Merge pull request #3616 from xy200303/fix/nested-content-control-import

fix(super-editor): support nested content controls

Author: caio-pizzol

packages/super-editor/src/editors/v1/core/super-converter/v3/handlers/helpers/is-inline-node.js

@@ -24,6 +24,7 @@ const INLINE_FALLBACK_TYPES = new Set([
   'endnoteReference',
   'fieldAnnotation',
   'structuredContent',
+  'image',
   'mathInline',
   'passthroughInline',
   'page-number',

packages/super-editor/src/editors/v1/core/super-converter/v3/handlers/helpers/is-inline-node.test.js

@@ -8,6 +8,7 @@ describe('isInlineNode', () => {
     expect(isInlineNode({ type: 'bookmarkStart', attrs: { id: '1' } })).toBe(true);
     expect(isInlineNode({ type: 'bookmarkEnd', attrs: { id: '1' } })).toBe(true);
     expect(isInlineNode({ type: 'tab' })).toBe(true);
+    expect(isInlineNode({ type: 'image', attrs: { src: 'media/image1.png' } })).toBe(true);
     expect(isInlineNode({ type: 'footnoteReference', attrs: { id: '1' } })).toBe(true);
   });
 

packages/super-editor/src/editors/v1/core/super-converter/v3/handlers/w/sdt/helpers/handle-structured-content-node.js

@@ -1,6 +1,56 @@
 import { parseAnnotationMarks } from './handle-annotation-node';
 import { parseStrictStOnOff } from '../../../utils.js';
 import { BLOCK_FIELD_XML_NAMES } from '../../../sd/shared/block-field-xml-names.js';
+import { isInlineNode } from '../../../helpers/is-inline-node.js';
+
+const INLINE_CONTEXT_XML_NAMES = new Set(['w:p', 'w:r', 'w:hyperlink', 'w:smartTag']);
+
+function hasDirectBlockSignal(sdtContent) {
+  return Boolean(
+    sdtContent?.elements?.some(
+      (el) => el?.name === 'w:p' || el?.name === 'w:tbl' || BLOCK_FIELD_XML_NAMES.has(el?.name),
+    ),
+  );
+}
+
+function canEmitInlineStructuredContent(path = []) {
+  return path.some((entry) => INLINE_CONTEXT_XML_NAMES.has(entry?.name) || entry?.name === 'w:sdtContent');
+}
+
+function hasTranslatedBlockContent(content = [], schema) {
+  return content.some((node) => node?.type && !isInlineNode(node, schema));
+}
+
+function wrapInlineRunsAsParagraphs(content = [], schema) {
+  const normalized = [];
+  let pendingInline = [];
+
+  const flushInline = () => {
+    if (!pendingInline.length) return;
+    normalized.push({
+      type: 'paragraph',
+      attrs: null,
+      content: pendingInline,
+      marks: [],
+    });
+    pendingInline = [];
+  };
+
+  for (const node of content) {
+    if (!node) continue;
+
+    if (isInlineNode(node, schema)) {
+      pendingInline.push(node);
+      continue;
+    }
+
+    flushInline();
+    normalized.push(node);
+  }
+
+  flushInline();
+  return normalized;
+}
 
 /**
  * Detect the semantic control type from sdtPr child elements.
@@ -113,25 +163,25 @@ export function handleStructuredContentNode(params) {
     return null;
   }
 
-  const paragraph = sdtContent.elements?.find((el) => el.name === 'w:p');
-  const table = sdtContent.elements?.find((el) => el.name === 'w:tbl');
-  // SD-3005: a content control wrapping a block field (e.g. BIBLIOGRAPHY) has
-  // no direct w:p after preprocessing — its child is an sd:* block node. It is
-  // block content and must not be emitted as an inline structuredContent.
-  const blockField = sdtContent.elements?.find((el) => BLOCK_FIELD_XML_NAMES.has(el?.name));
   const { marks } = parseAnnotationMarks(sdtContent);
   const translatedContent = nodeListHandler.handler({
     ...params,
     nodes: sdtContent.elements,
     path: [...(params.path || []), sdtContent],
   });
 
