Math accent handling

library/src/math/accent.rs

@@ -1,12 +1,15 @@
 use super::*;
 
+/// How much the accent can be shorter than the base.
+const ACCENT_SHORT_FALL: Em = Em::new(0.5);
+
 /// # Accent
 /// An accented node.
 ///
 /// ## Example
 /// ```
-/// $acc(a, ->) != acc(a, ~)$ \
-/// $acc(a, `) = acc(a, grave)$
+/// $accent(a, ->) != accent(a, ~)$ \
+/// $accent(a, `) = accent(a, grave)$
 /// ```
 ///
 /// ## Parameters
@@ -16,97 +19,154 @@ use super::*;
 ///
 ///   ### Example
 ///   ```
-///   $acc(A B C, ->)$
+///   $accent(A B C, ->)$
 ///   ```
 ///
 /// - accent: Content (positional, required)
 ///   The accent to apply to the base.
 ///
 ///   Supported accents include:
-///   - Grave: `` ` ``
-///   - Acute: `´`
-///   - Circumflex: `^`
-///   - Tilde: `~`
-///   - Macron: `¯`
-///   - Overline: `‾`
-///   - Breve: `˘`
+///   - Plus: `` + ``
+///   - Overline: `` - ``, `‾`
 ///   - Dot: `.`
+///   - Circumflex: `^`
+///   - Acute: `´`
+///   - Low Line: `_`
+///   - Grave: `` ` ``
+///   - Tilde: `~`
 ///   - Diaeresis: `¨`
+///   - Macron: `¯`
+///   - Acute: `´`
+///   - Cedilla: `¸`
 ///   - Caron: `ˇ`
-///   - Arrow: `→`
+///   - Breve: `˘`
+///   - Double acute: `˝`
+///   - Left arrow: `<-`
+///   - Right arrow: `->`
 ///
 /// ## Category
 /// math
 #[func]
-#[capable(Texify)]
+#[capable(LayoutMath)]
 #[derive(Debug, Hash)]
-pub struct AccNode {
+pub struct AccentNode {
     /// The accent base.
     pub base: Content,
-    /// The Unicode accent character.
-    pub accent: char,
+    /// The accent.
+    pub accent: Content,
 }
 
 #[node]
-impl AccNode {
+impl AccentNode {
     fn construct(_: &Vm, args: &mut Args) -> SourceResult<Content> {
         let base = args.expect("base")?;
-        let Spanned { v, span } = args.expect::<Spanned<Content>>("accent")?;
-        let accent = match extract(&v) {
-            Some(Ok(c)) => c,
-            Some(Err(msg)) => bail!(span, "{}", msg),
-            None => bail!(span, "not an accent"),
-        };
+        let accent = args.expect("accent")?;
         Ok(Self { base, accent }.pack())
     }
 }
 
-#[rustfmt::skip]
-fn extract(content: &Content) -> Option<Result<char, &'static str>> {
-    let MathNode { children, .. } = content.to::<MathNode>()?;
-    let [child] = children.as_slice() else { return None };
-    let c = if let Some(atom) = child.to::<AtomNode>() {
-        let mut chars = atom.0.chars();
-        chars.next().filter(|_| chars.next().is_none())?
-    } else if let Some(symbol) = child.to::<SymbolNode>() {
-        match symmie::get(&symbol.0) {
-            Some(c) => c,
-            None => return Some(Err("unknown symbol")),
+impl LayoutMath for AccentNode {
+    fn layout_math(&self, ctx: &mut MathContext) -> SourceResult<()> {
+        ctx.style(ctx.style.with_cramped(true));
+        let base = ctx.layout_fragment(&self.base)?;
+        ctx.unstyle();
+
+        let base_attach = match base {
+            MathFragment::Glyph(base) => {
+                attachment(ctx, base.id, base.italics_correction)
             }
-    } else {
-        return None;
+            _ => (base.width() + base.italics_correction()) / 2.0,
         };
 
