fix: splitting run with header adds empty row (#2229)

harbournick · web-flow · commit e1965fc0ccdd · 2026-02-28T12:49:37.000-08:00
diff --git a/packages/super-editor/src/core/commands/splitBlock.js b/packages/super-editor/src/core/commands/splitBlock.js
@@ -3,6 +3,23 @@ import { canSplit } from 'prosemirror-transform';
 import { defaultBlockAt } from '../helpers/defaultBlockAt.js';
 import { Attribute } from '../Attribute.js';
 
+const isHeadingStyleId = (styleId) => typeof styleId === 'string' && /^heading\s*[1-6]$/i.test(styleId.trim());
+
+const clearHeadingStyleId = (attrs) => {
+  if (!attrs || typeof attrs !== 'object') return attrs;
+  const paragraphProperties = attrs.paragraphProperties;
+  const styleId = paragraphProperties?.styleId;
+  if (!isHeadingStyleId(styleId)) return attrs;
+
+  const nextParagraphProperties = { ...paragraphProperties };
+  delete nextParagraphProperties.styleId;
+
+  return {
+    ...attrs,
+    paragraphProperties: nextParagraphProperties,
+  };
+};
+
 const ensureMarks = (state, splittableMarks) => {
   const marks = state.storedMarks || (state.selection.$to.parentOffset && state.selection.$from.marks());
   if (marks) {
@@ -62,11 +79,18 @@ export const splitBlock =
       if (can) {
         tr.split(tr.mapping.map($from.pos), 1, types);
 
-        if (deflt && !atEnd && !$from.parentOffset && $from.parent.type !== deflt) {
+        if (deflt && !atEnd && !$from.parentOffset) {
           const first = tr.mapping.map($from.before());
           const $first = tr.doc.resolve(first);
-          if ($from.node(-1).canReplaceWith($first.index(), $first.index() + 1, deflt)) {
-            tr.setNodeMarkup(tr.mapping.map($from.before()), deflt);
+          const shouldChangeType = $from.parent.type !== deflt;
+          const normalizedAttrs = clearHeadingStyleId($from.parent.attrs);
+          const shouldNormalizeAttrs = normalizedAttrs !== $from.parent.attrs;
+
+          if (
+            $from.node(-1).canReplaceWith($first.index(), $first.index() + 1, deflt) &&
+            (shouldChangeType || shouldNormalizeAttrs)
+          ) {
+            tr.setNodeMarkup(first, deflt, normalizedAttrs);
           }
         }
       }
@@ -79,19 +103,34 @@ export const splitBlock =
   };
 