-  const isBlockNode = paragraph || table || blockField;
+  const schema = params.editor?.schema;
+  const content = Array.isArray(translatedContent) ? translatedContent : [];
+  const isBlockNode =
+    hasTranslatedBlockContent(content, schema) ||
+    hasDirectBlockSignal(sdtContent) ||
+    !canEmitInlineStructuredContent(params.path);
   const sdtContentType = isBlockNode ? 'structuredContentBlock' : 'structuredContent';
+  const normalizedContent = isBlockNode ? wrapInlineRunsAsParagraphs(content, schema) : content;
 
   let result = {
     type: sdtContentType,
-    content: translatedContent,
+    content: normalizedContent,
     marks,
     attrs: {
       id: id?.attributes?.['w:val'] || null,

packages/super-editor/src/editors/v1/core/super-converter/v3/handlers/w/sdt/helpers/handle-structured-content-node.test.js

@@ -1,6 +1,8 @@
-import { describe, it, expect, vi, beforeEach } from 'vitest';
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
 import { handleStructuredContentNode } from './handle-structured-content-node';
 import { parseAnnotationMarks } from './handle-annotation-node';
+import { defaultNodeListHandler } from '../../../../../v2/importer/docxImporter.js';
+import { initTestEditor } from '@tests/helpers/helpers.js';
 
 // Mock dependencies
 vi.mock('./handle-annotation-node', () => ({
@@ -31,6 +33,10 @@ describe('handleStructuredContentNode', () => {
     parseAnnotationMarks.mockReturnValue({ marks: [] });
   });
 
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
   it('returns null when nodes array is empty', () => {
     const params = { nodes: [], nodeListHandler: mockNodeListHandler };
     const result = handleStructuredContentNode(params);
@@ -79,7 +85,7 @@ describe('handleStructuredContentNode', () => {
     const params = {
       nodes: [node],
       nodeListHandler: mockNodeListHandler,
-      path: [],
+      path: [{ name: 'w:p' }],
     };
 
