Hyphenation minimun left/right constraints should be language-specific

[Omikhleia]

Issue

The Knuth-Liang hyphenation always defaults to (2, 2) for left hyphen and right hyphen minima, upon initialization:

sile/core/hyphenator-liang.lua

Line 105 in b2cc084

SILE._hyphenators[lang] = { minWord = 5, leftmin = 2, rightmin = 2, trie = {}, exceptions = {} }

These are quite sane defaults for the algorithm... but most languages would beg to differ and use different values...

Typically, for instance, English would likely prefer (2, 3), as Babel (LaTeX) implements it:

https://github.com/latex3/babel/blob/d4d55826cd264220b7a8d92b453748564affea54/locale/en/babel-en-GB.ini#L152-L153

Besides segmentation rules and patterns, SILE should likely implement such "per-language" default preferences:

In Babel some are at (2, 2) (e.g. Finnish), most at (2, 3), some at (1,1), etc.

• Appropriate values could be possibly derived from Babel's ini files for all supported languages....
• Perhaps dubious sometimes, e.g. for Georgian patterns are generated for (1,2) according to their comments, but Babel uses (2,2) regardless; while Typst (see below) seems to use (1,2)...

Workaround

(Not a general solution)

\lua{
-- To do after having switched to English language i.e. the "en" hyphenator got instantiated
SILE._hyphenators['en'].rightmin = 3
}

Further thought

• This was probably overlooked (due to other issues), but (language-specific / custom) left/right hyphen minima were actually mentioned in an existing issue (now closed): Justification for Indic scripts (Malayalam) #308, with rather extreme values in the LaTeX example (3, 5).

• AFAIK, Typst (hypher) seems to implement these right/left minima per languages (in one big file):
  https://github.com/typst/hypher/blob/6b40344866f2d7b2e156db93e91cf105cb75f7a2/src/lang.rs#L201-L205C1.

• While at it, the current Knuth-Plass line breaker use a single hyphenPenalty (probably as TeX does), but we could use variable penalties depending on initial/final segment lengths. That is to say, rather than being behind LaTeX (and/or TeX, which we are here), there would be a way to have improvements.

[Omikhleia]

Linking to #1994 and #1631 -- I do think this should be part of the same "language refactoring".
Perhaps we should have these in a dedicated "project"?

[Omikhleia]

So we might need additional per-language typography tuning files too?
E.g. for French:

{
   lefthyphenmin = 2,
   righthyphenmin = 3,
   identfirst = false,
}

(For the last one, see #1991 (comment))

Notwithstanding the capability to override them if the user wants it.

[alerque]

Yes, this setting should be tunable per language.

And yes the language code related issues are so intertwined they are hard to track and work on. It's hard to sit down and get my head around the problem or know when an individual issue is actionable. Grouping them all in a "project" sounds like a good idea.

[Omikhleia]

Linking to #2001 (comment) - We are not in (2, 2) but likely in (2, 3) by default due to another bug, it seems.