 function deleteAttributes(attrs, attrsToRemove) {
-  const newAttrs = { ...attrs };
-  attrsToRemove.forEach((attrName) => {
+  let nextAttrs = { ...attrs };
+  for (const attrName of attrsToRemove) {
     const parts = attrName.split('.');
     if (parts.length === 1) {
-      delete newAttrs[attrName];
-    } else {
-      let current = newAttrs;
-      for (let i = 0; i < parts.length - 1; i++) {
-        if (current[parts[i]] == null) return;
-        current = current[parts[i]];
+      delete nextAttrs[attrName];
+      continue;
+    }
+
+    let source = nextAttrs;
+    for (let i = 0; i < parts.length - 1; i += 1) {
+      if (source == null || typeof source !== 'object') {
+        source = null;
+        break;
       }
-      delete current[parts[parts.length - 1]];
+      source = source[parts[i]];
     }
-  });
-  return newAttrs;
+
+    if (source == null || typeof source !== 'object') continue;
+
+    let target = nextAttrs;
+    for (let i = 0; i < parts.length - 1; i += 1) {
+      const key = parts[i];
+      const value = target[key];
+      target[key] = { ...value };
+      target = target[key];
+    }
+
+    delete target[parts[parts.length - 1]];
+  }
+  return nextAttrs;
 }
diff --git a/packages/super-editor/src/core/commands/splitBlock.test.js b/packages/super-editor/src/core/commands/splitBlock.test.js
@@ -233,5 +233,99 @@ describe('splitBlock', () => {
       // ensureMarks should NOT be called
       expect(mockTr.ensureMarks).not.toHaveBeenCalled();
     });
+
+    it('clears heading style on leading block when splitting at start of heading paragraph', () => {
+      const canReplaceWith = vi.fn(() => true);
+      const paragraphType = { name: 'paragraph', isTextblock: true, hasRequiredAttrs: vi.fn(() => false) };
+      const parentNode = {
+        contentMatchAt: vi.fn(() => ({
+          edgeCount: 1,
+          edge: vi.fn(() => ({ type: paragraphType })),
+        })),
+        canReplaceWith,
+      };
+      const headingAttrs = { paragraphProperties: { styleId: 'Heading2' } };
+      const $from = createMockResolvedPos({
+        depth: 1,
+        parent: {
+          isBlock: true,
+          content: { size: 10 },
+          type: { name: 'paragraph' },
+          inlineContent: true,
+          attrs: headingAttrs,
+        },
+        parentOffset: 0,
+        node: vi.fn((depth) => {
+          if (depth === -1) return parentNode;
+          return { type: { name: 'paragraph' }, attrs: headingAttrs };
+        }),
+      });
+      const $to = createMockResolvedPos({
+        pos: 5,
+        parent: { isBlock: true, content: { size: 10 }, type: { name: 'paragraph' }, inlineContent: true },
+        parentOffset: 0,
+      });
+
+      mockTr.selection = { $from, $to };
+      mockState.selection = mockTr.selection;
+      mockTr.doc = {
+        resolve: vi.fn(() => ({ index: vi.fn(() => 0) })),
+      };
+
+      const command = splitBlock();
+      command({ tr: mockTr, state: mockState, dispatch: () => {}, editor: mockEditor });
+
+      expect(mockTr.setNodeMarkup).toHaveBeenCalled();
+      const attrs = mockTr.setNodeMarkup.mock.calls[0][2];
+      expect(attrs.paragraphProperties?.styleId).toBeUndefined();
+    });
+
+    it('does not mutate source attrs when removing nested override attributes', () => {
+      const paragraphType = { name: 'paragraph', isTextblock: true, hasRequiredAttrs: vi.fn(() => false) };
+      const parentNode = {
+        contentMatchAt: vi.fn(() => ({
+          edgeCount: 1,
+          edge: vi.fn(() => ({ type: paragraphType })),
+        })),
+      };
+      const sourceAttrs = {
+        paragraphProperties: { styleId: 'Heading2', keep: true },
+        preserve: true,
+      };
+      const $from = createMockResolvedPos({
+        depth: 1,
+        parent: {
+          isBlock: true,
+          content: { size: 10 },
+          type: { name: 'paragraph' },
+          inlineContent: true,
+          attrs: sourceAttrs,
+        },
+        parentOffset: 0,
+        node: vi.fn((depth) => {
+          if (depth === -1) return parentNode;
+          return { type: { name: 'paragraph' }, attrs: sourceAttrs };
+        }),
+      });
+      const $to = createMockResolvedPos({
+        pos: 5,
+        parent: { isBlock: true, content: { size: 10 }, type: { name: 'paragraph' }, inlineContent: true },
+        parentOffset: 10,
+      });
+
+      mockTr.selection = { $from, $to };
+      mockState.selection = mockTr.selection;
+      mockTr.doc = {
+        resolve: vi.fn(() => ({ index: vi.fn(() => 0) })),
+      };
+
+      const command = splitBlock({ attrsToRemoveOverride: ['paragraphProperties.styleId'] });
+      command({ tr: mockTr, state: mockState, dispatch: () => {}, editor: mockEditor });
+
+      const splitTypes = mockTr.split.mock.calls[0][2];
+      expect(splitTypes?.[0]?.attrs?.paragraphProperties?.styleId).toBeUndefined();
+      expect(splitTypes?.[0]?.attrs?.paragraphProperties?.keep).toBe(true);
+      expect(sourceAttrs.paragraphProperties.styleId).toBe('Heading2');
+    });
   });
 });
diff --git a/packages/super-editor/src/core/helpers/importMarkdown.integration.test.js b/packages/super-editor/src/core/helpers/importMarkdown.integration.test.js
@@ -52,6 +52,23 @@ function paragraphByText(paragraphs, expectedText) {
 }
 
