Fix hyphen duplication rule for some languages (#4058)

11 files changed, 179 insertions, 22 deletions

diff --git a/crates/typst/src/layout/inline/mod.rs b/crates/typst/src/layout/inline/mod.rs
index 0b73eef6..f8b17f46 100644
--- a/crates/typst/src/layout/inline/mod.rs
+++ b/crates/typst/src/layout/inline/mod.rs
@@ -298,6 +298,19 @@ impl SpanMapper {
     }
 }
 
+/// A dash at the end of a line.
+#[derive(Debug, Copy, Clone, Eq, PartialEq)]
+pub(super) enum Dash {
+    /// A hyphen added to break a word.
+    SoftHyphen,
+    /// Regular hyphen, present in a compound word, e.g. beija-flor.
+    HardHyphen,
+    /// An em dash.
+    Long,
+    /// An en dash.
+    Short,
+}
+
 /// A layouted line, consisting of a sequence of layouted paragraph items that
 /// are mostly borrowed from the preparation phase. This type enables you to
 /// measure the size of a line in a range before committing to building the
@@ -327,7 +340,7 @@ struct Line<'a> {
     justify: bool,
     /// Whether the line ends with a hyphen or dash, either naturally or through
     /// hyphenation.
-    dash: bool,
+    dash: Option<Dash>,
 }
 
 impl<'a> Line<'a> {
@@ -814,8 +827,10 @@ fn linebreak_simple<'a>(
     let mut last = None;
 
     breakpoints(p, |end, breakpoint| {
+        let prepend_hyphen = lines.last().map(should_repeat_hyphen).unwrap_or(false);
+
         // Compute the line and its size.
-        let mut attempt = line(engine, p, start..end, breakpoint);
+        let mut attempt = line(engine, p, start..end, breakpoint, prepend_hyphen);
 
         // If the line doesn't fit anymore, we push the last fitting attempt
         // into the stack and rebuild the line from the attempt's end. The
@@ -824,7 +839,7 @@ fn linebreak_simple<'a>(
             if let Some((last_attempt, last_end)) = last.take() {
                 lines.push(last_attempt);
                 start = last_end;
-                attempt = line(engine, p, start..end, breakpoint);
+                attempt = line(engine, p, start..end, breakpoint, prepend_hyphen);
             }
         }
 
@@ -894,7 +909,7 @@ fn linebreak_optimized<'a>(
     let mut table = vec![Entry {
         pred: 0,
         total: 0.0,
-        line: line(engine, p, 0..0, Breakpoint::Mandatory),
+        line: line(engine, p, 0..0, Breakpoint::Mandatory, false),
     }];
 
     let em = p.size;
@@ -908,8 +923,9 @@ fn linebreak_optimized<'a>(
         for (i, pred) in table.iter().enumerate().skip(active) {
             // Layout the line.
             let start = pred.line.end;
+            let prepend_hyphen = should_repeat_hyphen(&pred.line);
 
-            let attempt = line(engine, p, start..end, breakpoint);
+            let attempt = line(engine, p, start..end, breakpoint, prepend_hyphen);
 
             // Determine how much the line's spaces would need to be stretched
             // to make it the desired width.
@@ -987,7 +1003,7 @@ fn linebreak_optimized<'a>(
             cost = (0.01 + cost).powi(2);
 
             // Penalize two consecutive dashes (not necessarily hyphens) extra.
-            if attempt.dash && pred.line.dash {
+            if attempt.dash.is_some() && pred.line.dash.is_some() {
                 cost += CONSECUTIVE_DASH_COST;
             }
 
@@ -1022,6 +1038,7 @@ fn line<'a>(
     p: &'a Preparation,
     mut range: Range,
     breakpoint: Breakpoint,
+    prepend_hyphen: bool,
 ) -> Line<'a> {
     let end = range.end;
     let mut justify =
@@ -1037,7 +1054,7 @@ fn line<'a>(
             last: None,
             width: Abs::zero(),
             justify,
-            dash: false,
+            dash: None,
         };
     }
 
@@ -1047,7 +1064,7 @@ fn line<'a>(
 
     // Reshape the last item if it's split in half or hyphenated.
     let mut last = None;
-    let mut dash = false;
+    let mut dash = None;
     if let Some((Item::Text(shaped), before)) = inner.split_last() {
         // Compute the range we want to shape, trimming whitespace at the
         // end of the line.
@@ -1062,7 +1079,17 @@ fn line<'a>(
         // Deal with hyphens, dashes and justification.
         let shy = trimmed.ends_with('\u{ad}');
         let hyphen = breakpoint == Breakpoint::Hyphen;
-        dash = hyphen || shy || trimmed.ends_with(['-', '–', '—']);
+        dash = if hyphen || shy {
+            Some(Dash::SoftHyphen)
+        } else if trimmed.ends_with('-') {
+            Some(Dash::HardHyphen)
+        } else if trimmed.ends_with('–') {
+            Some(Dash::Short)
+        } else if trimmed.ends_with('—') {
+            Some(Dash::Long)
+        } else {
+            None
+        };
         justify |= text.ends_with('\u{2028}');
 
