[lexical-markdown][lexical-playground] Bug Fix: Keep indent when switching with markdown

[lytion]

Description

Reported issue: #7820
To summarize, if you use indent button and then switch to markdown, indent will be lost.

I decided to fix this issue by adding a transformer by I think it could also be fixed by handling it directly in createMarkdownExport and createMarkdownImport

This transformer simply add '\t' when exporting to markdown and remove those '\t' and set indent when importing from markdown. This only happen if '\t' is found at the beginning of the string and if the parent is a paragraph to avoid applying it to lists.

Closes #7820

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
lexical	Ready	Preview, Comment	Feb 11, 2026 8:46am
lexical-playground	Ready	Preview, Comment	Feb 11, 2026 8:46am

[etrepum]

Failing lint https://github.com/facebook/lexical/actions/runs/17919434734/job/50950181819?pr=7845

/home/runner/work/lexical/lexical/packages/lexical-markdown/src/index.ts
Error:   31:3  error  'INDENT' is defined but never used. Allowed unused vars must match /^_/u  @typescript-eslint/no-unused-vars

[etrepum]

Now the prettier check is failing https://github.com/facebook/lexical/actions/runs/17919849090/job/50951631139

You can run the integrity checks locally with npm run ci-check, and if the husky git commit hooks are installed then it will catch these formatting issues locally as well. You can fix the prettier issue with npm run prettier:fix

[lytion]

Is there some way to run tests in a linux environment or should I test it in a vm ?
Edit: Looks like updating the branch fix the e2e test

[etrepum]

This seems like maybe there's a bug elsewhere because tabs are usually supposed to use TabNode

[etrepum]

I noticed that this PR was updated with main but this comment hasn't been addressed

[lytion]

Sorry I missed the comment.
I think you're right, when clicking on "left indent" button in normal mode, there's no INDENT_CONTENT_COMMAND which I believe should happen and therefore should add a TabNode.

[lytion]

I want to make sure we're on the same page.
From what I see:
Markdown use \t for indent and tab key.
CodeHighlight convert \t into TabNode when converting to markdown.
Nodes use __indent and TabNode.
Here the transformer set the correct number of indent and remove all \t found from the start of the line.
If I understand correctly, what you're talking about is to convert all \t into TabNode even if CodeHighlight is disabled.
I can create a TAB transformer that does that, but maybe that's a separate issue

[etrepum]

Generally speaking there should never be '\t' or a '\n' in a TextNode, anything that creates a TextNode with those characters in the contents probably has a bug. The solution isn't a transformer, it's to fix the code that's creating those TextNodes. I don't think there is currently a helper to do this other than RangeSelection.insertRawText but one could be added to make this a little easier.

[lytion]

This is a quick and dirty refactor but is this the direction you are thinking about ?

diff --git a/packages/lexical-markdown/src/MarkdownImport.ts b/packages/lexical-markdown/src/MarkdownImport.ts
index d64037d91..851707205 100644
--- a/packages/lexical-markdown/src/MarkdownImport.ts
+++ b/packages/lexical-markdown/src/MarkdownImport.ts

+// Copy of LexicalSelection insertRawText function
+function insertRawText(text: string): LexicalNode[] {
+  const parts = text.split(/(\r?\n|\t)/);
+  const nodes = [];
+  const length = parts.length;
+  for (let i = 0; i < length; i++) {
+  const part = parts[i];
+  if (part === '\n' || part === '\r\n') {
+    nodes.push($createLineBreakNode());
+  } else if (part === '\t') {
+    nodes.push($createTabNode());
+  } else {
+    nodes.push($createTextNode(part));
+  }
+}
+  return nodes;
+}
+
 function $importBlocks(
   lineText: string,
   rootNode: ElementNode,
@@ -227,27 +248,39 @@ function $importBlocks(
   textMatchTransformers: Array<TextMatchTransformer>,
   shouldPreserveNewLines: boolean,
 ) {
-  const textNode = $createTextNode(lineText);
+  const nodes = insertRawText(lineText);
   const elementNode = $createParagraphNode();
-  elementNode.append(textNode);
+  elementNode.append(...nodes);
   rootNode.append(elementNode);
 
   for (const {regExp, replace} of elementTransformers) {
-    const match = lineText.match(regExp);
-
-    if (match) {
-      textNode.setTextContent(lineText.slice(match[0].length));
-      if (replace(elementNode, [textNode], match, true) !== false) {
+    for (const node of nodes) {
+      if (!$isTextNode(node)) {
         break;
       }
+      const match = node.getTextContent().match(regExp);
+
+      if (match) {
+        node.setTextContent(lineText.slice(match[0].length));
+        if (replace(elementNode, [node], match, true) !== false) {
+          break;
+        }
+      }
     }
   }
 
-  importTextTransformers(
-    textNode,
-    textFormatTransformersIndex,
-    textMatchTransformers,
-  );
+  for (const node of nodes) {
+    if (!$isTextNode(node)) {
+      break;
+    }
+    // TODO: Do this for TabNode too (maybe Line
+    importTextTransformers(
+      node,
+      textFormatTransformersIndex,
+      textMatchTransformers,
+    );
+  }
+

[etrepum]

Hard to say without a thorough audit of that code but probably something like that. It might be the case that it’s better to fix it up afterwards in the markdown case depending on how that code works.