 describe('markdown to DOCX integration', () => {
+  it('retains blank lines between root blocks as empty paragraphs', () => {
+    const markdown = `First paragraph.
+
+
+Second paragraph.
+
+
+
+Third paragraph.`;
+
+    const doc = createDocFromMarkdown(markdown, editor);
+    const paragraphs = collectTopLevelParagraphs(doc);
+    const texts = paragraphs.map((node) => node.textContent);
+
+    expect(texts).toEqual(['First paragraph.', '', '', 'Second paragraph.', '', '', '', 'Third paragraph.']);
+  });
+
   it('converts complete markdown document with headings and lists', () => {
     const markdown = `# Main Title
 
diff --git a/packages/super-editor/src/core/helpers/markdown/mdastToProseMirror.ts b/packages/super-editor/src/core/helpers/markdown/mdastToProseMirror.ts
@@ -44,7 +44,7 @@ import type { MdastConversionContext, MarkdownDiagnostic } from './types.js';
  * suitable for constructing a full doc or a fragment.
  */
 export function convertMdastToBlocks(root: Root, ctx: MdastConversionContext): JsonNode[] {
-  return flatMapChildren(root, ctx);
+  return flatMapRootChildrenPreserveBlankLines(root, ctx);
 }
 
 // ---------------------------------------------------------------------------
@@ -68,6 +68,26 @@ interface JsonMark {
 // Block-level converters
 // ---------------------------------------------------------------------------
 
+function flatMapRootChildrenPreserveBlankLines(root: Root, ctx: MdastConversionContext): JsonNode[] {
+  const children = root.children ?? [];
+  const blocks: JsonNode[] = [];
+
+  for (let i = 0; i < children.length; i += 1) {
+    const child = children[i];
+    blocks.push(...convertBlockNode(child, ctx));
+
+    const next = children[i + 1];
+    if (!next) continue;
+
+    const blankLines = countBlankLinesBetweenSiblings(child, next);
+    for (let j = 0; j < blankLines; j += 1) {
+      blocks.push(makeParagraph([]));
+    }
+  }
+
+  return blocks;
+}
+
 function flatMapChildren(parent: MdastNode & { children?: MdastNode[] }, ctx: MdastConversionContext): JsonNode[] {
   if (!parent.children) return [];
   const blocks: JsonNode[] = [];
@@ -77,6 +97,20 @@ function flatMapChildren(parent: MdastNode & { children?: MdastNode[] }, ctx: Md
   return blocks;
 }
 
+function countBlankLinesBetweenSiblings(previous: MdastNode, next: MdastNode): number {
+  const previousEndLine = previous.position?.end?.line;
+  const nextStartLine = next.position?.start?.line;
+
+  if (typeof previousEndLine !== 'number' || typeof nextStartLine !== 'number') {
+    return 0;
+  }
+
+  // mdast line numbers are 1-based and inclusive:
+  //   previous ends on line A, next starts on line B.
+  // Lines strictly between them are explicit blank lines in the source.
+  return Math.max(0, nextStartLine - previousEndLine - 1);
+}
+
 function convertBlockNode(node: MdastNode, ctx: MdastConversionContext): JsonNode[] {
   switch (node.type) {
     case 'paragraph':
diff --git a/packages/super-editor/src/extensions/run/commands/split-run.js b/packages/super-editor/src/extensions/run/commands/split-run.js
@@ -5,6 +5,25 @@ import { defaultBlockAt } from '@core/helpers/defaultBlockAt.js';
 import { resolveRunProperties, encodeMarksFromRPr } from '@core/super-converter/styles.js';
 import { extractTableInfo } from '../calculateInlineRunPropertiesPlugin.js';
 
