9. Parse math field access in the lexer

crates/typst-syntax/src/lexer.rs

@@ -109,7 +109,10 @@ impl Lexer<'_> {
             Some('`') if self.mode != LexMode::Math => return self.raw(),
             Some(c) => match self.mode {
                 LexMode::Markup => self.markup(start, c),
-                LexMode::Math => self.math(start, c),
+                LexMode::Math => match self.math(start, c) {
+                    (kind, None) => kind,
+                    (kind, Some(node)) => return (kind, node),
+                },
                 LexMode::Code => self.code(start, c),
             },
 
@@ -507,8 +510,8 @@ impl Lexer<'_> {
 
 /// Math.
 impl Lexer<'_> {
-    fn math(&mut self, start: usize, c: char) -> SyntaxKind {
-        match c {
+    fn math(&mut self, start: usize, c: char) -> (SyntaxKind, Option<SyntaxNode>) {
+        let kind = match c {
             '\\' => self.backslash(),
             '"' => self.string(),
 
@@ -561,11 +564,41 @@ impl Lexer<'_> {
             // Identifiers.
             c if is_math_id_start(c) && self.s.at(is_math_id_continue) => {
                 self.s.eat_while(is_math_id_continue);
-                SyntaxKind::MathIdent
+                let (kind, node) = self.math_ident_or_field(start);
+                return (kind, Some(node));
             }
 
             // Other math atoms.
             _ => self.math_text(start, c),
+        };
+        (kind, None)
+    }
+
+    /// Parse a single `MathIdent` or an entire `FieldAccess`.
+    fn math_ident_or_field(&mut self, start: usize) -> (SyntaxKind, SyntaxNode) {
+        let mut kind = SyntaxKind::MathIdent;
+        let mut node = SyntaxNode::leaf(kind, self.s.from(start));
+        while let Some(ident) = self.maybe_dot_ident() {
+            kind = SyntaxKind::FieldAccess;
+            let field_children = vec![
+                node,
+                SyntaxNode::leaf(SyntaxKind::Dot, '.'),
+                SyntaxNode::leaf(SyntaxKind::Ident, ident),
+            ];
+            node = SyntaxNode::inner(kind, field_children);
+        }
+        (kind, node)
+    }
+
+    /// If at a dot and a math identifier, eat and return the identifier.
+    fn maybe_dot_ident(&mut self) -> Option<&str> {
+        if self.s.scout(1).is_some_and(is_math_id_start) && self.s.eat_if('.') {
+            let ident_start = self.s.cursor();
+            self.s.eat();
+            self.s.eat_while(is_math_id_continue);
+            Some(self.s.from(ident_start))
+        } else {
+            None
         }
     }
 

crates/typst-syntax/src/parser.rs

@@ -6,9 +6,7 @@ use ecow::{eco_format, EcoString};
 use unicode_math_class::MathClass;
 
 use crate::set::{syntax_set, SyntaxSet};
-use crate::{
-    ast, is_ident, is_newline, set, LexMode, Lexer, SyntaxError, SyntaxKind, SyntaxNode,
-};
+use crate::{ast, is_newline, set, LexMode, Lexer, SyntaxError, SyntaxKind, SyntaxNode};
 
 /// Parses a source file as top-level markup.
 pub fn parse(text: &str) -> SyntaxNode {
@@ -261,21 +259,11 @@ fn math_expr_prec(p: &mut Parser, min_prec: usize, stop: SyntaxKind) {
     let mut continuable = false;
     match p.current() {
         SyntaxKind::Hash => embedded_code_expr(p),
-        SyntaxKind::MathIdent => {
+        // The lexer manages creating full FieldAccess nodes if needed.
+        SyntaxKind::MathIdent | SyntaxKind::FieldAccess => {
             continuable = true;
             p.eat();
-            while p.directly_at(SyntaxKind::Text) && p.current_text() == "." && {
-                let mut copy = p.lexer.clone();
-                let start = copy.cursor();
-                let next = copy.next().0;
-                let end = copy.cursor();
-                matches!(next, SyntaxKind::MathIdent | SyntaxKind::Text)
-                    && is_ident(&p.text[start..end])
-            } {
-                p.convert_and_eat(SyntaxKind::Dot);
-                p.convert_and_eat(SyntaxKind::Ident);
-                p.wrap(m, SyntaxKind::FieldAccess);
-            }
+            // Parse a function call for an identifier or field access.
             if min_prec < 3 && p.directly_at(SyntaxKind::Text) && p.current_text() == "("
             {
                 math_args(p);

crates/typst-syntax/src/set.rs

@@ -58,6 +58,7 @@ pub const STMT: SyntaxSet = syntax_set!(Let, Set, Show, Import, Include, Return)
 pub const MATH_EXPR: SyntaxSet = syntax_set!(
     Hash,
     MathIdent,
+    FieldAccess,
     Text,
     MathShorthand,
     Linebreak,

tests/suite/math/symbols.typ

@@ -0,0 +1,29 @@
+// Test math symbol edge cases.
+
+--- math-symbol-basic ---
+#let sym = symbol("s", ("basic", "s"))
+#test($sym.basic$, $#"s"$)
+
+--- math-symbol-underscore ---
+#let sym = symbol("s", ("test_underscore", "s"))
+// Error: 6-10 unknown symbol modifier
+$sym.test_underscore$
+
+--- math-symbol-dash ---
+#let sym = symbol("s", ("test-dash", "s"))
+// Error: 6-10 unknown symbol modifier
+$sym.test-dash$
+
+--- math-symbol-double ---
+#let sym = symbol("s", ("test.basic", "s"))
+#test($sym.test.basic$, $#"s"$)
+
+--- math-symbol-double-underscore ---
+#let sym = symbol("s", ("one.test_underscore", "s"))
+// Error: 10-14 unknown symbol modifier
+$sym.one.test_underscore$
+
+--- math-symbol-double-dash ---
+#let sym = symbol("s", ("one.test-dash", "s"))
+// Error: 10-14 unknown symbol modifier
+$sym.one.test-dash$