When working with large codebases (e.g. the GHC codebase) using HLS, inserting something as simple as a comment can sometimes feel noticeably laggy.
It seems this is due to multiple Shake restarts being triggered in quick succession, where each restart kills and spawns many threads. This repeated teardown/startup amplifies latency.
Proposal:
Consider merging consecutive Shake restarts that occur close together into a single restart, instead of handling each one individually. This should help smooth out the editing experience and reduce unnecessary thread churn.
When working with large codebases (e.g. the GHC codebase) using HLS, inserting something as simple as a comment can sometimes feel noticeably laggy.
It seems this is due to multiple Shake restarts being triggered in quick succession, where each restart kills and spawns many threads. This repeated teardown/startup amplifies latency.
Proposal:
Consider merging consecutive Shake restarts that occur close together into a single restart, instead of handling each one individually. This should help smooth out the editing experience and reduce unnecessary thread churn.