'liga' should not support only latin words #697
Description
'liga' works in latin words well. But for non Latin characters, it's not work.
For example, https://learn.microsoft.com/en-us/typography/opentype/otspec183/gsub#example-7-contextsubstformat1-subtable-and-substlookuprecord
NewFont-Regular.ttf.zip
Expected Behavior
Current Behavior
import opentype from 'opentype.js';
import { readFileSync } from 'fs';
import path from 'path';
const { parse } = opentype;
const loadSync = (url, opt) => parse(readFileSync(url), opt);
let newFont = loadSync('./fonts/NewFont-Regular.ttf');
console.log(newFont.stringToGlyphIndexes(' – ')); // [ 1, 88, 1 ] wrong, should be [86,88,86]
console.log(newFont.stringToGlyphIndexes('ABA')); // [ 25, 3, 25 ] right
Possible Solution
script latn;
language dflt;
lookup SUB_11 {
sub A by X;
sub space by thinspace;
} SUB_11;
These two rules are very similar, the only different is 'latin characters'. I think the problem lies in https://github.com/opentypejs/opentype.js/blob/master/src/features/latn/contextCheck/latinWord.js
function latinWordStartCheck(contextParams) {
const char = contextParams.current;
const prevChar = contextParams.get(-1);
return (
// ? latin first char
(prevChar === null && isLatinChar(char)) ||
// ? latin char preceded with a non latin char
(!isLatinChar(prevChar) && isLatinChar(char))
);
}
function latinWordEndCheck(contextParams) {
const nextChar = contextParams.get(1);
return (
// ? last latin char
(nextChar === null) ||
// ? next char is not latin
(!isLatinChar(nextChar))
);
}
I think we mixed up ‘script’ and ‘language’.
-
'script 'is more like a preference.
https://learn.microsoft.com/en-us/typography/script-development/standard#features
If you choosed 'latn', Equivalent to you choosed ['ccmp','liga','clig'] by default. You can also manually choose to close or open certain tags if you want. When the program queries these tags first or returns ' latn ', but cannot find them, it needs to return' DFLT ' -
'Language' is more like optimization for certain characters.
Selective processing based on Unicode. And I think we needn't support it now.
