Michael Milton

A prize to the first person who can successfully allocate a >1 quettabyte (10^30 bytes) vector in #rstats and manage to call object.size() on it! (I take no responsibility for bankruptcy from cloud charges and/or murder by sysadmin that may result from attempting this 😅)

Today I encountered the extremely rare blood-red #mamba snake. Surely an ill-omen for the fate of my #conda environments.

Not if they actually read the document

Just another day using the programming language with the world's shortest name. #RStats

I briefly looked at that, and I think because that function tries to access slots within S4 classes, it won't work well for a few purposes, such as Matrix::sparseMatrix which is an S4 class.

It's endlessly annoying that #RStats's native pipe doesn't support piping into operators. You can almost hear the record scratch sound effect when you realise you need to use [ or [[ and have to rewrite your 20-line pipeline into two separate statements.

Love to see @CSIRO using two of my favourite technologies, Dask (@dask_dev) and Apptainer (@SingularityApp) at the #eResAU2022 Conference! Link: conference.eresearch.edu.au/…

Today Salmon (RNA-seq) and @MultiQC conspired to create my first git branch that comes with the risk of tapeworm infection. #bioinformatics

Inspired by a recent thread I've now added support for the rest of the main @uniprot APIs, including ID mapping, allowing you to convert between any of the accession formats supported by UniProt! Many other improvements as well in version 1.0.0 of Unipressed: #bioinformatics

I would go with the id-mapper Uniprot API. My wrapper library doesn't yet support this, since it's separate to the new REST API, but it's still useful:

Re-branding Allen's Killer Python lollies as "YumPy Arrays" because they're already segmented into elements, and selling them to non-Australians who have no idea what I'm talking about

I've developed a #python package for querying UniProt's new REST API! Maybe the first to fully support the new format. Check it out at github.com/multimeric/Unipre…. In particular I've tried hard to integrate with Python tooling, giving you great code completion: #bioinformatics

I can't say I'm a fan of the fastapi/typer issue template. If you want to want to file any issue at all, even if it's a critical bug, you have to commit to helping with unrelated issues. I'm very aware of the demands on maintainers, but punishing bug reports isn't the way.

I increasingly think "how well does this function autocomplete" is a crucial factor in #python API design. So I'm inclined to use dataclasses as argument objects, since they give you great IDE autocomplete, unlike **kwargs, but also let you forward the args to another function:

It took me a while to grasp why #python's Concatenate type can be used to define a higher order function that can not only add, but also remove arguments from another function's signature. But eventually it clicked that it works the same as Python's tuple unpacking:

If you are using #python packages with incomplete types, you can make a stubs directory, recreate the package hierarchy, and populate it with .pyi type stubs. Then you just configure the stubs directory with your tools and both autocomplete and type checking will magically work!