@lokascript/framework 2.7.1 → 2.8.0

package/dist/index.js

@@ -1420,7 +1420,8 @@ function buildFormatString(schema, profile, keyword) {
   if (profile.wordOrder === "SVO" || profile.wordOrder === "VSO") {
     parts.push(keyword);
   }
-  for (const role of schema.roles) {
+  const sortedRoles = sortRolesByWordOrder(schema.roles, profile.wordOrder);
+  for (const role of sortedRoles) {
     const marker = getMarkerForRole(role, profile);
     const roleName = `{${role.role}}`;
     if (marker) {
@@ -2878,7 +2879,9 @@ function buildProtocolSection() {
 3. No spaces around the colon in role:value pairs
 4. Strings with spaces must be quoted: \`patient:"hello world"\`
 5. Selectors start with \`#\`, \`.\`, \`[\`, \`@\`, or \`*\`
-6. Output must be valid bracket syntax: \`[action role:value ...]\``;
+6. A selector containing a space, a combinator (\`>\` \`+\` \`~\`), or a comma must use a selector literal: \`patient:<ul > li/>\`, \`patient:<.a, .b/>\`
+7. Inside a structural role (\`body\`, \`then\`, \`else\`, \`condition\`, \`loop-body\`, \`variable\`, \`catch\`, \`finally\`) a \`[...]\` value is always a nested command. Write an attribute selector there as \`condition:<[data-active]/>\`
+8. Output must be valid bracket syntax: \`[action role:value ...]\``;
   return {
     id: "protocol",
     title: "LSE Protocol",
@@ -2891,6 +2894,7 @@ function buildValueTypeSection() {
 
 | Type | Syntax | Example |
 |------|--------|---------|
+| Selector literal | Delimited by \`<\` and \`/>\` | \`<ul > li/>\`, \`<.a, .b/>\`, \`<[data-id]/>\` |
 | Selector | Starts with \`#\` \`.\` \`[\` \`@\` \`*\` | \`#button\`, \`.active\`, \`[data-id]\` |
 | String | Quoted with \`"\` or \`'\` | \`"hello world"\`, \`'json'\` |
 | Boolean | Exact: \`true\` / \`false\` | \`visible:true\` |
@@ -5086,7 +5090,39 @@ var _BaseTokenizer = class _BaseTokenizer {
         pos++;
       }
     }
-    return new TokenStreamImpl(tokens, this.language);
+    return new TokenStreamImpl(this.mergeColonQualifiedNames(tokens), this.language);
+  }
+  /**
+   * Fuse `name` + `:qualifier` into ONE identifier (`draggable:start`).
+   *
+   * `:name` is hyperscript's local-variable sigil, but a colon IMMEDIATELY
+   * preceded by an identifier is a qualifier (custom event namespace), not a
+   * sigil. The English tokenizer already merges these inside
+   * EnglishKeywordExtractor; this post-pass gives the other 23 languages the
+   * same stream. Strict position adjacency is the discriminator: whitespace
+   * between the tokens (`trigger :start`) breaks `end === start`, so a spaced
+   * local-variable reference survives untouched.
+   *
+   * Self-gating for non-hyperscript tokenizers (domain DSLs): their extractor
+   * sets tokenize `:` as bare punctuation (length 1), which never matches
+   * COLON_QUALIFIER, so this pass is a no-op for them.
+   */
+  mergeColonQualifiedNames(tokens) {
+    const out = [];
+    for (const tok of tokens) {
+      const prev = out[out.length - 1];
+      if (prev && _BaseTokenizer.ASCII_WORD.test(prev.value) && _BaseTokenizer.COLON_QUALIFIER.test(tok.value) && prev.position.end === tok.position.start) {
+        const merged = prev.value + tok.value;
+        out[out.length - 1] = createToken(
+          merged,
+          this.classifyToken(merged),
+          createPosition(prev.position.start, tok.position.end)
+        );
+        continue;
+      }
+      out.push(tok);
+    }
+    return out;
   }
   /**
    * Classify an unknown character when no extractor matches.
@@ -5592,6 +5628,14 @@ var _BaseTokenizer = class _BaseTokenizer {
     return null;
   }
 };
+/**
+ * ASCII word of the shape the English word-walker produces. Excludes `:`, so a
+ * token that already carries a qualifier never merges again — `a:b:c` yields
+ * `a:b` + `:c`, byte-matching the English extractor's single-segment merge.
+ */
+_BaseTokenizer.ASCII_WORD = /^[A-Za-z_][A-Za-z0-9_]*$/;
+/** `:name` — only a variable-ref-style extractor ever emits this token shape. */
+_BaseTokenizer.COLON_QUALIFIER = /^:[A-Za-z_][A-Za-z0-9_]*$/;
 /**
  * Configuration for native language time units.
  * Maps patterns to their standard suffix (ms, s, m, h).