MUFI-CodeChart-2-0.pdf

William Overington

28 April 2007

[sevry7]

http://www.mufi.info/specs/MUFI-Alphabetic-2-0.pdf

As you said, already:

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

0152 LATIN CAPITAL LIGATURE OE
0153 LATIN SMALL LIGATURE OE
017F LATIN SMALL LETTER LONG S
E259 LATIN CAPITAL LIGATURE OE WITH ACUTE
E25D LATIN CAPITAL LIGATURE OE WITH MACRON
E659 LATIN SMALL LIGATURE OE WITH ACUTE
E65D LATIN SMALL LIGATURE OE WITH MACRON
EBA1 LATIN SMALL LIGATURE LONG S H
EBA2 LATIN SMALL LIGATURE LONG S I
EBA3 LATIN SMALL LIGATURE LONG S L
EBA6 LATIN SMALL LIGATURE LONG S LONG S
EBA7 LATIN SMALL LIGATURE LONG S LONG S I
EBA8 LATIN SMALL LIGATURE LONG S LONG S L
EBC8 LATIN CAPITAL LIGATURE OE WITH DOUBLE ACUTE
EBC9 LATIN SMALL LIGATURE OE WITH DOUBLE ACUTE
EBE4 LATIN CAPITAL LIGATURE OO WITH DIAERESIS
EBE5 LATIN SMALL LIGATURE OO WITH DIAERESIS
EEC5 LATIN SMALL LIGATURE CTEECE LATIN SMALL LIGATURE FFT
EECF LATIN SMALL LIGATURE FFY
EFAD LATIN SMALL LIGATURE OC
EFE9 LATIN SMALL LIGATURE OO WITH ACUTE
F20A LATIN CAPITAL LIGATURE OO

These cover a lot. I left out the AE ligatures. And oddly, I've never seen an "fj".

But you can see even in the short example that there is a ligature for "us", and not shown for "es", "as", etc. in the italic. And the wide long S before a letter that is full height, like "h", or "k", "l", etc, 'stands off', as opposed to when followed by "o", "e", etc. In my font, I didn't attempt to duplicate these as separate ligatures, but rather used a wide long S for the one, and a narrow one for the other. There's double-L (LC). There's even a double s with a leading long s and a trailing, s. And I'm sure I'm leaving some out here, as well.

There are also any number of variations. The ending "s" in the italic is shown using two separate symbols. In the same font, their are different 'g's, different 'z's, and so on. Different cap "W", and sometimes a larger LC "w". Sometimes as leading caps, there are fancier versions. Now I guess these and other variations could all be included as separate fonts, rather than Unicode cells. But it might get confusing.

There's also abbreviations and ligatures in old Greek texts, 100s I think, which I presume made it easier to quickly write the Greek by hand and was transferred to early print.

There's another thing, too. Many of these ligatures, abbreviations and alternatives were dropped by many presses even by the 17th century. Some of the standard "fi", "ff", etc survive to this day. And there's still quite a few. But many were dropped.

To answer one of your questions, no I didn't use the same cells for many of these, except for the few called out at the FB block, at the top here, which you already mentioned.

[William]

Thank you for your reply.

At first, as I read through your post, I thought that you had used the MUFI recommendations in your font.

However, at the end you wrote "To answer one of your questions, no I didn't use the same cells for many of these, except for the few called out at the FB block, at the top here, which you already mentioned.", so it appears that you used some other codepoint allocations? Is that correct?

I have been thinking about these new MUFI recommendations in relation to my golden ligatures collections. MUFI has a different codepoint for a ct ligature. With long s ligatures, the golden ligatures collection has some which are in MUFI and some which are not in MUFI, and MUFI has some which golden ligatures does not have.

For example, the golden ligatures collection has a long s k ligature at U+E754, whereas, as far as I have found at present, MUFI does not have that ligature.

I am hoping to produce a version of my Chronicle Text font with those Private Use Area ligature glyphs and some others from the font that are included in MUFI also mapped to the MUFI recommendations. Also I am hoping to add into the font a few of the ligature glyphs which MUFI has which the golden ligatures collection does not have. As well as ligatures there are also a few other items to double map, such as the precomposed glyphs for M macron, m macron, N macron and n macron and some other ligatures such as pp.

I have come to the conclusion that a central issue is that, as at the present time, with OpenType support not in wide use, that entering transcripts using Private Use Area precomposed glyphs is a good idea, yet that a method is needed so that, in the future, transcriptions to regular Unicode can be made, preferably by an automated method.

There are issues such as how one would decompose a ct ligature. For example, is it as ct or is it as c ZWJ t because one did have a ligature in the first place?

