Use bucklescript string interpolation instead of ocaml stdlib

[thangngoc89]

As I can see, nice is a string concatenation library and you used ocaml function like string_of_int a lot. But calling these function should have some performance cost.

Example:

Link here https://reasonml.github.io/try/?reason=DYUwLgBAZg9jEF4IAoAeBKRA+CBnMATgJYB2A5gPoxQWlhqYDUjEARAA6qsDcAUKJABGAQwKIUqAFwQ6mBDgDeAKwA+AEgYqlAX25A

Bucklescript docs on string interpolation https://bucklescript.github.io/docs/en/common-data-types.html#interpolation

[thangngoc89]

I created a benchmark here https://rebench.github.io/?rebench-data=NoIgNgpgLgBAHjAvDALAJgNwHoBUMDOAFgIYBOEAJjAMYD2FEMA5rRPjIROTDliADTBQAZgEgAVvjEIA1DAA6IAA5xFIALqCQaMeTEBvcQB8AJAAo4AShVHxAXw1aAjLohj8UUgEsAdkwD6tABm-r5QFpYwMnKKKmrqCUA

Since I’m on mobile, the result maybe unreliable. Could any of you run it on a laptop and report it back here?

[threepointone]

thanks! I'm only just learning reason, so this is good to know. I'm trying to keep this to pure reason, so any idea how this will compile to native?

I also checked with rollup + closure(simple), and string_of_int gets inlined, making it as fast as the 3rd one.

I also messed with the benchmark, and it appears the advantage disappears when you're actually doing something with the value (note that the 3rd output doesn't return anything). eg - adding a console.log(value) brings them all closer together.

[ulrikstrid]

If you use the bucklescript specific implementation I don't think it will compile to native. Maybe bucklescript should compile string_of_int down to the second example if possible. cc @bobzhang

[thangngoc89]

After posting this issue, it appears that this is string_of_int:

function string_of_int (x) {
  return "" + x
}

So it's actually what I would write in JS. Sorry for the noises. So the only advantage of using {j I can think of is it's much easier to use

[threepointone]

I'll close this out for now, but feel free to carry on