Merge pull request #2670 from dotCooCoo/fix/joiners-stay-with-their-run

Type: a zero width non-joiner is shaped in the face of the letters around it

crates/text/src/lib.rs

@@ -557,6 +557,8 @@ mod tests_fit;
 #[cfg(test)]
 mod tests_inline;
 #[cfg(test)]
+mod tests_joiners;
+#[cfg(test)]
 mod tests_justify;
 #[cfg(test)]
 mod tests_scripts;

crates/text/src/shape.rs

@@ -392,7 +392,8 @@ pub(crate) fn shape_range(
             let next_script = shaping_script(c);
             let strong_script = !matches!(next_script, Script::Common | Script::Inherited);
             let script_change = strong_script && script != Script::Common && script != next_script;
-            // Combining marks stay with their base.
+            // Combining marks, zero width joiners and non-joiners, and variation selectors stay
+            // with the character before them.
             if (face.id() != seg.face.id() || small != seg.small || level != seg.level || script_change) && !is_mark(c) {
                 if i > seg.range.start {
                     seg.range.end = i;
@@ -580,7 +581,7 @@ struct Shaped {
 fn is_mark(c: char) -> bool {
     use unicode_general_category::{GeneralCategory, get_general_category};
     matches!(get_general_category(c), GeneralCategory::NonspacingMark | GeneralCategory::SpacingMark | GeneralCategory::EnclosingMark)
-        || matches!(c, '\u{200D}' | '\u{FE00}'..='\u{FE0F}')
+        || matches!(c, '\u{200C}' | '\u{200D}' | '\u{FE00}'..='\u{FE0F}')
 }
 
 /// Shape `seg` and append its glyphs to `out`. Each glyph takes the tracking, kerning, baseline

