collective
Browse / github.com/storytold/craft-libs / patch
Add Hebrew locale normalization, BiDi RTL/LTR helpers, and Hebrew font manifest build support
Add Hebrew locale normalization, BiDi RTL/LTR helpers, and Hebrew font manifest build support ## Summary Adds shared RTL, mixed RTL/LTR (BiDi), and Hebrew font catalog support across `craft-i18n`, `craft-fonts-build`, and `craft-ui`: ### `craft-i18n` - **Legacy ISO 639-1 tag normalization**: Normalizes deprecated primary language subtags (`iw` -> `he`, `ji` -> `yi`, `in` -> `id`) in `scripts()`/`direction()` and `I18n::match_tag()`, so `iw_IL.UTF-8` and `iw-IL` resolve to `he` (`Script::Hebr`, `Direction::Rtl`). - **BiDi mixed RTL/LTR helpers**: - `auto_direction(text, fallback)`: resolves paragraph direction from the first strong character outside isolates (UAX #9 P2/P3 via `base_
- Status
- open
- Score
- 0.00
- Adopters
- 0 other operators
- Author
- hrb9 (imported from pull request #342, open)
- Base
9ff719499d8f4730b93071abed993e5f12fe1be8- Size
- +204 −12 in 7 files
- Review
- pending
- Shared
- 2026-10-28
- Id
d329e40fe6777774c05367e3
How it went
No other operator has reported using it yet.
Diff
raw · crates/fonts-build/src/embed.rs crates/fonts-build/src/lib.rs crates/fonts-build/src/manifest.rs crates/fonts-build/src/tests.rs crates/i18n/src/lib.rs crates/i18n/src/scripts.rs crates/ui/src/text.rs
@@ -9,7 +9,7 @@ use std::io; use std::path::{Path, PathBuf}; use crate::cmap::Cmap;-use crate::manifest::{ManifestEntry, ParseError, parse_manifest, script_matches};+use crate::manifest::{DEFAULT_MANIFEST, HEBREW_MANIFEST, LATIN_MANIFEST, ManifestEntry, ParseError, parse_manifest, script_matches}; const DIR_VAR: &str = "CRAFT_FONTS_DIR"; const REQUIRED_VAR: &str = "CRAFT_FONTS_REQUIRED";@@ -25,6 +25,7 @@ type Filter = Box<dyn Fn(&ManifestEntry) -> bool>; /// Which faces to embed, and from where. [`Embed::write`] does the work. #[derive(Default)] pub struct Embed {+ manifest: Option<PathBuf>, scripts: Option<Vec<String>>, filter: Option<Filter>, dir: Option<PathBuf>,@@ -72,6 +73,37 @@ impl Embed { Self { families: Some(families.iter().map(|f| (*f).to_owned()).collect()), ..Self::default() } } + /// Read an opt-in manifest file relative to the `craft-fonts` checkout instead of+ /// [`DEFAULT_MANIFEST`] (`"fonts/manifest.txt"`), such as [`HEBREW_MANIFEST`] or+ /// [`LATIN_MANIFEST`].+ pub fn manifest(mut self, rel_path: impl Into<PathBuf>) -> Self {+ self.manifest = Some(rel_path.into());+ self+ }++ /// Every face in the opt-in Hebrew catalog ([`HEBREW_MANIFEST`], `"fonts/hebrew-manifest.txt"`).+ /// Narrow it with [`Embed::filter`] or select specific families with+ /// [`Embed::hebrew_documents`].+ pub fn hebrew() -> Self {+ Self::new().manifest(HEBREW_MANIFEST)+ }++ /// Every face of these families from the opt-in Hebrew catalog ([`HEBREW_MANIFEST`]).+ pub fn hebrew_documents(families: &[&str]) -> Self {+ Self::documents(families).manifest(HEBREW_MANIFEST)+ }++ /// Every face in the opt-in Latin presentation catalog ([`LATIN_MANIFEST`],+ /// `"fonts/latin-manifest.txt"`).+ pub fn latin() -> Self {+ Self::new().manifest(LATIN_MANIFEST)+ }++ /// Every face of these families from the opt-in Latin presentation catalog ([`LATIN_MANIFEST`]).+ pub fn latin_documents(families: &[&str]) -> Self {+ Self::documents(families).manifest(LATIN_MANIFEST)+ }+ /// Embed only faces with at least one of these script tags (see /// [`ManifestEntry::has_script`]). With `CRAFT_FONTS_REQUIRED=1`, each tag must match a face. pub fn scripts(mut self, tags: &[&str]) -> Self {@@ -233,9 +265,14 @@ impl Embed { Ok(faces) } + /// Path to the selected manifest file inside `dir`.