[XeTeX] Bug or design flaw in Polyglossia 2012/04/29 v1.2.1

[Ian-Mathew Hornburg]

2013/3/30 Khaled Hosny <khaledhosny at eglug.org>:
> OpenType does not allow this, if the font has the character they will be
> used (font developer may, say, map them to blank glyphs in feature with
> mlym tag, though that makes no sense, but XeTeX wouldn’t still report
> missing characters; as far as the engine is concerned the font has those
> character).

Ditto to Khaled and Arthur. And polyglossia doesn’t have the
functionality to flip between fonts for single characters like Latin
numbers or periods that’re used in many scripts and languages.

Polyglossia only reselects fonts that you predefine when you
explicitly use a command that changes the language being used. Rachana
contains numbers and some basic punctuation, so that’s what is
rendered. If they weren’t in the font, you’d just see boxes.

The kind of complex font-fallback that you’re describing isn’t even
truly implemented in operating-system software like fontconfig. You’re
gonna have to manually switch between languages or fonts, or code up
some sort of workaround for what you’re wanting.

When I’m typesetting Indic scripts whose font-internal Latin
punctuation I don’t like, I’ll use something like:
    \textmalayalam{<Malayalam text>}?

where document-wide language is something else, and the question-mark
glyph is taken from the default Latin fault. Conversely, you could do
something like:
    <Malayalam text>\textenglish{?}

everytime punctuation like that comes up, but I’ll admit, that’s
really cumbersome.