         // Deal with CJK punctuation at line ends.
@@ -1079,7 +1106,11 @@ fn line<'a>(
         // need the shaped empty string to make the line the appropriate
         // height. That is the case exactly if the string is empty and there
         // are no other items in the line.
-        if hyphen || start + shaped.text.len() > range.end || maybe_adjust_last_glyph {
+        if hyphen
+            || start + shaped.text.len() > range.end
+            || maybe_adjust_last_glyph
+            || prepend_hyphen
+        {
             if hyphen || start < range.end || before.is_empty() {
                 let mut reshaped = shaped.reshape(engine, &p.spans, start..range.end);
                 if hyphen || shy {
@@ -1131,7 +1162,10 @@ fn line<'a>(
         let end = range.end.min(base + shaped.text.len());
 
         // Reshape if necessary.
-        if range.start + shaped.text.len() > end || maybe_adjust_first_glyph {
+        if range.start + shaped.text.len() > end
+            || maybe_adjust_first_glyph
+            || prepend_hyphen
+        {
             // If the range is empty, we don't want to push an empty text item.
             if range.start < end {
                 let reshaped = shaped.reshape(engine, &p.spans, range.start..end);
@@ -1143,6 +1177,15 @@ fn line<'a>(
         }
     }
 
+    if prepend_hyphen {
+        let reshaped = first.as_mut().or(last.as_mut()).and_then(Item::text_mut);
+        if let Some(reshaped) = reshaped {
+            let width_before = reshaped.width;
+            reshaped.prepend_hyphen(engine, p.fallback);
+            width += reshaped.width - width_before;
+        }
+    }
+
     if maybe_adjust_first_glyph {
         let reshaped = first.as_mut().or(last.as_mut()).and_then(Item::text_mut);
         if let Some(reshaped) = reshaped {
@@ -1446,3 +1489,49 @@ fn overhang(c: char) -> f64 {
         _ => 0.0,
     }
 }
+
+/// Whether the hyphen should repeat at the start of the next line.
+fn should_repeat_hyphen(pred_line: &Line) -> bool {
+    // If the predecessor line does not end with a Dash::HardHyphen, we shall
+    // not place a hyphen at the start of the next line.
+    if pred_line.dash != Some(Dash::HardHyphen) {
+        return false;
+    }
+
+    // If there's a trimmed out space, we needn't repeat the hyphen. That's the
+    // case of a text like "...kebab é a -melhor- comida que existe", where the
+    // hyphens are a kind of emphasis marker.
+    if pred_line.trimmed.end != pred_line.end {
+        return false;
+    }
+
+    // The hyphen should repeat only in the languages that require that feature.
+    // For more information see the discussion at https://github.com/typst/typst/issues/3235
+    let Some(Item::Text(shape)) = pred_line.last.as_ref() else { return false };
+
+    match shape.lang {
+        // - Lower Sorbian: see https://dolnoserbski.de/ortografija/psawidla/K3
+        // - Czech: see https://prirucka.ujc.cas.cz/?id=164
+        // - Croatian: see http://pravopis.hr/pravilo/spojnica/68/
+        // - Polish: see https://www.ortograf.pl/zasady-pisowni/lacznik-zasady-pisowni
+        // - Portuguese: see https://www2.senado.leg.br/bdsf/bitstream/handle/id/508145/000997415.pdf (Base XX)
+        // - Slovak: see https://www.zones.sk/studentske-prace/gramatika/10620-pravopis-rozdelovanie-slov/
+        Lang::LOWER_SORBIAN
+        | Lang::CZECH
+        | Lang::CROATIAN
+        | Lang::POLISH
+        | Lang::PORTUGUESE
+        | Lang::SLOVAK => true,
+        // In Spanish the hyphen is required only if the word next to hyphen is
+        // not capitalized. Otherwise, the hyphen must not be repeated.
+        //
+        // See § 4.1.1.1.2.e on the "Ortografía de la lengua española"
+        // https://www.rae.es/ortografía/como-signo-de-división-de-palabras-a-final-de-línea
+        Lang::SPANISH => pred_line.bidi.text[pred_line.end..]
+            .chars()
+            .next()
+            .map(|c| !c.is_uppercase())
+            .unwrap_or(false),
+        _ => false,
+    }
+}
diff --git a/crates/typst/src/layout/inline/shaping.rs b/crates/typst/src/layout/inline/shaping.rs
index ff13f776..57b94230 100644
--- a/crates/typst/src/layout/inline/shaping.rs
+++ b/crates/typst/src/layout/inline/shaping.rs
@@ -447,6 +447,15 @@ impl<'a> ShapedText<'a> {
 