+ fn manifest_path(&self, dir: &Path) -> PathBuf {+ dir.join(self.manifest.as_deref().unwrap_or_else(|| Path::new(DEFAULT_MANIFEST)))+ }+ /// The wanted manifest entries with their canonical file paths. fn select(&self, dir: &Path, cargo: &mut Vec<String>) -> Result<Vec<Face>, Error> {- let manifest = dir.join("fonts/manifest.txt");+ let manifest = self.manifest_path(dir); let text = std::fs::read_to_string(&manifest).map_err(|source| Error::Read { path: manifest.clone(), source })?; // Cargo treats a missing watched file as changed, which would rerun the build every time. cargo.push(format!("cargo::rerun-if-changed={}", manifest.display()));@@ -364,7 +401,7 @@ impl Embed { if self.elsewhere.is_empty() { return Vec::new(); }- let manifest = dir.join("fonts/manifest.txt");+ let manifest = self.manifest_path(dir); // `select` read and parsed this manifest already. let entries = std::fs::read_to_string(&manifest).ok().and_then(|text| parse_manifest(&text).ok()).unwrap_or_default(); for family in self.elsewhere.iter().filter(|f| !entries.iter().any(|e| &e.family == *f)) {@@ -420,7 +457,7 @@ impl Embed { } fn check_required(&self, dir: &Path, faces: Vec<Face>) -> Result<Vec<Face>, Error> {- let manifest = dir.join("fonts/manifest.txt");+ let manifest = self.manifest_path(dir); if faces.is_empty() { return Err(Error::NoFaces { manifest }); }@@ -63,7 +63,7 @@ mod subset; mod tests; pub use embed::{Embed, Error, target_is_wasm};-pub use manifest::{CJK_SCRIPTS, ManifestEntry, ParseError, UI_ROLE, parse_manifest};+pub use manifest::{CJK_SCRIPTS, DEFAULT_MANIFEST, HEBREW_MANIFEST, LATIN_MANIFEST, ManifestEntry, ParseError, UI_ROLE, parse_manifest}; /// A craft-fonts face embedded by [`Embed`]: one line of `fonts/manifest.txt`. ///@@ -60,6 +60,15 @@ impl ManifestEntry { /// The `role` of an interface face. pub const UI_ROLE: &str = "ui"; +/// Default manifest file relative to a `craft-fonts` checkout.+pub const DEFAULT_MANIFEST: &str = "fonts/manifest.txt";++/// Opt-in Latin presentation manifest relative to a `craft-fonts` checkout.+pub const LATIN_MANIFEST: &str = "fonts/latin-manifest.txt";++/// Opt-in Hebrew presentation and document manifest relative to a `craft-fonts` checkout.+pub const HEBREW_MANIFEST: &str = "fonts/hebrew-manifest.txt";+ /// Script tags of the CJK faces, which craft-fonts also ships as interface faces. `Hant` covers /// `Hant-HK`. pub const CJK_SCRIPTS: &[&str] = &["Hans", "Hant", "Jpan", "Kore", "Hani"];@@ -123,7 +123,12 @@ fn rejects_a_checkout_under_a_non_utf8_path() { use std::os::unix::ffi::OsStrExt; let scratch = Scratch::new("non-utf8"); let dir = scratch.0.join(std::ffi::OsStr::from_bytes(b"craft-fonts-\xff"));- std::fs::create_dir_all(dir.join("fonts")).unwrap();+ // Filesystems such as macOS APFS reject non-UTF-8 paths at creation time (EILSEQ).+ match std::fs::create_dir_all(dir.join("fonts")) {+ Ok(()) => {}+ Err(e) if matches!(e.raw_os_error(), Some(84 | 92)) || e.kind() == std::io::ErrorKind::InvalidInput => return,+ Err(e) => panic!("creating non-UTF-8 test directory failed: {e}"),+ } std::fs::copy(fixture().join("fonts/sans-arabic/SansArabic.ttf"), dir.join("fonts/a.ttf")).unwrap(); std::fs::write(dir.join("fonts/manifest.txt"), "A | Regular | fonts/a.ttf | Arab | OFL-1.1 | OFL.txt | 00 | src\n").unwrap(); let r = run(Embed::new(), Some(&dir), Some("1"));@@ -253,6 +258,29 @@ fn integrity_is_base64_of_the_hex_hash() { } } +/// Opt-in Hebrew and Latin manifests embed their selected faces independently of the default manifest.