Clj-kondo just got 10% faster due to switching from a map to a record representation for "var usages"
This post inspired me to try the same thing on my TCP server library, TeensyP, along with some other micro-optimisations such as eliminated boxed arithmetic. I managed to get over a 20% speedup, so kudos for reminding me about this!
Curious if the 1.13 array map size performance improvement makes any difference
I havenβt upgraded
it was very interesting to read through the PR; such a simple change but large perf gain
Iβm curious though is there a specific reason for putting the ADR for this under doc/ai/adr instead of the existing /doc/adr? Was the doc and PR LLM generated?
The doc is LLM generated, so I can lock down my decisions and they don't come up again in a code review
I put "ai" there so people know they are dealing with AI generated content