Multilingual font sample

Text samples

Every span in this section was written into this file by pixi run format, from lines that were typed as plain text; the one in the next section is the hand-written exception, and is there to show what the formatter leaves alone. Nothing runs at render time: what the four recipes read is the ordinary Pandoc markup checked in here, which is also what a reviewer reads in the diff.

וַיֹּ֥אמֶר אֱלֹהִים יְהִ֣י א֑וֹר וַֽיְהִי־אֽוֹר׃

καὶ εἶπεν ὁ θεός Γενηθήτω φῶς. καὶ ἐγένετο φῶς.

And God said, Let there be light: and there was light.

神說：「要有光」，就有了光。

Runs inside a sentence are found the same way: 神說 was rendered καὶ εἶπεν ὁ θεός by the Septuagint, from the Hebrew וַיֹּ֥אמֶר אֱלֹהִים.

Overriding the map

A script carries no more than a script: Han is written by Traditional Chinese, Simplified Chinese and Japanese alike, and no rule can tell them apart. Where the language is not the one the map names, write it out. The formatter leaves a span that already carries a lang — and everything inside it — alone:

神说：“要有光。”就有了光。

That is the fourth line above again, written the way Simplified Chinese writes it: one Han character differs, and the rest is punctuation convention. No rule over the script can see either, so left to the map this line would have come out labelled zh-Hant — the right script, the wrong language.

Nothing else in the file marks the span as different from the ones the formatter wrote, and nothing needs to: once formatting has run, its output is markup like any other, editable in place. Noto Sans CJK TC carries the glyphs either way, so the override changes what the document says about the text rather than which face draws it.