-    Some(Ok(match c {
-        '`' | '\u{300}' => '\u{300}',              // Grave
-        '´' | '\u{301}' => '\u{301}',              // Acute
-        '^' | '\u{302}' => '\u{302}',              // Circumflex
-        '~' | '\u{223C}' | '\u{303}' => '\u{303}', // Tilde
-        '¯' | '\u{304}' => '\u{304}',              // Macron
-        '‾' | '\u{305}' => '\u{305}',              // Overline
-        '˘' | '\u{306}' => '\u{306}',              // Breve
-        '.' | '\u{22C5}' | '\u{307}' => '\u{307}', // Dot
-        '¨' | '\u{308}' => '\u{308}',              // Diaeresis
-        'ˇ' | '\u{30C}' => '\u{30C}',              // Caron
-        '→' | '\u{20D7}' => '\u{20D7}',            // Arrow
-        _ => return None,
-    }))
-}
-
-impl Texify for AccNode {
-    fn texify(&self, t: &mut Texifier) -> SourceResult<()> {
-        if let Some(sym) = unicode_math::SYMBOLS.iter().find(|sym| {
-            sym.codepoint == self.accent
-                && sym.atom_type == unicode_math::AtomType::Accent
-        }) {
-            t.push_str("\\");
-            t.push_str(sym.name);
-            t.push_str("{");
-            self.base.texify(t)?;
-            t.push_str("}");
-        } else {
-            self.base.texify(t)?;
+        let Some(c) = extract(&self.accent) else {
+            ctx.push(base);
+            if let Some(span) = self.accent.span() {
+                bail!(span, "not an accent");
             }
+            return Ok(());
+        };
+
+        // Forcing the accent to be at least as large as the base makes it too
+        // wide in many case.
+        let glyph = GlyphFragment::new(ctx, c);
+        let short_fall = ACCENT_SHORT_FALL.scaled(ctx);
+        let variant = glyph.stretch_horizontal(ctx, base.width(), short_fall);
+        let accent = variant.frame;
+        let accent_attach = match variant.id {
+            Some(id) => attachment(ctx, id, variant.italics_correction),
+            None => accent.width() / 2.0,
+        };
+
+        // Descent is negative because the accent's ink bottom is above the
+        // baseline. Therefore, the default gap is the accent's negated descent
+        // minus the accent base height. Only if the base is very small, we need
+        // a larger gap so that the accent doesn't move too low.
+        let accent_base_height = scaled!(ctx, accent_base_height);
+        let gap = -accent.descent() - base.height().min(accent_base_height);
+        let size = Size::new(base.width(), accent.height() + gap + base.height());
+        let accent_pos = Point::with_x(base_attach - accent_attach);
+        let base_pos = Point::with_y(accent.height() + gap);
+        let baseline = base_pos.y + base.ascent();
+
+        let mut frame = Frame::new(size);
+        frame.set_baseline(baseline);
+        frame.push_frame(accent_pos, accent);
+        frame.push_frame(base_pos, base.to_frame(ctx));
+        ctx.push(frame);
+
         Ok(())
     }
 }
+
+/// The horizontal attachment position for the given glyph.
+fn attachment(ctx: &MathContext, id: GlyphId, italics_correction: Abs) -> Abs {
+    ctx.table
+        .glyph_info
+        .and_then(|info| info.top_accent_attachments)
+        .and_then(|attachments| attachments.get(id))
+        .map(|record| record.value.scaled(ctx))
+        .unwrap_or_else(|| {
+            let advance = ctx.ttf.glyph_hor_advance(id).unwrap_or_default();
+            (advance.scaled(ctx) + italics_correction) / 2.0
+        })
+}
+
+/// Extract a single character from content.
+fn extract(accent: &Content) -> Option<char> {
+    let atom = accent.to::<MathNode>()?.body.to::<AtomNode>()?;
+    let mut chars = atom.0.chars();
+    let c = chars.next().filter(|_| chars.next().is_none())?;
+    Some(combining(c))
+}
+
+/// Convert from a non-combining accent to a combining one.
+///
+/// https://www.w3.org/TR/mathml-core/#combining-character-equivalences
+fn combining(c: char) -> char {
+    match c {
+        '\u{002b}' => '\u{031f}',
+        '\u{002d}' => '\u{0305}',
+        '\u{002e}' => '\u{0307}',
+        '\u{005e}' => '\u{0302}',
+        '\u{005f}' => '\u{0332}',
+        '\u{0060}' => '\u{0300}',
+        '\u{007e}' => '\u{0303}',
+        '\u{00a8}' => '\u{0308}',
+        '\u{00af}' => '\u{0304}',
+        '\u{00b4}' => '\u{0301}',
+        '\u{00b8}' => '\u{0327}',
+        '\u{02c6}' => '\u{0302}',
+        '\u{02c7}' => '\u{030c}',
+        '\u{02d8}' => '\u{0306}',
+        '\u{02d9}' => '\u{0307}',
+        '\u{02db}' => '\u{0328}',
+        '\u{02dc}' => '\u{0303}',
+        '\u{02dd}' => '\u{030b}',
+        '\u{203e}' => '\u{0305}',
+        '\u{2190}' => '\u{20d6}',
+        '\u{2192}' => '\u{20d7}',
+        '\u{2212}' => '\u{0305}',
+        '\u{27f6}' => '\u{20d7}',
+        _ => c,
+    }
+}

library/src/math/mod.rs

@@ -47,6 +47,7 @@ use crate::text::{families, variant, FallbackList, FontFamily, SpaceNode, Symbol
 /// Hook up all math definitions.
 pub fn define(scope: &mut Scope) {
     scope.def_func::<MathNode>("math");
+    scope.def_func::<AccentNode>("accent");
     scope.def_func::<FracNode>("frac");
     scope.def_func::<BinomNode>("binom");
     scope.def_func::<ScriptNode>("script");