Fix hyphenation for Czech, Polish, Portuguese and Spanish

Fix typst#3235

crates/typst/src/layout/inline/mod.rs

@@ -816,8 +816,10 @@ 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
@@ -826,7 +828,7 @@ 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);
  }
  }
 
@@ -896,7 +898,7 @@ 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;
@@ -910,8 +912,9 @@ 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.
@@ -1024,6 +1027,7 @@ fn line<'a>(
  p: &'a Preparation,
  mut range: Range,
  breakpoint: Breakpoint,
+ prepend_hyphen: bool,
 ) -> Line<'a> {
  let end = range.end;
  let mut justify =
@@ -1091,13 +1095,25 @@ 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 < range.end || before.is_empty() {
+ if hyphen
+ || start + shaped.text.len() > range.end
+ || maybe_adjust_last_glyph
+ || (prepend_hyphen && before.is_empty())
+ {
+ if hyphen
+ || start < range.end
+ || before.is_empty()
+ || (prepend_hyphen && before.is_empty())
+ {
  let mut reshaped = shaped.reshape(engine, &p.spans, start..range.end);
  if hyphen || shy {
  reshaped.push_hyphen(engine, p.fallback);
  }
 
+ if prepend_hyphen && before.is_empty() {
+ reshaped.prepend_hyphen(engine, p.fallback);
+ }
+
  if let Some(last_glyph) = reshaped.glyphs.last() {
  if last_glyph.is_cjk_left_aligned_punctuation(gb_style) {
  // If the last glyph is a CJK punctuation, we want to shrink it.
@@ -1143,10 +1159,18 @@ 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);
+ let mut reshaped = shaped.reshape(engine, &p.spans, range.start..end);
+
+ if prepend_hyphen {
+ reshaped.prepend_hyphen(engine, p.fallback)
+ }
+
  width += reshaped.width;
  first = Some(Item::Text(reshaped));
  }
@@ -1458,3 +1482,44 @@ fn overhang(c: char) -> f64 {
  _ => 0.0,
  }
 }
+
+/// Whether the hyphen should repeat at the begin of the next line
+fn should_repeat_hyphen(pred_line: &Line) -> bool {
+ // If the predecessor line does not end with with a Dash::HardHyphen, we shouldn't put a hyphen
+ // on 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 requires that feature.
+ // For more information see the discussion at https://github.com/typst/typst/issues/3235
+ if let Some(Item::Text(shape)) = pred_line.last.as_ref() {
+ match shape.lang {
+ // Czech: see https://prirucka.ujc.cas.cz/?id=164
+ //
+ // Polish: see https://www.ortograf.pl/zasady-pisowni/lacznik-zasady-pisowni
+ //
+ // Portuguese: see Base XX of "Acordo Ortográfico da Língua Portuguesa de 1990"
+ // https://www2.senado.leg.br/bdsf/bitstream/handle/id/508145/000997415.pdf
+ Lang::CZECH | Lang::POLISH | Lang::PORTUGUESE => true,
+ // In Spanish the hyphen is required only if the word next to hyphen isn't capitalized.
+ //
+ // 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,
+ }
+ } else {
+ false
+ }
+}

crates/typst/src/layout/inline/shaping.rs

@@ -492,6 +492,56 @@ impl<'a> ShapedText<'a> {
  });
  }
 
+ /// Prepend a hyphen to begin of the text.
+ pub fn prepend_hyphen(&mut self, engine: &Engine, fallback: bool) {
+ let world = engine.world;
+ let book = world.book();
+ let fallback_func = if fallback {
+ Some(|| book.select_fallback(None, self.variant, "-"))
+ } else {
+ None
+ };
+ let mut chain = families(self.styles)
+ .map(|family| book.select(family, self.variant))
+ .chain(fallback_func.iter().map(|f| f()))
+ .flatten();
+
+ chain.find_map(|id| {
+ let font = world.font(id)?;
+ 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
+ .first()
+ .map(|g| g.range.start..g.range.start)
+ // 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().insert(
+ 0,
+ ShapedGlyph {
+ font,
+ glyph_id: glyph_id.0,
+ x_advance,
+ x_offset: Em::zero(),
+ y_offset: Em::zero(),
+ adjustability: Adjustability::default(),
+ range,
+ safe_to_break: true,
+ c: '-',
+ span: (Span::detached(), 0),
+ is_justifiable: false,
+ script: Script::Common,
+ },
+ );
+ Some(())
+ });
+ }
+
  /// Find the subslice of glyphs that represent the given text range if both
  /// sides are safe to break.
  fn slice_safe_to_break(&self, text_range: Range<usize>) -> Option<&[ShapedGlyph]> {

tests/suite/layout/inline/hyphenate.typ

@@ -50,6 +50,55 @@ 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, height: 2cm, margin: 2mm)
+#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, height: 2cm, margin: 2mm)
+#set text(lang: "pt", hyphenate: true)
+
+Alguma coisa no arco-da-velha é algo que está muito longe.
+
+--- hyphenate-pt-no-repeat-hyphen ---
+// LARGE
+#set page(width: 4cm, height: 2cm, margin: 2mm)
+#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, height: 2cm, margin: 2mm)
+#set text(lang: "pt", hyphenate: true)
+
+Quebabe é a -melhor- comida que existe.
+
+--- hyphenate-es-repeat-hyphen ---
+// LARGE
+#set page(width: 6.25cm, height: 6cm, margin: 2mm)
+#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 ---
+// LARGE
+// If the hyphen is followed by a capitalized word we shall not repeat
+// the hyphen at the next line
+#set page(width: 6.5cm, height: 3.5cm, margin: 2mm)
+#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)