Question: steps to production from this project about mini-redis HOT 4 OPEN

Perhaps the most important part missing in being production-ready is an intention to support the project in the future. If someone reports an issue with this project, and we don't believe actually fixing it would improve the project in the direction of being a learning resource, it might just turn into another one of those comments you mentioned.

As for your lru frame, it seems like the parsing spends the majority of the time inside a call to Bytes::copy_from_slice.

from mini-redis.

Looking at the code,

from mini-redis.

You would have to change several other parts of the codebase to elide that copy if it's possible at all.

from mini-redis.

I try to make the parsing code without copy:
https://github.com/photoszzt/cache_server/blob/522a2fc3fcb8ba57fd7e24a7505b41cd60a0321e/src/frame.rs#L274-L332
I change the parse in connections.rs to https://github.com/photoszzt/cache_server/blob/522a2fc3fcb8ba57fd7e24a7505b41cd60a0321e/src/connection.rs#L102-L132

My modified version records the start and end position in the slice and uses split_to and freeze to get Bytes from BytesMut.
But I find that this modification is leaking memory. I'm running a workload that inserts key/value and deletes them. The one that uses copy_from_slice doesn't leak memory. Is there anything wrong on this modification?

from mini-redis.