Plural rules are looked up by language alone: CLDR's parent of sr-Latn
and zh-Hant is the root, whose only category is other, so sr-latn rows
were rejected. The baked data covers every CLDR locale (10 KB): plural
rules for he, ro, nb, ca, sk, th, ms and the rest, and decimal symbols
for nb-NO, sk-SK, bn-BD, ar-MA and English in Europe. Format applies a
Region's separators only where the system set them, and an empty
grouping separator writes none.
trn_plain writes counts in 0-9 without grouping, for documents and file
names. reset_format goes back to the system region. trn_decimal and
Format write infinities and NaN as CLDR does and take the other form.
Tests that don't set a format use en-US, not the machine's region. The
ICU4X crates are pinned to ~2.3, the minor version the data is for.
Collator::new takes sv_SE.UTF-8 and sr_RS@latin. Folding maps every
script's decimal digits to ASCII and, ignoring accents, Persian keheh,
Farsi yeh, alef maksura, teh marbuta and heh with hamza to the Arabic
letters. The collation data grows 4.8 KB (465 bytes gzipped).
Format writes integers, decimals and percentages for a Region on ICU4X's
decimal formatter: the locale's digits (Arabic-Indic in Egypt, Persian in
Iran) or Latin ones on request, its separators and grouping sizes (Indian
12,34,567), CLDR's percent patterns, and the region's own separators where
the system sets them. I18n::format holds the one trn, file_size and
measure use, the system region's until set_format.
Plural rules move from intl_pluralrules to ICU4X, and trn_decimal picks
the form from the digits shown (1.0 files, 1,5 файла).
The CLDR data (decimal symbols, digits, plural rules, percent patterns)
covers the shipped languages and their regional variants; data/generate.sh
rebuilds it.
`Folder` folds text for case- and accent-insensitive search (Turkish and
Azerbaijani i by locale, ß, final sigma, ligatures, full-width forms) and
maps matches back to the original text for highlighting. `Collator`
(feature `collate`) sorts lists in the locale's order with digits compared
by value, on ICU4X with collation data for the shipped locales only.
The Turkish i and sigma handling come from PdfCraft's `fold_case`.
Part of #198.
Co-authored-by: snipeTR <31975916+snipeTR@users.noreply.github.com>
Co-authored-by: nevicia2010-hub <nevicia2010@gmail.com>
aiongg
