[XeTeX] OT features

[Paul Isambert]

Le 15/02/2011 11:14, Georg A. Duffner a écrit :
> Am 2011-02-15 04:38, schrieb David J. Perry:
>> I had the exact same problem last weekend. I constructed an OT feature
>> to replace a long s with a regular s if followed by a comma, period,
>> space, etc. The longs-space combination did not work. I am not a
>> low-level TeX programming kind of guy, but I had learned a bit somewhere
>> about TeX's "glue" and I wondered if the OT parser did not see the space
>> character. Thanks to Khaled for confirming this.
>>
>> David
>>
>>
>> ----- Original Message ----- From: "Khaled Hosny" 
>> <khaledhosny at eglug.org>
>> To: "Unicode-based TeX for Mac OS X and other platforms" <xetex at tug.org>
>> Sent: Monday, February 14, 2011 6:35 PM
>> Subject: Re: [XeTeX] OT features
>>
>>
>>> On Mon, Feb 14, 2011 at 10:24:14PM +0100, Georg A. Duffner wrote:
>>>> For my font project (see www.georgduffner.at/ebgaramond) i have been
>>>> playing around with some opentype features. It seems, that
>>>> contextual features involving the character "space" don’t work as
>>>> expected.
>>>
>>> AFAIK, XeTeX processes each word in isolation, so space never part of
>>> context (also in TeX space is not a character/glyph but a kind of a
>>> property named glue, so it does not get handled by OT layout).
>>>
>>> Generally speaking, using space in context is fragile and not 
>>> guaranteed
>>> to work everywhere.
>>>
>>>> I wanted to implement a feature for latin texts that
>>>> replaces "u" at the beginning of a word by a letter that looks like
>>>> "v". The rule looks like this:
>>>
>>> 'init' feature should be more suitable here, but almost all OT engine
>>> supports it for few selected scripts (Arabic mainly). With 'init' it is
>>> the responsibility of the engine to detect word beginning.
>>>
>
> Thanks Khaled,
>
> I already feared such implication being the culprit. In this case, the 
> solution was simple since it’s easy to reverse the lookup: replace 
> every "u" by a "v"-like glyph and then make a contextual rule that 
> changes them to a "u"-like one after a letter. But I don’t think I’m 
> gonna keep this as default for latin but rather put it in a stylistic 
> set.

I'm quite new to the internals of OT, but what I've seen implemented in 
similar cases looks more like this: keep "u" as is, and make a 
contextual rule on it with two lookups: the first is triggered whenever 
there's something before the "u", and does nothing; the second is 
triggered in any context and changes "u" to "v". If the first applies, 
the second doesn't. That's the way word-final variants are implemented 
in Minion, except the context of the first lookup is not all glyphs, but 
almost all glyphs, excluding for instance punctuation marks. Thinking 
about it, this should be the case for you too: the first lookup 
shouldn't be triggered by a left parenthesis.

(Sorry if I don't have the terminology right, I've been investigated 
fonts for some weeks only.)

Best,
Paul