Address review: single provider resolution + figure template for kind-less captions

- Resolve the providing page state once per container instead of two
  linear searches across states (target lookup + numbering selection)
- Containers without a kind are figures (kindFromNode); their caption
  label now honours numbering.figure.template instead of a hard-coded
  default

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: mmcky

packages/myst-transforms/src/enumerate.spec.ts

@@ -1286,6 +1286,21 @@ describe('code numbering (#47)', () => {
     expect(captionParagraph.children[0].type).toBe('captionNumber');
     expect(toText(captionParagraph.children[0])).toBe('Listing 1.1:');
   });
+  test('kind-less containers honour the figure template', () => {
+    const tree = u('root', [
+      u('container', { identifier: 'fig-1' }, [
+        u('caption', [u('paragraph', [u('text', 'My figure')])]),
+      ]),
+    ]);
+    const state = new ReferenceState('main.md', {
+      frontmatter: { numbering: { all: { enabled: true }, figure: { template: 'Fig. %s' } } },
+      vfile: new VFile(),
+    });
+    enumerateTargetsTransform(tree, { state });
+    addContainerCaptionNumbersTransform(tree, new VFile(), { state });
+    const captionParagraph = (tree as any).children[0].children[0].children[0];
+    expect(toText(captionParagraph.children[0])).toBe('Fig. 1:');
+  });
   test('caption noun falls back to the default without a template', () => {
     const tree = u('root', [
       u('container', { kind: 'code', identifier: 'list-1' }, [

packages/myst-transforms/src/enumerate.ts

@@ -1033,10 +1033,12 @@ export const enumerateTargetsPlugin: Plugin<[StateOptions], GenericParent, Gener
 function getCaptionLabel(kind?: Container['kind'], subcontainer?: boolean, numbering?: Numbering) {
   if (subcontainer && (kind === 'equation' || kind === 'subequation')) return `(%s)`;
   if (subcontainer) return `({subEnumerator})`;
-  if (!kind) return 'Figure %s:';
   // The caption noun follows the configured template, matching what
-  // cross-references render (e.g. `numbering.code.template: "Listing %s"`)
-  const template = numbering?.[kind]?.template ?? getDefaultNumberedReferenceTemplate(kind);
+  // cross-references render (e.g. `numbering.code.template: "Listing %s"`);
+  // containers without a kind are figures (matching kindFromNode)
+  const effectiveKind = kind || 'figure';
+  const template =
+    numbering?.[effectiveKind]?.template ?? getDefaultNumberedReferenceTemplate(effectiveKind);
   return `${template}:`;
 }
 
@@ -1056,7 +1058,11 @@ export function addContainerCaptionNumbersTransform(
   containers
     .filter((container: Container) => container.enumerator)
     .forEach((container: Container) => {
-      const target = opts.state.getTarget(container.identifier)?.node;
+      // Resolve the providing page state once: it serves both the target
+      // lookup and the numbering selection below. Single-page states do not
+      // resolve without a page argument but carry the numbering directly.
+      const stateProvider = opts.state.resolveStateProvider(container.identifier);
+      const target = (stateProvider ?? opts.state).getTarget(container.identifier)?.node;
       if (!target?.enumerator) return;
       // Only look for direct caption children
       let para = select(
@@ -1080,11 +1086,8 @@ export function addContainerCaptionNumbersTransform(
           html_id: (container as any).html_id,
           enumerator: target.enumerator,
         };
-        // Page-level resolvers need the identifier's page; a single-page
-        // ReferenceState carries the numbering directly
         const numbering =
-          opts.state.resolveStateProvider(container.identifier)?.numbering ??
-          (opts.state as { numbering?: Numbering }).numbering;
+          stateProvider?.numbering ?? (opts.state as { numbering?: Numbering }).numbering;
         fillReferenceEnumerators(
           file,
           captionNumber,