Tweak default ignorable fix (#5105)

3 files changed, 17 insertions, 23 deletions

diff --git a/crates/typst-pdf/src/content.rs b/crates/typst-pdf/src/content.rs
index 4f2baafb..79323b6a 100644
--- a/crates/typst-pdf/src/content.rs
+++ b/crates/typst-pdf/src/content.rs
@@ -493,7 +493,7 @@ fn write_normal_text(
 
     let glyph_set = ctx.resources.glyph_sets.entry(text.item.font.clone()).or_default();
     for g in text.glyphs() {
-        glyph_set.entry(g.id).or_insert_with(|| text.glyph_text(&g));
+        glyph_set.entry(g.id).or_insert_with(|| text.glyph_text(g));
     }
 
     let fill_transform = ctx.state.transforms(Size::zero(), pos);
@@ -638,7 +638,7 @@ fn write_color_glyphs(
 
         ctx.content.show(Str(&[index]));
 
-        glyph_set.entry(glyph.id).or_insert_with(|| text.glyph_text(&glyph));
+        glyph_set.entry(glyph.id).or_insert_with(|| text.glyph_text(glyph));
     }
     ctx.content.end_text();
 
diff --git a/crates/typst/src/layout/place.rs b/crates/typst/src/layout/place.rs
index ce13a0c9..67b50926 100644
--- a/crates/typst/src/layout/place.rs
+++ b/crates/typst/src/layout/place.rs
@@ -46,7 +46,7 @@ use crate::layout::{Alignment, Em, Length, Rel};
 /// you can wrap the `place` call in a [`box`] when the call is made
 /// in the middle of a paragraph. The alignment and offsets will then be
 /// relative to this zero-size box. To make sure it doesn't interfere with
-/// spacing, the box should be attached to a word using a zero-width joiner.
+/// spacing, the box should be attached to a word using a word joiner.
 ///
 /// For example, the following defines a function for attaching an annotation
 /// to the following word:
@@ -55,7 +55,7 @@ use crate::layout::{Alignment, Em, Length, Rel};
 /// >>> #set page(height: 70pt)
 /// #let annotate(..args) = {
 ///   box(place(..args))
-///   sym.zwj
+///   sym.wj
 ///   h(0pt, weak: true)
 /// }
 ///
diff --git a/crates/typst/src/text/item.rs b/crates/typst/src/text/item.rs
index f6044b2f..b2f9e272 100644
--- a/crates/typst/src/text/item.rs
+++ b/crates/typst/src/text/item.rs
@@ -89,33 +89,27 @@ impl<'a> TextItemView<'a> {
 
     /// Returns an iterator over the glyphs of the slice.
     ///
-    /// The range of text that each glyph represents is remapped to be relative
-    /// to the start of the slice.
-    pub fn glyphs(&self) -> impl Iterator<Item = Glyph> + '_ {
-        let first = self.item.glyphs[self.glyph_range.start].range();
-        let last = self.item.glyphs[self.glyph_range.end - 1].range();
-        let base = first.start.min(last.start) as u16;
-        (0..self.glyph_range.len()).map(move |index| {
-            let g = &self.item.glyphs[self.glyph_range.start + index];
-            Glyph {
-                range: g.range.start - base..g.range.end - base,
-                ..*g
-            }
-        })
+    /// Note that the ranges are not remapped. They still point into the
+    /// original text.
+    pub fn glyphs(&self) -> &[Glyph] {
+        &self.item.glyphs[self.glyph_range.clone()]
     }
 
-    /// The plain text for the given glyph. This is an approximation since
-    /// glyphs do not correspond 1-1 with codepoints.
+    /// The plain text for the given glyph from `glyphs()`. This is an
+    /// approximation since glyphs do not correspond 1-1 with codepoints.
     pub fn glyph_text(&self, glyph: &Glyph) -> EcoString {
+        // Trim default ignorables which might have ended up in the glyph's
+        // cluster. Keep interior ones so that joined emojis work. All of this
+        // is a hack and needs to be reworked. See
+        // https://github.com/typst/typst/pull/5099
         self.item.text[glyph.range()]
-            .chars()
-            .filter(|&c| !is_default_ignorable(c))
-            .collect()
+            .trim_matches(is_default_ignorable)
+            .into()
     }
 
     /// The total width of this text slice
     pub fn width(&self) -> Abs {
-        self.item.glyphs[self.glyph_range.clone()]
+        self.glyphs()
             .iter()
             .map(|g| g.x_advance)
             .sum::<Em>()