Re: [lmi] Continuing deboostification with removing dependency on Boost.Regex

[Vadim Zeitlin]

On Thu, 27 May 2021 19:04:26 +0200 I wrote:

Me> 2. Rewrite test_coding_rules in some other language and use its built-in
Me>    regex support.
[...]
Me>    + Performance should be pretty good:
Me>      * Nim is based on the same PCRE as I'd like to use for C++ anyhow.
Me>      * Rust has a fast build-in regex implementation (used by ripgrep,
Me>        a.k.a. rg (https://github.com/BurntSushi/ripgrep), which is one grep
Me>        replacement that I've finally switched to after considering
Me>        switching to ack, ag, ... in the past -- but they just didn't seem
Me>        sufficiently better than grep to justify it, while rg definitely did
Me>        seem, and is).
Me>      * I'm more doubtful about Raku, as it's regex support is crazy
Me>        powerful and nice to use, but not necessarily known (at least to me)
Me>        for its performance prowesses.

 Just a small addendum to the last part: there is a benchmark of regex
implementations in various languages here that I found after posting the
original message which shows that both Rust and Nim greatly outperform
Boost.Regex in C++:

        https://github.com/mariomka/regex-benchmark/tree/optimized

 Of course, the benchmark is not checking for exactly the same kind of
regexes that we use, so there is no guarantee that we get the same kind of
speedup, but it's still pretty encouraging, and their results generally
make sense (although I have no idea how does Nim manage to get better
numbers than C, while using the same C library).

 Hope this can be useful,
VZ

pgppelPPNon2Y.pgp
Description: PGP signature