crates/text/src/tests_joiners.rs

@@ -0,0 +1,93 @@
+//! A zero width joiner or non-joiner between characters that the style's font lacks is shaped with
+//! those characters, in their fallback face. The Unicode Standard describes both joiners under
+//! Cursive Connection and Ligatures
+//! (https://www.unicode.org/versions/latest/core-spec/chapter-23/), and Microsoft's OpenType shaping
+//! specifications for Devanagari and Bengali include them in a syllable after a halant
+//! (https://learn.microsoft.com/en-us/typography/script-development/devanagari,
+//! https://learn.microsoft.com/en-us/typography/script-development/bengali).
+//! No font is committed: the tests use installed faces and print SKIPPED when none is found.
+use super::*;
+use kurbo::Point;
+use std::sync::Arc;
+use vectorcraft_doc::CharStyle;
+
+const ZWNJ: char = '\u{200C}';
+const ZWJ: char = '\u{200D}';
+
+/// A Latin face that maps both joiners but not `c`, and the face that the fallback gives for `c`.
+/// With `tags`, the fallback face's GSUB script list must contain one of those script tags. Which
+/// face the fallback gives depends on the faces loaded so far, so the tests compare with that face.
+fn joiner_test_faces(test: &str, c: char, tags: &[&[u8; 4]]) -> Option<(Arc<FontFace>, Arc<FontFace>)> {
+    let db = FontDb::global();
+    let opentype = |face: &FontFace| {
+        use skrifa::raw::TableProvider;
+        tags.is_empty()
+            || face
+                .skrifa()
+                .and_then(|f| f.gsub().ok())
+                .and_then(|gsub| gsub.script_list().ok())
+                .is_some_and(|list| list.script_records().iter().any(|r| tags.contains(&&r.script_tag().to_be_bytes())))
+    };
+    let faces = ["Source Sans 3", "Helvetica", "Arial", "Times New Roman", "Liberation Sans", "DejaVu Sans"]
+        .into_iter()
+        .filter_map(|name| db.face(name, "Regular").filter(|face| face.family == name))
+        .filter(|primary| primary.covers(ZWNJ) && primary.covers(ZWJ) && !primary.covers(c))
+        .find_map(|primary| db.fallback_for(c, primary.id()).filter(|f| opentype(f)).map(|fallback| (primary, fallback)));
+    if let Some((primary, fallback)) = &faces {
+        let path = fallback.path().map_or_else(|| "bundled font".to_string(), |p| p.display().to_string());
+        eprintln!("ASSERTIONS RUN: {test}; {} {} with the fallback {} {} ({path})", primary.family, primary.style, fallback.family, fallback.style);
+    } else {
+        eprintln!("SKIPPED: {test}; no installed Latin face maps both joiners and falls back to a face for {c} with the lookups it needs");
+    }
+    faces
+}
+
+fn layout_in(face: &FontFace, text: &str, tracking: f64) -> TextLayout {
+    let st = CharStyle { font_family: face.family.clone(), font_style: face.style.clone(), size: 20.0, tracking, ..CharStyle::default() };
+    layout(FontDb::global(), &TextObject::point(Point::ZERO, text, st))
+}
+
+/// `text` set in `primary` is drawn, spaced and stepped through exactly as `text` set in `fallback`.
+fn assert_shaped_as_in_the_fallback(primary: &FontFace, fallback: &FontFace, text: &str) {
+    for tracking in [0.0, 200.0] {
+        let (mixed, single) = (layout_in(primary, text, tracking), layout_in(fallback, text, tracking));
+        // The caret stops where it stops in the fallback face.
+        assert_eq!(caret_horizontal(&mixed, 0, true), caret_horizontal(&single, 0, true), "{text:?} with tracking {tracking}");
+        // Tracking adds no extra space around the joiner.
+        let width = |l: &TextLayout| l.glyphs.iter().map(|g| g.advance).sum::<f64>();
+        assert_eq!(width(&mixed), width(&single), "{text:?} with tracking {tracking}");
+        let glyphs = |l: &TextLayout| l.glyphs.iter().map(|g| (g.font_id, g.gid, g.byte, g.len, g.origin, g.advance)).collect::<Vec<_>>();
+        assert_eq!(glyphs(&mixed), glyphs(&single), "{text:?} with tracking {tracking}");
+        assert!(mixed.glyphs.iter().all(|g| g.font_id == fallback.id()), "{text:?}: {:?}", mixed.glyphs);
+    }
+}
+
+#[test]
+fn a_zero_width_non_joiner_in_devanagari_is_shaped_with_the_letters_around_it() {
+    let test = "a_zero_width_non_joiner_in_devanagari_is_shaped_with_the_letters_around_it";
+    let Some((primary, fallback)) = joiner_test_faces(test, 'क', &[]) else { return };
+    // Ka + halant + ZWNJ + Ssa: Ka keeps its explicit halant instead of forming the KaSsa conjunct.
+    let s = format!("क्{ZWNJ}ष");
+    assert_shaped_as_in_the_fallback(&primary, &fallback, &s);
+}
+
+#[test]
+fn a_zero_width_non_joiner_in_bangla_is_shaped_with_the_letters_around_it() {
+    let test = "a_zero_width_non_joiner_in_bangla_is_shaped_with_the_letters_around_it";
+    let Some((primary, fallback)) = joiner_test_faces(test, 'ক', &[]) else { return };
+    let s = format!("ক্{ZWNJ}ষ");
+    assert_shaped_as_in_the_fallback(&primary, &fallback, &s);
+}
+
+#[test]
+fn a_zero_width_joiner_in_devanagari_asks_for_a_half_form() {
+    let test = "a_zero_width_joiner_in_devanagari_asks_for_a_half_form";
+    let Some((primary, fallback)) = joiner_test_faces(test, 'क', &[b"dev2", b"deva"]) else { return };
+    // Ka + halant + ZWJ + Ssa: the half form of Ka before a full Ssa.
+    let s = format!("क्{ZWJ}ष");
+    assert_shaped_as_in_the_fallback(&primary, &fallback, &s);
+    let l = layout_in(&primary, &s, 0.0);
+    let half = l.glyphs.first().map(|g| g.gid);
+    assert!(half.is_some_and(|gid| gid != fallback.glyph_for('क') && gid != fallback.glyph_for('\u{094D}')), "no half form: {:?}", l.glyphs);
+    assert_eq!(l.glyphs.last().map(|g| g.gid), Some(fallback.glyph_for('ष')), "{:?}", l.glyphs);
+}