Why is limit u64 instead of usize?

[kriswuollett]

Given that non-streaming results need to go into a Vec, shouldn't limit use usize instead? If it was usize then it would be more convenient to compare the length of returned results versus the limit. Use case is over fetching by one to determine whether or not to return a query continuation token in an API response.

For use case of streaming more than usize, then would it be worth it it just have a tradeoff of not specifying the limit and just dropping the stream when done determined by the client?

Is there any particular reason why u64 is used?

[tyt2y3]

Thank you for your question. I assume you mean QuerySelect::limit. I think we've had the discussion somewhere in SeaQuery. Basically, usize is platform dependent, it's 32 bits on 32 bit systems. This matters in parameter binding. We could hide this fact and cast it to u64 internally anyway, instead of exposing this in the API. In practice, we just cast everything back and forth between u64 and usize.

Casting is common in Rust. I do find myself casting the i of enumerate into i32 a lot, for example. This is not specific to SeaORM, some Rust libraries want to be more memory efficient uses u32 for len explicitly. Ref: https://docs.rs/arraystring/latest/arraystring/prelude/struct.ArrayString.html#method.len