     parseAnnotationMarks.mockReturnValue({ marks: [{ type: 'bold' }] });
@@ -363,3 +369,177 @@ describe('handleStructuredContentNode', () => {
     });
   });
 });
+
+describe('handleStructuredContentNode nested SDT import regression', () => {
+  let editor;
+
+  const textRun = (text) => ({
+    name: 'w:r',
+    elements: [{ name: 'w:t', elements: [{ type: 'text', text }] }],
+  });
+
+  const paragraph = (text) => ({
+    name: 'w:p',
+    elements: [textRun(text)],
+  });
+
+  const sdtPr = ({ id, tag, alias, lockMode = 'unlocked', controlType = 'w:richText' }) => ({
+    name: 'w:sdtPr',
+    elements: [
+      { name: 'w:id', attributes: { 'w:val': id } },
+      { name: 'w:tag', attributes: { 'w:val': tag } },
+      { name: 'w:alias', attributes: { 'w:val': alias } },
+      { name: 'w:lock', attributes: { 'w:val': lockMode } },
+      { name: controlType },
+    ],
+  });
+
+  const sdt = (props, contentElements) => ({
+    name: 'w:sdt',
+    elements: [sdtPr(props), { name: 'w:sdtContent', elements: contentElements }],
+  });
+
+  const table = (text) => ({
+    name: 'w:tbl',
+    elements: [
+      {
+        name: 'w:tblPr',
+        elements: [{ name: 'w:tblW', attributes: { 'w:w': '2400', 'w:type': 'dxa' } }],
+      },
+      {
+        name: 'w:tblGrid',
+        elements: [{ name: 'w:gridCol', attributes: { 'w:w': '2400' } }],
+      },
+      {
+        name: 'w:tr',
+        elements: [
+          {
+            name: 'w:tc',
+            elements: [
+              {
+                name: 'w:tcPr',
+                elements: [{ name: 'w:tcW', attributes: { 'w:w': '2400', 'w:type': 'dxa' } }],
+              },
+              paragraph(text),
+            ],
+          },
+        ],
+      },
+    ],
+  });
+
+  const importNodes = (nodes) => {
+    const nodeListHandler = defaultNodeListHandler();
+    return nodeListHandler.handler({
+      nodes,
+      nodeListHandler,
+      docx: {},
+      editor,
+      path: [],
+    });
+  };
+
+  const expectSchemaValid = (content) => {
+    let pmDoc;
+    expect(() => {
+      pmDoc = editor.schema.nodeFromJSON({ type: 'doc', content });
+      pmDoc.check();
+    }).not.toThrow();
+    return pmDoc;
+  };
+
+  const findFirstJson = (node, predicate) => {
+    if (!node) return null;
+    if (predicate(node)) return node;
+    for (const child of node.content || []) {
+      const found = findFirstJson(child, predicate);
+      if (found) return found;
+    }
+    return null;
+  };
+
+  beforeEach(() => {
+    ({ editor } = initTestEditor({
+      isHeadless: true,
+      loadFromSchema: true,
+      content: { type: 'doc', content: [{ type: 'paragraph' }] },
+    }));
+    parseAnnotationMarks.mockReturnValue({ marks: [] });
+  });
+
+  afterEach(() => {
+    editor?.destroy();
+    editor = null;
+    vi.restoreAllMocks();
+  });
+
+  it('imports nested block SDT when outer sdtContent directly contains w:sdt wrapping a paragraph', () => {
+    const inner = sdt({ id: 'inner-block', tag: 'inner-tag', alias: 'Inner Alias', lockMode: 'contentLocked' }, [
+      paragraph('Nested paragraph'),
+    ]);
+    const outer = sdt({ id: 'outer-block', tag: 'outer-tag', alias: 'Outer Alias', lockMode: 'sdtLocked' }, [inner]);
+
+    const result = importNodes([outer]);
+
+    expect(result).toHaveLength(1);
+    expect(result[0].type).toBe('structuredContentBlock');
+    expect(result[0].attrs).toMatchObject({
+      id: 'outer-block',
+      tag: 'outer-tag',
+      alias: 'Outer Alias',
+      lockMode: 'sdtLocked',
+      controlType: 'richText',
+    });
+
+    const nested = result[0].content?.[0];
+    expect(nested?.type).toBe('structuredContentBlock');
+    expect(nested.attrs).toMatchObject({
+      id: 'inner-block',
+      tag: 'inner-tag',
+      alias: 'Inner Alias',
+      lockMode: 'contentLocked',
+      controlType: 'richText',
+    });
+    expect(nested.attrs.sdtPr?.elements?.find((el) => el.name === 'w:alias')?.attributes?.['w:val']).toBe(
+      'Inner Alias',
+    );
+
+    expectSchemaValid(result);
+  });
+
+  it('wraps nested inline SDT safely when an outer block SDT also contains paragraph and table content', () => {
+    const inlineNested = sdt(
+      { id: 'inner-inline', tag: 'inline-tag', alias: 'Inline Alias', lockMode: 'sdtContentLocked' },
+      [textRun('Inline value')],
+    );
+    const outer = sdt({ id: 'outer-mixed', tag: 'outer-mixed-tag', alias: 'Outer Mixed', lockMode: 'sdtLocked' }, [
+      inlineNested,
+      paragraph('Outer paragraph'),
+      table('Cell text'),
+    ]);
+
+    const result = importNodes([outer]);
+
+    expect(result).toHaveLength(1);
+    expect(result[0].type).toBe('structuredContentBlock');
+    expect(result[0].content?.map((node) => node.type)).toEqual(['paragraph', 'paragraph', 'table']);
+
+    const nested = findFirstJson(
+      result[0],
+      (node) => node.type === 'structuredContent' && node.attrs?.id === 'inner-inline',
+    );
+    expect(nested).toBeTruthy();
+    expect(nested.attrs).toMatchObject({
+      id: 'inner-inline',
+      tag: 'inline-tag',
+      alias: 'Inline Alias',
+      lockMode: 'sdtContentLocked',
+      controlType: 'richText',
+    });
+    expect(nested.attrs.sdtPr?.elements?.find((el) => el.name === 'w:lock')?.attributes?.['w:val']).toBe(
+      'sdtContentLocked',
+    );
+
+    expectSchemaValid(result);
+  });
+});

