[@lexical/utils] Bug Fix: cursor positioning issue with CJK

[mrdivyansh]

Description

It fixes #6946. However, I am not sure why TextNode.updateDOM should cause any side-effect.

Closes #6946

[vercel]

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

Name	Status	Preview	Comments	Updated (UTC)
lexical	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Dec 12, 2024 10:27pm
lexical-playground	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Dec 12, 2024 10:27pm

[github-actions]

size-limit report 📦

Path	Size
lexical - cjs	31.09 KB (0%)
lexical - esm	30.96 KB (0%)
@lexical/rich-text - cjs	40.13 KB (0%)
@lexical/rich-text - esm	32.8 KB (0%)
@lexical/plain-text - cjs	38.71 KB (0%)
@lexical/plain-text - esm	30.12 KB (0%)
@lexical/react - cjs	42.05 KB (0%)
@lexical/react - esm	34.21 KB (0%)

[etrepum]

I think you're right that this updateDOM call doesn't really make sense. What updateDOM does is return true if the node should be re-mounted, otherwise it returns false and does some in-place mutations (which is what is likely causing the issues here). If it would've returned true you can detect that because that DOM would be unmounted and createDOM would be called to create a new one so currentAnchorNodeDOM !== previousAnchorNodeDOM would be true. I think the proper fix would be to track those previous DOM variables in the closure outside of the compute function just like previousAnchorNode and previousFocusNode. This way the updateDOM call can be eliminated but the intent of that check can be preserved.

/cc @zurfyx

[etrepum]

@mrdivyansh I created a draft PR #6961 with what I think this should look like, but I can't easily repro this bug so if you could take a look that would be great