     /// Push a hyphen to end of the text.
     pub fn push_hyphen(&mut self, engine: &Engine, fallback: bool) {
+        self.insert_hyphen(engine, fallback, Side::Right)
+    }
+
+    /// Prepend a hyphen to start of the text.
+    pub fn prepend_hyphen(&mut self, engine: &Engine, fallback: bool) {
+        self.insert_hyphen(engine, fallback, Side::Left)
+    }
+
+    fn insert_hyphen(&mut self, engine: &Engine, fallback: bool, side: Side) {
         let world = engine.world;
         let book = world.book();
         let fallback_func = if fallback {
@@ -464,17 +473,17 @@ impl<'a> ShapedText<'a> {
             let ttf = font.ttf();
             let glyph_id = ttf.glyph_index('-')?;
             let x_advance = font.to_em(ttf.glyph_hor_advance(glyph_id)?);
-            let range = self
-                .glyphs
-                .last()
-                .map(|g| g.range.end..g.range.end)
-                // In the unlikely chance that we hyphenate after an empty line,
-                // ensure that the glyph range still falls after self.base so
-                // that subtracting either of the endpoints by self.base doesn't
-                // underflow. See <https://github.com/typst/typst/issues/2283>.
-                .unwrap_or_else(|| self.base..self.base);
+            let range = match side {
+                Side::Left => self.glyphs.first().map(|g| g.range.start..g.range.start),
+                Side::Right => self.glyphs.last().map(|g| g.range.end..g.range.end),
+            }
+            // In the unlikely chance that we hyphenate after an empty line,
+            // ensure that the glyph range still falls after self.base so
+            // that subtracting either of the endpoints by self.base doesn't
+            // underflow. See <https://github.com/typst/typst/issues/2283>.
+            .unwrap_or_else(|| self.base..self.base);
             self.width += x_advance.at(self.size);
-            self.glyphs.to_mut().push(ShapedGlyph {
+            let glyph = ShapedGlyph {
                 font,
                 glyph_id: glyph_id.0,
                 x_advance,
@@ -487,7 +496,11 @@ impl<'a> ShapedText<'a> {
                 span: (Span::detached(), 0),
                 is_justifiable: false,
                 script: Script::Common,
-            });
+            };
+            match side {
+                Side::Left => self.glyphs.to_mut().insert(0, glyph),
+                Side::Right => self.glyphs.to_mut().push(glyph),
+            }
             Some(())
         });
     }
diff --git a/crates/typst/src/text/lang.rs b/crates/typst/src/text/lang.rs
index 67df0c6e..6809238a 100644
--- a/crates/typst/src/text/lang.rs
+++ b/crates/typst/src/text/lang.rs
@@ -57,6 +57,7 @@ impl Lang {
     pub const BOKMÅL: Self = Self(*b"nb ", 2);
     pub const CATALAN: Self = Self(*b"ca ", 2);
     pub const CHINESE: Self = Self(*b"zh ", 2);
+    pub const CROATIAN: Self = Self(*b"hr ", 2);
     pub const CZECH: Self = Self(*b"cs ", 2);
     pub const DANISH: Self = Self(*b"da ", 2);
     pub const DUTCH: Self = Self(*b"nl ", 2);
@@ -70,12 +71,14 @@ impl Lang {
     pub const HUNGARIAN: Self = Self(*b"hu ", 2);
     pub const ITALIAN: Self = Self(*b"it ", 2);
     pub const JAPANESE: Self = Self(*b"ja ", 2);
+    pub const LOWER_SORBIAN: Self = Self(*b"dsb", 3);
     pub const NYNORSK: Self = Self(*b"nn ", 2);
     pub const POLISH: Self = Self(*b"pl ", 2);
     pub const PORTUGUESE: Self = Self(*b"pt ", 2);
     pub const ROMANIAN: Self = Self(*b"ro ", 2);
     pub const RUSSIAN: Self = Self(*b"ru ", 2);
     pub const SERBIAN: Self = Self(*b"sr ", 2);
+    pub const SLOVAK: Self = Self(*b"sk ", 2);
     pub const SLOVENIAN: Self = Self(*b"sl ", 2);
     pub const SPANISH: Self = Self(*b"es ", 2);
     pub const SWEDISH: Self = Self(*b"sv ", 2);
diff --git a/tests/ref/hyphenate-es-captalized-names.png b/tests/ref/hyphenate-es-captalized-names.png
new file mode 100644
index 00000000..803d6795
--- /dev/null
+++ b/tests/ref/hyphenate-es-captalized-names.png
diff --git a/tests/ref/hyphenate-es-repeat-hyphen.png b/tests/ref/hyphenate-es-repeat-hyphen.png
new file mode 100644
index 00000000..a4c5a060
--- /dev/null
+++ b/tests/ref/hyphenate-es-repeat-hyphen.png
diff --git a/tests/ref/hyphenate-pt-dash-emphasis.png b/tests/ref/hyphenate-pt-dash-emphasis.png
new file mode 100644
index 00000000..cab13ea4
--- /dev/null
+++ b/tests/ref/hyphenate-pt-dash-emphasis.png
diff --git a/tests/ref/hyphenate-pt-no-repeat-hyphen.png b/tests/ref/hyphenate-pt-no-repeat-hyphen.png
new file mode 100644
index 00000000..d0e34c9b
--- /dev/null
+++ b/tests/ref/hyphenate-pt-no-repeat-hyphen.png
diff --git a/tests/ref/hyphenate-pt-repeat-hyphen-hyphenate-true-with-emphasis.png b/tests/ref/hyphenate-pt-repeat-hyphen-hyphenate-true-with-emphasis.png
new file mode 100644
index 00000000..0bb23ab1
--- /dev/null
+++ b/tests/ref/hyphenate-pt-repeat-hyphen-hyphenate-true-with-emphasis.png
diff --git a/tests/ref/hyphenate-pt-repeat-hyphen-hyphenate-true.png b/tests/ref/hyphenate-pt-repeat-hyphen-hyphenate-true.png
new file mode 100644
index 00000000..d08859fb
--- /dev/null
+++ b/tests/ref/hyphenate-pt-repeat-hyphen-hyphenate-true.png
diff --git a/tests/ref/hyphenate-pt-repeat-hyphen-natural-word-breaking.png b/tests/ref/hyphenate-pt-repeat-hyphen-natural-word-breaking.png
new file mode 100644
index 00000000..d08859fb
--- /dev/null
+++ b/tests/ref/hyphenate-pt-repeat-hyphen-natural-word-breaking.png
diff --git a/tests/suite/layout/inline/hyphenate.typ b/tests/suite/layout/inline/hyphenate.typ
index bcad4d93..c366b38f 100644
--- a/tests/suite/layout/inline/hyphenate.typ
+++ b/tests/suite/layout/inline/hyphenate.typ
@@ -50,6 +50,58 @@ It's a #emph[Tree]beard.
 #set text(hyphenate: true)
 #h(6pt) networks, the rest.
 
