MarkdownPlugin: serialize \n within a text child of a paragraph as line break

[dschoorl]

See #4834 for background

Checklist

• yarn typecheck
• yarn lint:fix
• yarn test
• yarn brl
• yarn changeset
• ui changelog

[codesandbox]

Review or Edit in CodeSandbox

Open the branch in Web Editor • VS Code • Insiders

Open Preview

[changeset-bot]

🦋 Changeset detected

Latest commit: 281ca4d

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
@platejs/markdown	Patch
@platejs/ai	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
plate	Ready	Preview, Comment	Apr 10, 2026 10:09pm

[zbeyens]

Thanks for the PR. Looks like test is failing.

[dschoorl]

Thanks for the PR. Looks like test is failing.

Yes, I noticed. Those tests pass on my local machine. I am looking into it.

[JiwaniZakir]

The handling of empty parts produced by consecutive or trailing \n characters has a subtle bug. When splitting 'text\n' (text ending with a newline), split('\n') produces ['text', ''], which causes the loop to emit [text:'text', break, break] — two break nodes instead of one — because the non-empty branch unconditionally appends a break before the empty trailing part, and then the empty-string branch appends another. A test covering a text node with a trailing \n would expose this.

The test description also has a minor mismatch: it says "two empty lines" but the input string 'Text followed by two empty lines\n\n\nFollowed by more text.' contains three \n characters, not two, and the expected output has three \\\n sequences. The test content itself is correct, but the description will mislead future readers.

Finally, the condition childParts.length !== index + 1 for "not last element" in defaultRules.ts is clearer as index < childParts.length - 1, which is the conventional idiom for this pattern in the codebase.

[dschoorl]

Thanks for the review! See my comments below your feedback.

The handling of empty parts produced by consecutive or trailing \n characters has a subtle bug. When splitting 'text\n' (text ending with a newline), split('\n') produces ['text', ''], which causes the loop to emit [text:'text', break, break] — two break nodes instead of one — because the non-empty branch unconditionally appends a break before the empty trailing part, and then the empty-string branch appends another. A test covering a text node with a trailing \n would expose this.

Good catch. I fixed this and added some tests to verify that the mdast is now created as expected.

The test description also has a minor mismatch: it says "two empty lines" but the input string 'Text followed by two empty lines\n\n\nFollowed by more text.' contains three \n characters, not two, and the expected output has three \\\n sequences. The test content itself is correct, but the description will mislead future readers.

It takes three \n to create two empty lines, esp. when the first \n appears on a line with text, so that is not an empty line.

Finally, the condition childParts.length !== index + 1 for "not last element" in defaultRules.ts is clearer as index < childParts.length - 1, which is the conventional idiom for this pattern in the codebase.

Done.