Conversation
Fixed three critical bugs that caused progressive performance degradation: - Extension.dispose() now properly cleans up subscriptions and resources to prevent event listener accumulation on reload - Fixed memory leak in editors.ts where array length comparison was used instead of assignment (length === 0 -> length = 0) - Fixed selection update race condition by deferring reveal() to prevent cursor jumping during navigation 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
Fixed several secondary performance and reliability issues: - Replaced setTimeout with queueMicrotask in mode changes for predictable event listener setup - Fixed stale document references by capturing document at function start in selectionsToCharacterMode and selectionsFromCharacterMode - Added disposal guard to prevent initialization race in Editors class These changes improve stability and prevent potential edge cases that could contribute to performance degradation. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
|
Hey there! Thanks for the investigation and fix. Some of the findings don't seem to make a lot of sense to me (like replacing a lazy Edit: I'm actually not sure the memory leak comes from here. We do leak memory there, but the array is cleared every time we open an editor, so in practice the array stays very small. Footnotes
|
|
I looked at the changes and ended up submitting #416 with some of them. In my case, this didn't solve the performance problem, but it's still nice to get rid of some memory leaks. Thanks! |
2 participants
I've been using Dance quite a lot recently, but sometimes it started to lag excessively, particularly after I would trigger a conflict between code and dance handling the cursor.
I'm not keen on delving into a lot of typescript and I'm not familiar with the vscode extension model, so I just asked Claude to look around for possible causes. The
===->=looks like a legit mistype, but I have no clue about the rest.My limited testing for now makes me think that it solved the issue; can somebody with experience with the codebase chime in and review it? The first commit should directly address the issue, the second one is more things there were found.
Probably fixes #343 (comment) (or #343 altogether)