>mar. 30 juin 2026 at 12:00, Arun Isaac <[email protected]> wrote:

> Hi Guix,
>
> I spoke about Guix at STEP-UP RSLondon 2026, a conference for research
> software engineers (RSEs) primarily based at UK universities. Abstract here:
> https://step-up.ac.uk/events/step-up-2026/abstracts/#software4 My
> thinking was to advocate for Guix in the primarily conda+pip dominated
> world of the RSE. I got some feedback that I'd like to share.
>
> I heard from someone who said that they tried Guix and were very
> frustrated with all the scheme file editing (manifest files, package
> definitions, etc.) that they had to do. They came from a python world
> where they are used to adding dependencies from the CLI—like `uv add
> packagename`. They felt that the CLI is the direction modern tooling has
> gone, and that's what they'd prefer.

Guix cli tools are rather complete, IMMO. For example:

guix shell -CW emacs-magit bash -- sh -c 'guix package -p $GUIX_ENVIRONMENT/ 
--export-manifest'

would return a manifest with inputs.

> Now, I know we pride ourselves on the expressive power of our scheme
> files. And, I absolutely prefer writing them. But, perhaps there's a
> case for providing a more declarative config—something like TOML—for
> simpler and common use cases.

Would WISP¹ be an option ? This is even closer to Python world.

--
Cayetano Santos

[1] https://hg.sr.ht/~arnebab/wisp

Attachment: signature.asc
Description: PGP signature

Reply via email to