I have been wondering whether those people who are interested could devise a coding, maybe in XML format, maybe in some other format, such that the meaning of Private Use Area ligature glyphs could be added into the font in the Description section of font, so that in the future the font file could, in principle, be accessed by a special purpose application program which could extract that information and use it for such purposes as producing an OpenType version of the font and for decomposing into regular Unicode any text which had been keyed using the font. For example, a text file containing a U+E707 character which had been keyed using my Chronicle Text font could be decomposed automatically by using information from the Description section of the font that U+E707 represents a ct ligature in that font. The coding would allow a default statement with a meaning such as "all Private Use Area encodings not specifically mentioned in the above list are those in the MUFI version 2.0 documentation" so that descriptions need not be included in each font. Thus if someone includes a ligature glyph not in MUFI version 2.0 in a font, the font could have the decomposition information available within the font in a format which some future automated system could understand. The coding system used would best be published and deposited somewhere.

Have you noticed the hypothetical ligatures which are in the Chronicle Text font? These are mapped to U+E716 to U+E71B. These are bi, di, hi, pi, pl, vi respectively.

viewtopic.php?t=1080

William Overington

29 April 2007

[sevry7]

At first, as I read through your post, I thought that you had used the MUFI recommendations in your font.

No, but it might make for a good standard. They do include a lot that I've never seen in old documents. But I'm sure they are found, maybe from German printers.

Since we have the seven at the start of FB, one can take those off the table. The 'high Latin' provides for the long S and both AE and OE small and cap. So those can go.

What remains? In my case, a cell is needed for the stand-off long S (your dupe in E7E0 Chronicle could serve that purpose, though you use the same right margin/right-bearing in each). I see that you have cells in Chronicle for long S B, etc. I thought, again, that simply having a stand-off long S and then a normal long S was a better solution. In yours you need various cells for the separate combinations, which you have in E7xx.

Missing, of course, are the rest of the long S ligatures and triples. While there is ffi, already, there is not yet long S fi, nor f long S i, etc.

MUFI has in EBAx long S h, long S i, long S l, long S S (though I've seen two variations on this), long S S i. Chronicle has these in various places in E7xx.

There is a need for ct. And the example font above suggests some might like an "it" ligature. Chronicle includes many variations on the fs or sf with a third letter. MUFI does not.

MUFI suggests quite a variation in old texts, many of which variations I have never seen. And those that I have seen and wished to duplicate were not part of MUFI, though many are included in your Chronicle.

I have come to the conclusion that a central issue is that, as at the present time, with OpenType support not in wide use, that entering transcripts using Private Use Area precomposed glyphs is a good idea, yet that a method is needed so that, in the future, transcriptions to regular Unicode can be made, preferably by an automated method.

Don't the open type extensions that allow auto-substitution look to Unicode cells, anyway? I haven't set up a font that way. So I don't know.

meaning of Private Use Area ligature glyphs could be added into the font in the Description section of font, so that in the future the font file could, in principle, be accessed by a special purpose application program which could extract that information and use it for such purposes as producing an OpenType version of the font and for decomposing into regular Unicode any text which had been keyed using the font.

I would think some kind of table would be needed, even duplicating any MUFI or other 'standard'. People will surely have unexpected needs, and might need to put various symbols in private area which hadn't been anticipated by others. But I thought the extended open type features were based on table lookup? If that's not done, then one has to hardcode the Unicode right into the text.

Or are you suggesting that the Unicode be hardcoded in that way, but that table could be used by the app to somehow create a virtual text which is perfectly searchable? in addition to using perhaps another table for auto-substitution.

As I said before, what I do is rely on a particular ordering of regexp to run through the text and replace the Unicode with lower Latin letters. But I have to generate a separate document, which is then searchable.

[Bhikkhu Pesala]

There is a gap of 12 empty cells in Alphabetic Presentation Forms after ligature st from FB07 to FB12 — personally I think that is where they belong because that is what they are.

[William]

There is a gap of 12 empty cells in Alphabetic Presentation Forms after ligature st from FB07 to FB12 — personally I think that is where they belong because that is what they are.

This is the reasoning as to why the golden ligatures collection starts at ...7 in U+E707 rather than at ...0 so that the golden ligatures collection follows on from the ...0 to ...6 of the U+FB00 to U+FB06 range.

This issue was discussed at length in the Unicode mailing list about five years ago. Unicode was at that time, and I have no reason to think that that has changed, strongly opposed to adding any more Alphabetic Presentation Forms for things like a ct ligature. It is something to do with the seven items in U+FB00 to U+FB06 only being included so that there can be round trip capability with one of the older character sets which Unicode replaced. It is also something to do with the concept that if new codes were added for ligatures then software packages would need to know how to decompose the ligature into its constituent unligated parts. In the end, the Unicode Technical Committee discussed it at one of their meetings and there was a minute which, as I seem to remember, all but totally closed the door on any more ligatures being given regular Unicode code points.

William