+--- hyphenate-pt-repeat-hyphen-natural-word-breaking ---
+// The word breaker naturally breaks arco-da-velha at arco-/-da-velha,
+// so we shall repeat the hyphen, even that hyphenate is set to false.
+#set page(width: 4cm)
+#set text(lang: "pt")
+
+Alguma coisa no arco-da-velha é algo que está muito longe.
+
+--- hyphenate-pt-repeat-hyphen-hyphenate-true ---
+#set page(width: 4cm)
+#set text(lang: "pt", hyphenate: true)
+
+Alguma coisa no arco-da-velha é algo que está muito longe.
+
+--- hyphenate-pt-repeat-hyphen-hyphenate-true-with-emphasis ---
+#set page(width: 4cm)
+#set text(lang: "pt", hyphenate: true)
+
+Alguma coisa no _arco-da-velha_ é algo que está muito longe.
+
+--- hyphenate-pt-no-repeat-hyphen ---
+#set page(width: 4cm)
+#set text(lang: "pt", hyphenate: true)
+
+Um médico otorrinolaringologista cuida da garganta do paciente.
+
+--- hyphenate-pt-dash-emphasis ---
+// If the hyphen is followed by a space we shall not repeat the hyphen
+// at the next line
+#set page(width: 4cm)
+#set text(lang: "pt", hyphenate: true)
+
+Quebabe é a -melhor- comida que existe.
+
+--- hyphenate-es-repeat-hyphen ---
+#set page(width: 6cm)
+#set text(lang: "es", hyphenate: true)
+
+Lo que entendemos por nivel léxico-semántico, en cuanto su sentido más
+gramatical: es aquel que estudia el origen y forma de las palabras de
+un idioma.
+
+--- hyphenate-es-captalized-names ---
+// If the hyphen is followed by a capitalized word we shall not repeat
+//  the hyphen at the next line
+#set page(width: 6.2cm)
+#set text(lang: "es", hyphenate: true)
+
+Tras el estallido de la contienda Ruiz-Giménez fue detenido junto a sus
+dos hermanos y puesto bajo custodia por las autoridades republicanas, con
+el objetivo de protegerle de las patrullas de milicianos.
+
 --- costs-widow-orphan ---
 #set page(height: 60pt)