+#[test]+fn embeds_from_opt_in_hebrew_and_latin_manifests() {+ let scratch = Scratch::new("hebrew-manifest");+ let dir = scratch.0.join("craft-fonts");+ std::fs::create_dir_all(dir.join("fonts")).unwrap();+ std::fs::copy(fixture().join("fonts/sans-arabic/SansArabic.ttf"), dir.join("fonts/heebo.ttf")).unwrap();+ std::fs::copy(fixture().join("fonts/sans-arabic/SansArabic.ttf"), dir.join("fonts/david.ttf")).unwrap();+ std::fs::write(+ dir.join("fonts/hebrew-manifest.txt"),+ "Heebo | Variable | fonts/heebo.ttf | Hebr,Latn | OFL-1.1 | fonts/OFL.txt | 00 | src\n\+ David Libre | Regular | fonts/david.ttf | Hebr,Latn | OFL-1.1 | fonts/OFL.txt | 00 | src\n",+ )+ .unwrap();+ std::fs::write(dir.join("fonts/latin-manifest.txt"), "Latin Face | Regular | fonts/david.ttf | Latn | OFL-1.1 | fonts/OFL.txt | 00 | src\n")+ .unwrap();+ let r = run(Embed::hebrew(), Some(&dir), Some("1"));+ assert_eq!(r.result.unwrap(), ["Heebo Variable", "David Libre Regular"]);+ let r = run(Embed::hebrew_documents(&["David Libre"]), Some(&dir), Some("1"));+ assert_eq!(r.result.unwrap(), ["David Libre Regular"]);+ assert_eq!(run(Embed::latin(), Some(&dir), Some("1")).result.unwrap(), ["Latin Face Regular"]);+}+ /// A font with only a `cmap` table: a Windows format 4 subtable with `bmp` segments `(first, /// last, glyphs)`, mapped to glyphs 1.. when `glyphs` is `None` and through the glyph array /// otherwise, and a format 12 subtable with `groups` `(first, last)` when there are any.@@ -55,7 +55,7 @@ pub use fold::{Folded, Folder}; pub use format::{Digits, Format}; pub use number::{NumberError, parse_number}; pub use region::{HourCycle, Measurement, Paper, Region};-pub use scripts::{Direction, Han, Script, base_direction, direction, han, scripts};+pub use scripts::{Direction, Han, Script, auto_direction, base_direction, direction, han, scripts}; pub use size::{FileSize, SizeUnit}; /// The catalogs an app embeds, one entry per locale: (locale, shared catalog, the app's catalog).@@ -456,6 +456,14 @@ pub fn ltr(value: &str) -> String { wrap(LRI, value) } +/// `value` as a right-to-left isolate (RLI…PDI), for right-to-left phrases that start with a+/// number, Latin acronym or neutral character (`123 שלום`, `PDF בעברית`) so a first-strong+/// isolate ([`isolate`]) would otherwise treat them as left to right. [`fmt`] takes the isolate+/// off again in text with no right-to-left letter and no bidi control.+pub fn rtl_isolate(value: &str) -> String {+ wrap(RLI, value)+}+ /// `text` without isolate marks (LRI, RLI, FSI, PDI): a [`fmt`] or [`I18n::trn`] result made fit /// for a file name, a log or a comparison. pub fn strip_isolates(text: &str) -> String {@@ -584,9 +592,12 @@ fn candidates(tag: &str) -> Vec<String> { let tag = tag.trim().to_ascii_lowercase(); let base = tag.split(['.', '@']).next().unwrap_or("").replace('_', "-"); let mut parts: Vec<&str> = base.split('-').filter(|p| !p.is_empty()).collect();- let Some(&primary) = parts.first() else {+ let Some(first) = parts.first_mut() else { return Vec::new(); };+ let original = *first;+ *first = canonical_language(first);+ let primary = *first; let modifier = match tag.split_once('@').map(|(_, m)| m) { Some("latin") => Some("latn"), Some("cyrillic") => Some("cyrl"),@@ -606,7 +617,20 @@ fn candidates(tag: &str) -> Vec<String> { out.extend(candidates(&parts.join("-"))); return out; }- let mut out: Vec<String> = (1..=parts.len()).rev().map(|n| parts.get(..n).unwrap_or_default().join("-")).collect();+ let mut out = Vec::new();+ for n in (1..