polazarus / hipstr Goto Github PK

Arc<str> compares pointers first which normal reference equality checks do not perform.

See https://htmlpreview.github.io/?https://github.com/epage/string-benchmarks-rs/blob/master/runs/2023-10-10/eq/report/index.html for a benchmark showcasing the behavior from this

https://docs.rs/hipstr/0.4.0/hipstr/bytes/struct.HipByt.html#method.concat

If I have a [&str; 5] this would use 80 bytes to represent the string data correct? But with a HipStr if the actual data were 23 bytes or less I could compose that representation as a single HipStr and avoid allocating a new String?

I'm not sure if that would be any different than just creating a new HipStr and doing push_str()?:

let iter = ["key".as_bytes(), "value".as_bytes()];
iter.fold(HipByt::new(), |mut s, v| { s.push_slice(v); s })

let iter = ["key", "value"].into_iter();
iter.fold(HipStr::new(), |mut s, v| { s.push_str(v); s })

I've seen various methods for HipStr and HipByt that will output their type from a slice of the existing data, but not much for constructing the type from multiples of the type (or &str / &[u8]).

let parts = [key, "=\"", value, "\""];

// key="value"
let formatted = parts.concat();

Would it be worthwhile for HipStr and HipByt to support concat() to create an inline value more easily from multiple values?

In my case, I have a HashMap collection of KV pairs that I iterate through to create key="value" sequences, then the iterator composes those into a single String output.

That is fast with references (&str or &HipStr), and from what I understand this isn't something HipStr could improve on? So constructing a HipStr from an array of &str / &HipStr doesn't appear helpful due to the extra allocations/operations required?

How to get owned HipStr when using lines or split_whitespace on original HipStr?

Currently, lines returns &str, it causes HipStr to make a copy of &str. I want a slice of the original data.