packages/super-editor/src/editors/v1/tests/data/sdt-fixtures.README.md

@@ -0,0 +1,48 @@
+# SDT classification fixtures (PR #3616)
+
+Real `.docx` fixtures that validate the nested content-control classifier in
+`super-converter/v3/handlers/w/sdt/`. Exercised by
+`tests/editor/sdt-nested-classification.test.js`.
+
+The claim under test: block vs run/inline SDT classification is driven by the
+translated ProseMirror content shape plus import context, not only by the direct
+XML child names of `w:sdtContent`.
+
+## Provenance and conformance
+
+Each fixture's surrounding package (content types, rels, styles, theme, fonts, and
+image media) is taken verbatim from a Word-authored base already in this folder.
+Only `word/document.xml` is hand-authored to encode the exact OOXML shape, so the
+package stays valid while the structure is precise. All fixtures are therefore
+**schema-only** (hand-authored structure, not produced or validated by Word).
+
+The `conformance` column distinguishes shapes that are valid ECMA-376 from one that
+is deliberately malformed to exercise the PR's defensive normalization:
+
+| Fixture | Conformance | Base package | Shape under `w:body` |
+|---|---|---|---|
+| `sdt-nested-block.docx` | conformant | `blank-doc.docx` | block `w:sdt` whose `w:sdtContent` directly contains a nested block `w:sdt` (no direct `w:p`) wrapping a paragraph. Legal: `EG_ContentBlockContent` permits `sdt`. |
+| `sdt-nested-inline.docx` | conformant | `blank-doc.docx` | `w:p` containing an inline `w:sdt` that contains a nested inline `w:sdt` of runs, between two text runs. Legal: `CT_SdtContentRun` is `EG_PContent`. |
+| `sdt-mixed-block.docx` | **defensive (malformed)** | `blank-doc.docx` | block `w:sdt` whose `w:sdtContent` holds a bare inline `w:sdt`, a `w:p`, and a `w:tbl`. The bare inline `w:sdt` is **non-conformant**: a `w:sdt` directly under block content is positionally `CT_SdtBlock`, whose content may not be a bare `w:r` (`EG_ContentBlockContent` allows only `customXml/sdt/p/tbl/EG_RunLevelElts`, and `EG_RunLevelElts` excludes `w:r`). Included on purpose to drive `wrapInlineRunsAsParagraphs`, which the PR uses to normalize bare inline content inside a block SDT. |
+| `sdt-inline-picture.docx` | conformant | `anchor_images.docx` (reuses `media/image1.png`, `rId4`) | `w:p` > inline `w:sdt` with `<w:picture/>` marker > `w:sdtContent` > `w:r` > `w:drawing`. Legal per ECMA-376 §17.5.2.24 (picture content control wrapping a single DrawingML picture). |
+
+## Rebuild
+
+```
+node packages/super-editor/src/editors/v1/tests/data/sdt-fixtures.generate.cjs
+```
+
+The generator resolves all paths from its own location and reads the two base
+packages (`blank-doc.docx`, `anchor_images.docx`) from this folder, so it is
+portable. It re-reads each built file and asserts the intended shape. Set
+`SDT_FIXTURE_OUT=/some/dir` to write to a scratch dir instead of overwriting the
+committed fixtures (useful for a dry run). Regeneration is content-equivalent;
+only zip metadata may differ.
+
+## Out of scope
+
+Row-level SDTs (`w:tbl > w:sdt > w:sdtContent > w:tr`, Google Docs `goog_rdk_*`
+exports) are a separate table-walk concern tracked by SD-3118 / IT-1040. The real
+Google Docs artifact attached to those tickets should be used as that fixture, and
+its preservation checked through a Word round-trip before choosing transparent
+unwrap vs. `rowSdt` metadata. Not covered by these fixtures.