+function isHeadingStyleId(styleId) {
+  return typeof styleId === 'string' && /^heading\s*[1-6]$/i.test(styleId.trim());
+}
+
+function clearHeadingStyleId(attrs) {
+  if (!attrs || typeof attrs !== 'object') return attrs;
+  const paragraphProperties = attrs.paragraphProperties;
+  const styleId = paragraphProperties?.styleId;
+  if (!isHeadingStyleId(styleId)) return attrs;
+
+  const nextParagraphProperties = { ...paragraphProperties };
+  delete nextParagraphProperties.styleId;
+
+  return {
+    ...attrs,
+    paragraphProperties: nextParagraphProperties,
+  };
+}
+
 /**
  * Splits a run node at the current selection into two paragraphs.
  * @returns {import('@core/commands/types').Command}
@@ -99,11 +118,21 @@ export function splitBlockPatch(state, dispatch, editor) {
   }
   if (!can) return false;
   tr.split(splitPos, types.length, types);
-  if (!atEnd && atStart && $from.node(splitDepth).type != deflt) {
-    let first = tr.mapping.map($from.before(splitDepth)),
-      $first = tr.doc.resolve(first);
-    if (deflt && $from.node(splitDepth - 1).canReplaceWith($first.index(), $first.index() + 1, deflt))
-      tr.setNodeMarkup(tr.mapping.map($from.before(splitDepth)), deflt);
+  if (!atEnd && atStart) {
+    const first = tr.mapping.map($from.before(splitDepth));
+    const $first = tr.doc.resolve(first);
+    const sourceNode = $from.node(splitDepth);
+    const shouldChangeType = sourceNode.type != deflt;
+    const normalizedAttrs = clearHeadingStyleId(sourceNode.attrs);
+    const shouldNormalizeAttrs = normalizedAttrs !== sourceNode.attrs;
+
+    if (
+      deflt &&
+      $from.node(splitDepth - 1).canReplaceWith($first.index(), $first.index() + 1, deflt) &&
+      (shouldChangeType || shouldNormalizeAttrs)
+    ) {
+      tr.setNodeMarkup(first, deflt, normalizedAttrs);
+    }
   }
 
   applyStyleMarks(state, tr, editor, paragraphAttrs, tableInfo);
@@ -195,6 +224,10 @@ function applyStyleMarks(state, tr, editor, paragraphAttrs, tableInfo) {
   }
 }
 
+/**
+ * Splits the current run node into two sibling runs at the cursor position.
+ * @returns {import('@core/commands/types').Command}
+ */
 export const splitRunAtCursor = () => (props) => {
   let { state, dispatch, tr } = props;
   const sel = state.selection;
diff --git a/packages/super-editor/src/extensions/run/commands/split-run.test.js b/packages/super-editor/src/extensions/run/commands/split-run.test.js
@@ -170,6 +170,15 @@ describe('splitRunToParagraph command', () => {
 
     expect(handled).toBe(false);
   });
+
+  it('returns false for splitRunAtCursor when cursor is not in a run node', () => {
+    loadDoc(PLAIN_PARAGRAPH_DOC);
+    updateSelection(1);
+
+    const handled = editor.commands.splitRunAtCursor();
+
+    expect(handled).toBe(false);
+  });
 });
 
 describe('splitRunToParagraph with style marks', () => {
@@ -312,6 +321,37 @@ describe('splitRunToParagraph with style marks', () => {
     expect(paragraphTexts).toEqual(['Heading', ' Text']);
   });
 
+  it('clears heading style on the leading empty paragraph when splitting at heading start', () => {
+    const mockConverter = {
+      convertedXml: {},
+      numbering: {},
+      translatedNumbering: {},
+      translatedLinkedStyles: {},
+      documentGuid: 'test-guid-123',
+      promoteToGuid: vi.fn(),
+    };
+
+    editor.converter = mockConverter;
+    loadDoc(STYLED_PARAGRAPH_DOC);
+
+    const start = findTextPos('Heading Text');
+    expect(start).not.toBeNull();
+    updateSelection(start ?? 0);
+
+    const handled = editor.commands.splitRunToParagraph();
+    expect(handled).toBe(true);
+
+    const paragraphs = [];
+    editor.view.state.doc.descendants((node) => {
+      if (node.type.name === 'paragraph') paragraphs.push(node);
+    });
+
+    expect(paragraphs).toHaveLength(2);
+    expect(paragraphs[0].textContent).toBe('');
+    expect(paragraphs[0].attrs?.paragraphProperties?.styleId).toBeUndefined();
+    expect(paragraphs[1].attrs?.paragraphProperties?.styleId).toBe('Heading1');
+  });
+
   it('handles missing converter gracefully during split', () => {
     const mockConverter = {
       convertedXml: {},
diff --git a/tests/behavior/harness/main.ts b/tests/behavior/harness/main.ts
diff --git a/tests/behavior/tests/basic-commands/enter-page-boundary-line-count.spec.ts b/tests/behavior/tests/basic-commands/enter-page-boundary-line-count.spec.ts
diff --git a/tests/behavior/tests/basic-commands/fixtures/enter-page-boundary.md b/tests/behavior/tests/basic-commands/fixtures/enter-page-boundary.md
