search

luizirber / ukhs

Universal K-mer Hitting Set library in Rust
Apache License 2.0
10 stars 0 forks source link

Compare bbhash crate with rust-boomphf #1

Open luizirber opened 5 years ago

luizirber commented 5 years ago

I ended up wrapping BBHash into two crates (bbhash-sys and bbhash) because I wanted to maintain compatibility with C/C++ codebases using it. But if rust-boomphf is compatible already I would rather not keep mine around and just use it.

proptest is probably a good option for doing some oracle tests and check this.

cc @k3yavi =]

k3yavi commented 5 years ago

@luizirber , I don't think there were any cycles put into making rust-boomphf to be C++ compatible, at least until when I was working on it. We have performed many checks and benchmarks for the rust-lib but as you have the C++ version ready, may be this is a good time to have a direct comparison for performance between the two versions 😜 . In my knowledge, the rust version of mphf was used and tested extensively in the production environment of 10x. I was using it mainly for rust-debruijn and having a rust version of mphf gave us more freedom to tweak, like different kmer-size for the keys, online creation of mpfh directly from 2bit-encoded de-bruijn graph etc. I was mainly coming more from the aim of pufferfish type indexing, which is still a work in progress 😉 ....