Kerning Figures to Letters

Malcolm Wooden

Is kerning Figures to Letters (and Letters to Figures) something that most designers would add to their kerning pairs these days?

Ramiro Espinoza

I only kern some figures to letters like '7th' and a few others. That's all.

John Hudson

Life's too short (and I say that as someone who lives in a country that uses mixed alphanumeric postal codes).

Hrant Հրանդ Փափազեան Papazian

I don't know about that "most", but it's a good idea, for a number of reasons (pardon the pun).

Andreas Stötzner

Figures to letters: no. But a few instances with 7 and, maybe, 2 ( . , ) … and always assure yourself that you don’t kern any tabular figures, how painful it may ever look 

Hrant Հրանդ Փափազեան Papazian

And 4... and 1...

Mark Simonson

I do it with proportional figures, but not tabular. Not totally sure it's worth it, but it doesn't seem like that much trouble. I also kern proportional figures with each other and with punctuation.

Craig Eliason

Pretty easy to throw 8 in with the right-side-of-B group, zero and six in with the left-side-of-O group, etc. That can take you pretty far.

James Puckett

If Mark does it I’ll do it. I just added another page to my kerning proof.

Ray Larabie

This is what I kern to deal with the most distracting letter/numeral pairs.
7A7JF4T4V4W4Y4.

Vasil Stanev

How heavy would such additional kerning information actually be on a standart Latin Pro font?

Theunis de Jong

Vasil, if your font allows combining the figures with existing letter kerning, as Craig suggests, the impact is extremely minimal (a few more glyphs in the initial lookup). Worst case scenario is a couple more look-ups. Technically, this should not be a problem at all. Creating the kern pairs in the first place is a much bigger job.

Johannes Neumeier

It's tangential, but since I learned it only after the fact: Don't kern Latin figures with RTL scripts' letters (and vice versa). Since the kern pairs would span across scripts those features will never work in a font, even if your font editor might let you pursue such folly.

John Hudson

That’s the case of things now, Johannes. In addition to the scripts being processed in different glyph runs, the directional difference means kerning won't work across the bidi boundary.

However, there’s been a lot of talk about freeing up GPOS to process across the run boundaries used for GSUB, especially in the context of post-linebreak layout, in which a single direction could be presumed for an entire line of text. So at some future time there may be the option for cross-script and cross-direction kerning and other layout features.

Bhikkhu Pesala

Kerning pairs for Uppercase to superscripts are needed, and figures to superscript letters for ordinals. Note references use superscript figures after letters, and may also use superscript letters. 

Michael Rafailyk

What do you think about separating kerning by writing scripts?
Does this make any sense?

• Latin to Latin
• Latin to figures and punctiation (and vice versa)
• Greek to Greek
• Greek to figures and punctiation
• Figures to figures
• etc.

John Hudson

Hi Michael. That’s basically the approach I take, although I tend not to kern letters to numerals (because life is short, and the people who would care about such things are probably using software that allows them to manually kern). I create separate kerning lookups for each script, and a shared lookup for numerals-to-numerals.

At least, that is how I handled it in the Brill types. Some more recent projects use a unified source, in which all the kerning gets rolled into a single set of lookup subtables. I don’t like it, because it is inefficient at the text processing stage, but it is more convenient at the font tool stage.

Michael Rafailyk

Thanks for the answer, John, I think about performance too.

Why did I start thinking about it. I am currently working on a serif face (about 1500 glyphs including Small Caps) and some of the Small Caps characters borrow design elements from lowercase letters, in other words, the shape is close enough to be included in one class. I must say that serifs somewhat unify the shape of characters, and the amount of kerning is reduced. But for the same reason, the number of characters with a similar shape has increased.

As a result, the right-side class "A" was 69 glyphs, "vwy" — 70, and "bceop" — 198. Not sure if this is too large number for the class. Are there any restrictions and will it affect the rendering speed for the old computers?

John Hudson

OTL groups, including kerning classes, can be as big as you like, I think, and should not significantly affect processing speed. I just made some fonts with huge context groups, and they seem to behave fine.

John Savard

Andreas Stötzner said:

and always assure yourself that you don’t kern any tabular figures, how painful it may ever look

Yes. And given that lining figures are far more popular than old style figures, and lining figures are generally also made monospaced to serve for tabular use, this does mean there are few occasions to safely kern letters to digits.

Not that it can't be done, or wouldn't be worthwhile, for some kinds of typefaces, such as a Jenson or an Aldine or a Garalde, where old style figures are the norm.

John Hudson said:

OTL groups, including kerning classes, can be as big as you like, I think, and should not significantly affect processing speed. I just made some fonts with huge context groups, and they seem to behave fine.

On your computer. How many times do I read news items like "Computers with such-and-such peripheral no longer boot after latest Windows Update"... even Microsoft can't seem to test their updates against a sufficiently wide selection of the possible system configurations out there.

Not that I can realistically suggest a MIcrosoft-level testing regimen for every font publisher!