Slow performance in a loop

[closte]

A few days ago I was testing the .NET C# integration of the UA Parser and I`m getting way faster execution time compared to your solution and Rust language.
Rust language is well known to be faster than C# in almost all cases.

Iterations: 1000
Rust (cargo run --release:): ~280ms for 1000 iterations
C#: ~130ms for 1000 iterations

Rust code looks like this:

let ua_parser = UserAgentParser::from_yaml(&yamlFile).expect("Parser creation failed");
let user_agent_string =  String::from("Mozilla/5.0 (X11; Linux x86_64; rv:2.0b8pre) Gecko/20101031 Firefox-4.0/4.0b8pre");
let mut user_agent = ua_parser.parse_user_agent(&user_agent_string);
for n in 1..1000{
        user_agent = ua_parser.parse_user_agent(&user_agent_string);
}

C# code looks like this:

 uaParser = Parser.FromYaml(fileContent, new ParserOptions{ UseCompiledRegex = true });
 int count = 1000;
         
for (int i = 0; i < count; i++){
     ClientInfo c = uaParser.Parse("Mozilla/5.0 (X11; Linux x86_64; rv:2.0b8pre) Gecko/20101031 Firefox-4.0/4.0b8pre");
}

I`m very newbie to Rust but maybe you can make all regex compiled/static like on the following URL or why the Rust implementation is almost double slower?

[hamiltop]

So all regexes are already compiled during initialization, so that's not the cause. The cause is likely unicode support in Rust, which makes regex much more complicated.

My fork/PR here #14 disables unicode support and provides a significant speedup as well as some memory savings. You can see an example of how to set up the parser without unicode support here: https://github.com/davidarmstronglewis/uap-rs/pull/14/files#diff-27d11095d02b2115142e677214a2e2791f4a6158c7e12c367338dfa8378ceed7

I'd love to see your comparison with this new code. The benchmark suite has a 3x+ speedup for the user_agent parsers, so hopefully that makes rust beat c# in your example.

[oceanlewis]

With #14 now merged (thank you @hamiltop) and version 0.6.1 released I would be curious where performance lands compared to your C# example.

Feel free to share your findings and if you're satisfied close this issue, @closte.

I can also imagine a silly optimization of memoizing user agent strings to avoid re-running the regex; but I'm not sure how much of a difference in performance and may not be worth the optimization.

[closte]

I made a simple test and the same 1000 iterations ends is just +-30ms, faster than C# and way, way faster than the previous code.

Thanks
@hamiltop
@davidarmstronglewis

[oceanlewis]

Seriously, great work @hamiltop 🙌
Gonna close this one as it seems to be much more inline with performance expectations.
Thank you! ❤️