From d0120bf6a642beb92e31e026fc8a9bc64778bc13 Mon Sep 17 00:00:00 2001
From: Abdelrehman Fouad <97409061+MaridWSH@users.noreply.github.com>
Date: Sun, 11 Oct 2026 11:45:12 +0200
Subject: [PATCH] Text: Optical kerning leaves dashes and fullwidth forms alone
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Two dashes end to end have no ink gap to measure, so -- and the double
em dash opened up by the loosen cap; Chinese sets —— as one continuous
rule. Fullwidth Latin (U+FF21–FF5A) was kernable through its Latin
script and moved CJK text off its em grid. Both are now excluded from
kernable(), with a unit test and a compose test that failed before.
---
 crates/compose/src/optical.rs | 26 +++++++++++++++++++++++++-
 crates/compose/src/tests.rs   | 15 +++++++++++++++
 docs/typography-parity.md     |  2 +-
 3 files changed, 41 insertions(+), 2 deletions(-)

diff --git a/crates/compose/src/optical.rs b/crates/compose/src/optical.rs
index bd697c0dd..b7e4b8633 100644
--- a/crates/compose/src/optical.rs
+++ b/crates/compose/src/optical.rs
@@ -155,15 +155,24 @@ fn reference_gap(db: &ScopedFonts<'_>, face: &FontFace) -> Option<f64> {
 
 /// Whether `c` takes part in optical kerning: letters of the Latin, Greek and Cyrillic scripts,
 /// digits and Western punctuation. Spaces, marks, joined and ideographic scripts keep their
-/// font spacing.
+/// font spacing, as do dashes (two bars end to end have no ink gap to measure, and `——` must
+/// stay one continuous rule) and the fullwidth forms, which sit on the CJK em grid.
 pub fn kernable(c: char) -> bool {
     use unicode_script::{Script, UnicodeScript};
+    if is_dash(c) || ('\u{FF00}'..='\u{FFEF}').contains(&c) {
+        return false;
+    }
     if c.is_ascii_digit() || c.is_ascii_punctuation() || ('\u{2010}'..='\u{2027}').contains(&c) {
         return true;
     }
     c.is_alphabetic() && matches!(c.script(), Script::Latin | Script::Greek | Script::Cyrillic)
 }
 
+/// Dash punctuation (Unicode `Pd`) in the ranges optical kerning would otherwise touch.
+fn is_dash(c: char) -> bool {
+    matches!(c, '-' | '\u{2010}'..='\u{2015}' | '\u{2E3A}' | '\u{2E3B}' | '\u{FE58}' | '\u{FF0D}')
+}
+
 /// The adjustment for a pair, in font units: how far the first glyph's advance moves.
 fn adjustment(reference: f64, mean: f64, min: f64, upem: f64) -> f64 {
     let mut adj = (reference - mean) * PULL;
@@ -255,4 +264,19 @@ mod tests {
         assert!(kernable('A') && kernable('é') && kernable('Ж') && kernable('Ω') && kernable('7') && kernable('.') && kernable('“'));
         assert!(!kernable(' ') && !kernable('\t') && !kernable('漢') && !kernable('ب') && !kernable('\u{301}'));
     }
+
+    #[test]
+    fn dashes_and_fullwidth_forms_keep_their_font_spacing() {
+        // `--` and `——` have no ink gap to measure and would otherwise open up; `——` is one
+        // continuous dash in Chinese text.
+        for c in ['-', '\u{2010}', '\u{2013}', '\u{2014}', '\u{2015}', '\u{2E3A}', '\u{FF0D}'] {
+            assert!(!kernable(c), "{c:?} is a dash");
+        }
+        // Fullwidth Latin letters, digits and punctuation sit on the CJK em grid.
+        for c in ['\u{FF21}', '\u{FF41}', '\u{FF5A}', '\u{FF11}', '\u{FF0C}'] {
+            assert!(!kernable(c), "{c:?} is a fullwidth form");
+        }
+        // Hyphenated and dashed words still kern on their letters.
+        assert!(kernable('A') && kernable('.') && kernable('\u{2019}'));
+    }
 }
diff --git a/crates/compose/src/tests.rs b/crates/compose/src/tests.rs
index 664329a07..339c22477 100644
--- a/crates/compose/src/tests.rs
+++ b/crates/compose/src/tests.rs
@@ -312,6 +312,21 @@ fn optical_kerning_is_bounded_and_skips_spaces() {
     assert!((none[0].1 - optical[0].1).abs() < 1e-9);
 }
 
+#[test]
+fn optical_kerning_leaves_dash_pairs_alone() {
+    use designcraft_doc::Kerning;
+    // Two bars end to end have a near-zero ink gap; pulling them towards the stem gap would
+    // break `--` and the double em dash `——` that Chinese sets as one continuous rule.
+    for text in ["--", "\u{2014}\u{2014}", "a--b", "\u{2013}\u{2013}"] {
+        let (none_w, none) = set_with_kerning(text, Kerning::None);
+        let (optical_w, optical) = set_with_kerning(text, Kerning::Optical);
+        assert!((none_w - optical_w).abs() < 1e-9, "{text:?}: optical {optical_w:.3} vs none {none_w:.3}");
+        for ((c, a), (_, b)) in none.iter().zip(&optical) {
+            assert!((a - b).abs() < 1e-9, "{text:?} {c:?}: {a:.3} vs {b:.3}");
+        }
+    }
+}
+
 #[test]
 fn hyphen_limit_is_a_hard_constraint() {
     let text = CORPUS.join(" ");
diff --git a/docs/typography-parity.md b/docs/typography-parity.md
index a3afae27a..cd17dd7f5 100644
--- a/docs/typography-parity.md
+++ b/docs/typography-parity.md
@@ -14,7 +14,7 @@ against InDesign is **unmeasured**, which is the main gap.
 ## Fidelity: how we'd know
 
 There is no oracle yet. What exists: unit tests of break points and invariants in
-`crates/compose/src/tests.rs` (170 tests in the crate), IDML files built by DesignCraft and opened
+`crates/compose/src/tests.rs` (172 tests in the crate), IDML files built by DesignCraft and opened
 in InDesign 2026 by hand, and user reports comparing the two (#179, #180, #181, #282–#295).
 
 **Needed (gap 4 in [gaps.md](gaps.md)):** a set of synthetic stories (our own text and fonts: