What's broken?

Pressing the Delete key (forward-delete) when the cursor is at the end of the last block in the document throws:

RangeError: There is no position before the top-level node 

The error originates from getParentBlockInfo in mergeBlocks.ts (line 16).

This is a regression introduced in PR #2126 (v0.47.1). The old code had a depth guard:

if ($pos.depth <= 1) { return undefined; }

The new code calls $pos.before(depth) before any guard:

const depth = $pos.depth - 1; const parentBeforePos = $pos.before(depth); // crashes when depth === 0

When $pos.depth === 1, depth becomes 0, and $pos.before(0) throws because there is no position before the top-level document node.

What did you expect to happen?

Pressing Delete at the end of the last block should be a no-op (nothing to merge with), not throw an uncaught RangeError.

Steps to reproduce

1. Open any BlockNote editor
2. Type some text in a single block (e.g. "hello")
3. Place cursor at the end of the text
4. Press the Delete key (Fn+Backspace on Mac)
5. Open browser DevTools console → RangeError: There is no position before the top-level node

Or paste this in the browser console on any page with a BlockNote editor:

(() => { const el = document.querySelector('[contenteditable="true"]'); if (!el) return console.error('No editor found'); el.focus(); document.execCommand('insertText', false, 'hello'); const sel = window.getSelection(); sel.selectAllChildren(el); sel.collapseToEnd(); el.dispatchEvent(new KeyboardEvent('keydown', { key: 'Delete', code: 'Delete', bubbles: true })); console.log('Check console for RangeError'); })();

Suggested fix

Restore the depth guard before calling $pos.before(depth) in getParentBlockInfo:

export const getParentBlockInfo = (doc: Node, beforePos: number): BlockInfo | undefined => { const $pos = doc.resolve(beforePos); const depth = $pos.depth - 1; if (depth < 1) { return undefined; } const parentBeforePos = $pos.before(depth); // ...rest of function };

BlockNote version

0.47.1

Environment

All browsers (Chrome, Firefox, Safari). Reproducible in both browser and jsdom (vitest).

Additional context

Related to BLO-1012 and #2453. PR #2126 fixed forward-delete for most cases but introduced this regression for the last-block edge case.

Contribution

• I'd be interested in contributing a fix for this issue

Sponsor

• I'm a sponsor and would appreciate if you could look into this sooner than later 💖