From 47cf84f20b7fb486547cb3688d73d1b33763572a Mon Sep 17 00:00:00 2001
From: claude <claude@verso>
Date: Tue, 9 Jun 2026 13:57:10 +0000
Subject: [PATCH] fix(typst): highlight named arg keys after complex nested
 calls
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

canShift(CodeIdent) returns false in LALR-merged states that arise after
reducing a complex first argument (e.g. figure(table(...), caption: ...)).
The previous guard `!couldBeIdent && !canShift(CodeArgKey)` then caused
an early exit before the character-level scan ran, silently dropping the
CodeArgKey token for any named arg key that follows such a reduction.

Fix: run the backward character scan first and derive `couldBeArgKey`
from the raw predecessor char ('(' or ',') rather than from canShift.
The early-exit now reads `!couldBeIdent && !couldBeArgKey`, so arg-key
positions always proceed to the full scan regardless of parser state.
Also stop calling canShift(CodeArgKey) entirely — it is unreliable here.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 .../source-editor/lezer-typst/tokens.mjs      | 35 +++++++++++--------
 1 file changed, 20 insertions(+), 15 deletions(-)

diff --git a/services/web/frontend/js/features/source-editor/lezer-typst/tokens.mjs b/services/web/frontend/js/features/source-editor/lezer-typst/tokens.mjs
index 23205040ba..65d3098f0d 100644
--- a/services/web/frontend/js/features/source-editor/lezer-typst/tokens.mjs
+++ b/services/web/frontend/js/features/source-editor/lezer-typst/tokens.mjs
@@ -322,26 +322,36 @@ export const codeKeywordTokenizer = new ExternalTokenizer(
 //
 // Keywords and bools are excluded so codeKeywordTokenizer / CodeBool can
 // handle them without conflict.
+//
+// The backward scan runs BEFORE any canShift gate.  canShift(CodeArgKey) is
+// unreliable (LALR state merging can suppress it even at genuine arg-key
+// positions, e.g. 'caption:' after a complex nested call like 'table(...)').
+// We derive couldBeArgKey from character-level evidence ('(' or ',') and use
+// that to decide whether to continue even when canShift(CodeIdent) is false.
 export const codeIdentTokenizer = new ExternalTokenizer(
   (input, stack) => {
     const couldBeIdent = stack.canShift(CodeIdent)
-    // CodeArgKey detection is character-level (see below) — canShift is
-    // unreliable for it due to LALR merging, so don't gate on it here.
-    if (!couldBeIdent && !stack.canShift(CodeArgKey)) return
 
-    // Guard: only fire in code context.
-    // Walk back past whitespace to the nearest non-space character.
+    // Walk back past whitespace — primary context discriminator.
     let back = -1
     while (input.peek(back) === SPACE || input.peek(back) === TAB || input.peek(back) === NEWLINE) back--
     const prev = input.peek(back)
+
     if (prev !== HASH && prev !== DOT && prev !== OPEN_PAREN && prev !== COMMA) {
       // May be after a keyword like '#set' or '#show': scan back through the
       // keyword word itself and check that '#' immediately precedes it.
       if (!isIdentTail(prev)) return
-      while (isIdentTail(input.peek(back))) back--
-      if (input.peek(back) !== HASH) return
+      let b = back
+      while (isIdentTail(input.peek(b))) b--
+      if (input.peek(b) !== HASH) return
     }
 
+    // In arg-delimiter positions ('(' or ',') we may emit CodeArgKey regardless
+    // of canShift(CodeIdent) — LALR merging can suppress canShift(CodeIdent)
+    // after a complex first argument (e.g. figure(table(...), caption: ...)).
+    const couldBeArgKey = prev === OPEN_PAREN || prev === COMMA
+    if (!couldBeIdent && !couldBeArgKey) return
+
     // Must start with an identifier head character.
     if (!isIdentHead(input.next)) return
 
@@ -360,15 +370,10 @@ export const codeIdentTokenizer = new ExternalTokenizer(
     // Let codeKeywordTokenizer handle keywords; let CodeBool handle bools.
     if (KEYWORDS.has(word) || BOOLS.has(word)) return
 
-    // Emit CodeArgKey when this identifier is a named arg key (followed by ':').
-    // canShift(CodeArgKey) is unreliable here because LALR state merging can
-    // fold the codeArgItem start state into others where CodeArgKey is absent.
-    // Instead, use character-level context: we are inside a call argument list
-    // only when the nearest non-whitespace predecessor is '(' or ','.  That
-    // covers every named-arg position (#set text(size: …), align: …, etc.)
-    // while excluding keyword positions like '#show heading: …' (prev = 'g').
+    // Emit CodeArgKey when this identifier is immediately followed by ':'.
+    // Only applies in arg-delimiter positions (couldBeArgKey).
     let isArgKey = false
-    if (prev === OPEN_PAREN || prev === COMMA) {
+    if (couldBeArgKey) {
       let afterLen = len
       while (input.peek(afterLen) === SPACE || input.peek(afterLen) === TAB) afterLen++
       isArgKey = (input.peek(afterLen) === COLON)