To much String::clone()?

[matthiasbeyer]

In src/value.rs (https://github.com/mehcode/config-rs/blob/master/src/value.rs#L148) we clone() a String. That leads to a lot of String cloning when using this library.

Yesterday, I played around with cargo-flamegraph and cargo-bench a bit because of this. I am by no means an expert on how to do performance analysis and these kind of things, but from what I could see, String::clone() is not insignificant in the runtime cost of this crate.
Maybe, and I want to empathize this, maybe we could gain a lot from removing this one clone() call.

BUT the problem is, that it is not simple! I tried a bit, and using a &'a str in this place would basically put a lifetime on everything in the crate.
In some places, it is not even possible to have a lifetime... maybe I should have invested more time to gain some results,... I am not sure.

Maybe someone else has an idea, or is better with driving such analysis...

[szarykott]

Allow me to say something about this matter :)

Since most usually configurations are either created once at the beginning of application's lifetime or refreshed quite rarely later on, it shall not be a problem for users that library is allocating slightly too much. What would be a problem is non-intuitive API that could be the result of optimizations.

I am not saying that the issue should be disregarded, but it should be low-priority. If there are some easy wins - why not, but IMHO clean API is more important in this kind of library. The only problem is that this is Rust and Rust equals top performance, so something might be done to live up to its reputation.