Post

Doing some read clustering. The polished result is just directly visible :)

Currently we use needletail for IO, but I know that paraseq does have a bunch more backends. Not quite sure if it handles them completely transparently though. Either way the rust library API is quite minimalistic so easy to use and integrate in principle.

Interesting. I think what makes most sense that the API directly returns true/false whether a row matches with cost <=k. Or maybe it returns Option<Match> with the first or best match found. I don't think I have the experience and time to make such a plugin right now, but would be happy to help :)

>does it support multithreading? yes, polars has its own global allocator which is based on rayon, so it manages all reading and compute together. You can have a look at e.g. polars-distance for an inspiration: github.com/ion-elgreco/...

main reason for me to implement such plugins, apart from fun, is getting automatic access to many tabular formats (via e.g. polars-bio: github.com/biodatageeks...) and their lazy streaming from s3, gcp and so on, which is extremely useful in more production-like settings.

The library supports ASCII, but for amino acids you pretty much *need* the scoring matrices right? Currently I don't do scoring in _any_ of my libraries; it's the one too many degree of freedom unfortunately. But it has been requested many times... (cc @rickbitloo.bsky.social ;)