From d391a3a166850fcd13f2585fb17a00aa840ea679 Mon Sep 17 00:00:00 2001
From: Adam Kowalski <adam.kowalski.work@gmail.com>
Date: Thu, 8 Jan 2026 21:24:40 -0800
Subject: [PATCH] chore: add unit test and improve comment

---
 src/shape.rs                   |  5 ++++-
 tests/ligature_segmentation.rs | 30 ------------------------------
 tests/shaping_and_rendering.rs | 30 ++++++++++++++++++++++++++++++
 3 files changed, 34 insertions(+), 31 deletions(-)
 delete mode 100644 tests/ligature_segmentation.rs

diff --git a/src/shape.rs b/src/shape.rs
index 491c69b..19d0e31 100644
--- a/src/shape.rs
+++ b/src/shape.rs
@@ -826,7 +826,10 @@ impl ShapeSpan {
 
         let mut start_word = 0;
         for (end_lb, _) in unicode_linebreak::linebreaks(span) {
-            // Workaround for broken |> ligature in code fonts
+            // The unicode-linebreak crate treats the pipe character '|' as a break opportunity (BA/AL class).
+            // This causes ShapeSpan::build to split text like '|>' into separate ShapeWords.
+            // When these words are shaped independently, the font shaping engine cannot form ligatures that cross the word boundary.
+            // We manually check for the '|>' sequence during segmentation and skip the break opportunity to ensure they remain in the same shaping run.
             if end_lb > 0 && end_lb < span.len() {
                 let b = span.as_bytes();
                 if b[end_lb - 1] == b'|' && b[end_lb] == b'>' {
diff --git a/tests/ligature_segmentation.rs b/tests/ligature_segmentation.rs
deleted file mode 100644
index c1936ac..0000000
--- a/tests/ligature_segmentation.rs
+++ /dev/null
@@ -1,30 +0,0 @@
-use cosmic_text::{Attrs, Buffer, FontSystem, Metrics, Shaping};
-
-#[test]
-fn ligature_segmentation() {
-    let mut font_system =
-        FontSystem::new_with_locale_and_db("en-US".into(), fontdb::Database::new());
-    let font = std::fs::read("fonts/Inter-Regular.ttf").unwrap();
-    font_system.db_mut().load_font_data(font);
-    let metrics = Metrics::new(14.0, 20.0);
-
-    let mut buffer = Buffer::new(&mut font_system, metrics);
-    let mut buffer = buffer.borrow_with(&mut font_system);
-
-    buffer.set_text("|>", &Attrs::new(), Shaping::Advanced, None);
-    buffer.shape_until_scroll(false);
-
-    let line = &buffer.lines[0];
-    let shape = line.shape_opt().expect("ShapeLine not found");
-    let span = &shape.spans[0];
-    
-    // The pipe character | is typically a line break opportunity.
-    // This test ensures that our patch prevents splitting |> into separate words,
-    // which would break ligature formation in fonts that support it.
-    assert_eq!(
-        span.words.len(),
-        1,
-        "Expected '|>' to be a single word (preserved for ligature), but found {} words.",
-        span.words.len()
-    );
-}
diff --git a/tests/shaping_and_rendering.rs b/tests/shaping_and_rendering.rs
index cb90ff0..1841111 100644
--- a/tests/shaping_and_rendering.rs
+++ b/tests/shaping_and_rendering.rs
@@ -73,3 +73,33 @@ fn test_english_mixed_with_arabic_paragraph_rendering() {
         .canvas(400, 110)
         .validate_text_rendering();
 }
+
+#[test]
+fn test_ligature_segmentation() {
+    use cosmic_text::{Buffer, FontSystem, Metrics, Shaping};
+
+    let mut font_system = FontSystem::new_with_locale_and_db("en-US".into(), fontdb::Database::new());
+    let font = std::fs::read("fonts/Inter-Regular.ttf").unwrap();
+    font_system.db_mut().load_font_data(font);
+    let metrics = Metrics::new(14.0, 20.0);
+
+    let mut buffer = Buffer::new(&mut font_system, metrics);
+    let mut buffer = buffer.borrow_with(&mut font_system);
+
+    buffer.set_text("|>", &Attrs::new(), Shaping::Advanced, None);
+    buffer.shape_until_scroll(false);
+
+    let line = &buffer.lines[0];
+    let shape = line.shape_opt().expect("ShapeLine not found");
+    let span = &shape.spans[0];
+
+    // The pipe character | is typically a line break opportunity.
+    // This test ensures that our patch prevents splitting |> into separate words,
+    // which would break ligature formation in fonts that support it.
+    assert_eq!(
+        span.words.len(),
+        1,
+        "Expected '|>' to be a single word (preserved for ligature), but found {} words.",
+        span.words.len()
+    );
+}
\ No newline at end of file