=parts.len()).rev() {+ out.push(parts.get(..n).unwrap_or_default().join("-"));+ if original != primary {+ let mut raw = parts.get(..n).unwrap_or_default().to_vec();+ if let Some(first) = raw.first_mut() {+ *first = original;+ }+ let raw = raw.join("-");+ if !out.contains(&raw) {+ out.push(raw);+ }+ }+ } if let Some(s) = script && usual_script(primary).is_some_and(|usual| usual != s) {@@ -652,6 +676,17 @@ fn usual_script(language: &str) -> Option<&'static str> { }) } +/// Canonicalize deprecated ISO 639-1 primary language subtags (`iw` → `he`, `ji` → `yi`,+/// `in` → `id`).+pub(crate) fn canonical_language(language: &str) -> &str {+ match language {+ "iw" => "he",+ "ji" => "yi",+ "in" => "id",+ other => other,+ }+}+ /// A BCP 47 region subtag: two letters or three digits. fn is_region(subtag: &str) -> bool { (subtag.len() == 2 && subtag.bytes().all(|b| b.is_ascii_alphabetic())) || (subtag.len() == 3 && subtag.bytes().all(|b| b.is_ascii_digit()))@@ -1212,8 +1247,30 @@ mod tests { #[test] fn more_right_to_left() {- let i = all(&[("ps", "", ""), ("ckb", "", ""), ("yi", "", ""), ("ug", "", "")]).0;+ let i = all(&[+ ("he", "@id\tlanguage.name\tעברית\n@plural\t{n} page|{n} pages\tעמוד {n}|{n} עמודים|{n} עמודים\n", ""),+ ("ps", "", ""),+ ("ckb", "", ""),+ ("yi", "", ""),+ ("ug", "", ""),+ ])+ .0; assert!(i.languages().into_iter().skip(1).all(|l| i.rtl(l)));+ let he = lang(&i, "he");+ assert_eq!(i.name(he), "עברית");+ assert_eq!(i.match_tag("he_IL.UTF-8"), Some(he));+ assert_eq!(i.match_tag("iw_IL.UTF-8"), Some(he));+ assert_eq!(i.match_tag("iw-IL"), Some(he));+ assert_eq!(candidates("iw-IL"), ["he-il", "iw-il", "he", "iw"]);+ assert_eq!(i.trn(he, 1, "{n} page", "{n} pages"), "עמוד \u{2066}1\u{2069}");+ assert_eq!(i.trn(he, 2, "{n} page", "{n} pages"), "\u{2066}2\u{2069} עמודים");+ assert_eq!(i.trn(he, 10, "{n} page", "{n} pages"), "\u{2066}10\u{2069} עמודים");+ assert_eq!(i.trn_plain(he, 10, "{n} page", "{n} pages"), "10 עמודים");+ let mixed = fmt(+ "שמור את {file} בתיקייה {dir} ({note})",+ &[("file", "Report 2.pdf"), ("dir", <r("/Users/dana/docs")), ("note", &rtl_isolate("PDF בעברית"))],+ );+ assert_eq!(mixed, "שמור את \u{2068}Report 2.pdf\u{2069} בתיקייה \u{2066}/Users/dana/docs\u{2069} (\u{2067}PDF בעברית\u{2069})"); } #[test]@@ -173,6 +173,13 @@ pub fn base_direction(text: &str) -> Option<Direction> { } } +/// The resolved direction of a paragraph or mixed-direction snippet: its [`base_direction`] when+/// it contains a strong character outside isolates, or `fallback` (such as the interface or+/// document direction) when it contains only neutral characters or numbers.+pub fn auto_direction(text: &str, fallback: Direction) -> Direction {+ base_direction(text).unwrap_or(fallback)+}+ /// The forms a language tag's Han characters take: `zh-CN`, `zh-SG`, `zh-Hans` and plain `zh` → /// Simplified; `zh-TW`, `zh-Hant` and the other regions CLDR writes in Traditional characters /// (`zh-US`, `zh-AU`) → Traditional; `zh-HK`, `zh-MO`, `zh-Hant-HK` and `yue` → HongKong; `ja` →@@ -265,7 +272,7 @@ impl Subtags { let tag = tag.trim().to_ascii_lowercase(); let base = tag.split(['.', '@']).next().unwrap_or("").replace('_', "-"); let mut parts = base.split('-').filter(|p| !p.is_empty());- let language = parts.next().unwrap_or("").to_owned();+ let language = super::canonical_language(parts.next().unwrap_or("")).to_owned(); let mut script = match tag.split_once('@').map(|(_, m)| m) { Some("latin") => Some("latn".to_owned()), Some("cyrillic") => Some("cyrl".to_owned()),@@ -301,6 +308,11 @@ mod tests { assert_eq!(base_direction("123\nשלום"), None); assert_eq!(base_direction("1920 × 1080"), None); assert_eq!(base_direction(""), None);++ assert_eq!(auto_direction("שלום Hello", Direction::Ltr), Direction::Rtl);+ assert_eq!(auto_direction("Hello שלום", Direction::Rtl), Direction::Ltr);+ assert_eq!(auto_direction("1920 × 1080", Direction::Rtl), Direction::Rtl);+ assert_eq!(auto_direction("1920 × 1080", Direction::Ltr), Direction::Ltr); } /// Every locale the craft-i18n catalogs or an app ship, as the apps spell them.@@ -319,6 +331,7 @@ mod tests { ("et", &[Latn], None), ("fi", &[Latn], None), ("fr", &[Latn], None),+ ("he", &[Hebr], None), ("hr", &[Latn], None), ("hu", &[Latn], None), ("id", &[Latn], None),@@ -351,9 +364,15 @@ mod tests { ] { assert_eq!(scripts(tag), want, "{tag}"); assert_eq!(han(tag), han_forms, "{tag}");- let rtl = tag == "ar";+ let rtl = matches!(tag, "ar" | "he"); assert_eq!(direction(tag), if rtl { Direction::Rtl } else { Direction::Ltr }, "{tag}"); }+ // Legacy ISO 639-1 aliases normalize to their BCP 47 primary subtags.+ assert_eq!(scripts("iw_IL.UTF-8"), &[Hebr]);+ assert_eq!(direction("iw-IL"), Direction::Rtl);+ assert_eq!(scripts("ji"), &[Hebr]);+ assert_eq!(direction("ji"), Direction::Rtl);+ assert_eq!(scripts("in-ID"), &[Latn]); } #[test]@@ -123,6 +123,27 @@ pub fn align_galley_rows(mut galley: Arc<Galley>, width: f32, mut align: impl Fn AlignedGalley::Aligned { galley, paint_offset, reading_rect } } +/// The start-edge alignment for `row`'s paragraph direction: [`Align::Max`] (right) for a+/// right-to-left paragraph (`row.rtl_paragraph`), or [`Align::Min`] (left) for a left-to-right+/// paragraph.+pub fn row_start_align(row: &PlacedRow) -> Align {+ if row.rtl_paragraph { Align::Max } else { Align::Min }+}++/// The end-edge alignment for `row`'s paragraph direction: [`Align::Min`] (left) for a+/// right-to-left paragraph (`row.rtl_paragraph`), or [`Align::Max`] (right) for a left-to-right+/// paragraph.+pub fn row_end_align(row: &PlacedRow) -> Align {+ if row.rtl_paragraph { Align::Min } else { Align::Max }+}++/// Align each row inside `width` to the start edge of its paragraph ([`row_start_align`]):+/// right-to-left paragraphs flush with the right edge (`Align::Max`), and left-to-right+/// paragraphs flush with the left edge (`Align::Min`).+pub fn align_galley_rows_by_direction(galley: Arc<Galley>, width: f32) -> AlignedGalley {+ align_galley_rows(galley, width, row_start_align)+}+ #[cfg(test)] mod tests { use super::*;@@ -512,6 +533,27 @@ mod tests { assert_eq!(before, Arc::as_ptr(parts(&result).0)); } + #[test]+ /// Mixed RTL (Hebrew) and LTR (English) paragraphs align to their respective start edges within `width`.+ fn align_galley_rows_by_direction_aligns_mixed_rtl_and_ltr_paragraphs() {+ let original = layout_text("שלום עולם\nHello world\nשורה שלישית Hello", 1.0, Align::Min);+ assert_eq!(original.rows.len(), 3);+ assert!(original.rows[0].rtl_paragraph);+ assert!(!original.rows[1].rtl_paragraph);+ assert!(original.rows[2].rtl_paragraph);+ assert_eq!(row_start_align(&original.rows[0]), Align::Max);+ assert_eq!(row_start_align(&original.rows[1]), Align::Min);+ assert_eq!(row_end_align(&original.rows[0]), Align::Min);+ assert_eq!(row_end_align(&original.rows[1]), Align::Max);++ let result = align_galley_rows_by_direction(original, 300.0);+ let (aligned, offset, _) = parts(&result);+ assert_eq!(offset, Vec2::ZERO);+ assert!((aligned.rows[0].rect().right() - 300.0).abs() < 1.0, "RTL paragraph must align to right edge");+ assert_eq!(aligned.rows[1].pos.x, 0.0, "LTR paragraph must align to left edge");+ assert!((aligned.rows[2].rect().right() - 300.0).abs() < 1.0, "mixed paragraph starting with Hebrew must align to right edge");+ }+ const MIXED: &str = "First extraordinarily long reading row\nSecond independently aligned reading row\nThird right-aligned overwide reading row"; const ALIGNMENTS: [Align; 3] = [Align::Min, Align::Center, Align::Max];