Is there a reason the weigher only supports u32, while max capacity is u64?

[GalLalouche]

Basically the title. I'm trying to use moka with size-based-evictions, as detailed in https://docs.rs/moka/latest/moka/future/struct.Cache.html#example-size-based-eviction, but the limitation on u32 entry weights makes this impractical for some use-cases.

[tatsuya6502]

Thank you for using moka.

It uses u32 for entry weight rather than u64 to reduce memory overhead. A cache instance must hold the weight values of all entries so a small data type (u32 4 bytes) is preferable than a large data type (u64 8 bytes). It uses u64 for the max capacity and weighted_size (the sum of the weight values of all entries) to avoid arithmetic overflow (e.g. u32::MAX + 1u32) . The memory overhead are small for them because a cache instance only needs to hold one u64 for each.

Entry weights are relative values so you can scale your original value to fit in the 2^32 range, and scale the max capacity as well.