Bug: A tab in inline code leaves the code span, and a paragraph of only whitespace code is dropped #9409

Lexical version: 0.52.0 (main at 98e936d gives the same results)

Steps to reproduce

Import each document below with $convertFromMarkdownString(markdown, TRANSFORMERS), then export it with $convertToMarkdownString(TRANSFORMERS). ⇥ is a tab:

Current behaviour

1. A tab inside a code span is imported as a TabNode without the code format (or any other format of the span), so it drops out of the code. The span is split around it, and a span holding only tabs stops being code.
2. A paragraph whose only content is inline code of up to three whitespace characters, such as ` ` or `⇥`, counts as an empty paragraph. Import removes it, and export treats it as empty when separating blocks. Code is content even when it's whitespace.

Cause

In packages/lexical-markdown/src/:

• $normalizeMarkdownTextNode() in MarkdownImport.ts replaces each tab in every imported text node with $createTabNode(), and only the surrounding text keeps the node's format and style.
• isEmptyParagraph() in utils.ts tests the text of a paragraph's only text node against /^\s{0,3}$/, without checking the code format. $importMarkdownNodes() in MarkdownImport.ts uses it to drop paragraphs, and createMarkdownExport() uses it to decide block separators.

Suggested fix

--- a/packages/lexical-markdown/src/MarkdownImport.ts
+++ b/packages/lexical-markdown/src/MarkdownImport.ts
@@ -292,7 +292,8 @@ function $importBlocks(
 function $normalizeMarkdownTextNode(textNode: TextNode): void {
   // A TabNode is a TextNode whose content is a tab, so without this guard the
   // rebuild below would destroy it and create an equivalent one in its place.
-  if ($isTabNode(textNode)) {
+  // Inline code keeps its tabs as text, so they stay part of the code span.
+  if ($isTabNode(textNode) || textNode.hasFormat('code')) {
     return;
   }
   const text = textNode.getTextContent();
--- a/packages/lexical-markdown/src/utils.ts
+++ b/packages/lexical-markdown/src/utils.ts
@@ -72,6 +72,8 @@ export function isEmptyParagraph(node: LexicalNode): boolean {
     firstChild == null ||
     (node.getChildrenSize() === 1 &&
       $isTextNode(firstChild) &&
+      // Inline code of only whitespace is content.
+      !firstChild.hasFormat('code') &&
       MARKDOWN_EMPTY_LINE_REG_EXP.test(firstChild.getTextContent()))
   );
 }

I kept the tab as text in the code node rather than giving the TabNode the code format. Separate code nodes on each side of a tab would be exported as adjacent code spans ( `a⇥b` ), which read back as one span with backticks in it.

With this change:

• `a⇥b` round-trips unchanged.
• `⇥` and a `⇥` b stay code and are stable. They're exported with padding (` ⇥ `), which a reader strips back to the tab.
• The paragraph of only ` ` is kept. Its export still grows by two spaces on every save; that's a separate padding bug in getCodeSpanDelimiter() reported in Bug: A code span of only spaces grows by two spaces on every save, and loses its bold or italic #9405.
• Tabs outside code (a⇥b) and in code blocks are unchanged. A truly empty paragraph is still dropped.

Tested with headless Lexical against main at 98e936d. The unit tests for lexical-markdown, lexical-mdast and lexical-playground/__tests__/unit pass (1175), along with tsc, ESLint and Prettier. No new unit tests are added. I didn't check how the editor handles a text node containing a tab when typing in a browser. Code blocks already hold tabs as TabNodes and aren't affected.

No existing issue found.