Re: [boost] [Fit] Louis Dionne's formal review

Subject: Re: [boost] [Fit] Louis Dionne's formal review
From: Louis Dionne (ldionne.2_at_[hidden])
Date: 2016-03-12 16:38:01

Paul Fultz II <pfultz2 <at> yahoo.com> writes:

> 
> 
> [...]
> 
> Thanks Louis for the review. Do you have any feedback/thoughts about the
> compile-time performance of the library?

It's hard to tell without doing benchmarks, which I didn't do. Things I might
suggest benchmarking would be creating large `fit::pack`s and `fit::capture`s,
using `compress` with many arguments and using `fit::arg` with many arguments.

Just looking at the implementation quickly, it seems OK. For example `pack`
does not use a recursive implementation and all the functions I looked at 
used flat variadic expansion whenever possible instead of recursion. 
The C++11/14 way, basically. However, `detail::seq` is implemented using
recursion. It's not dramatic, but you could use a logarithmic approach 
instead to avoid hitting template instantiation depth limits.

Regards,
Louis