Closed artemrakov closed 4 months ago
But one thing I noticed if we use bigger data like 200 MB. The new code behaves significantly slower like 10x. Maybe we should add this code behind feature flag?
I think it is fine, it is not as big of a regression compared to the status quo, and it only happens in a less realistic case. (Realistically only when you are using stream decoder on an in-memory buffer, any kind of IO and you are unlikely to get so much data at a time).
I will try to make a better, more holistic fix soon-ish anyway.
Thanks. Released as 4.6.7!
Addressing the issue: https://github.com/Marwes/combine/issues/364
Avoid initializing continuously increasing buffers and limit the size to:
8 * 1024
But one thing I noticed if we use bigger data like 200 MB. The new code behaves significantly slower like 10x. Maybe we should add this code behind feature flag?
Before:
After:
Also did some tests in redis-rs lib: Before:
After: