Support for --dev

[andrader]

Hi, would be possible to support dev dependencies?

Like rv add --dev languageserver, rv sync --no-dev ...

To separate source dependencies from development dependencies.

[dpastoor]

yes! we want to add that - we are also considering the general concept of "profiles" where you might ahve a dev profile, a test profile etc, that might pull in other packages.

one admitted challenge is how complicated we make it - should we also be able to specifiy these dependencies come from other sources etc.

I think we can likely add something simple akin to a dev_dependencies array or the like.

One question for you as someone who has also identiifed this desire - meaningfully, how are the fact those dependencies exist or not impacting your workflow? is it wanting to pull in less dependencies for a CI run? or because they're not available in a certain environment? Is this for analysis based situations or apps?

Dev dependencies for packages and applications has always made a lot of sense to me, as you want things in your dev space you don't want to deploy into a production pipeline/app/connect/etc. But also trying to consider what "dev" might mean for a data scientist where rv is managing a data science project

[andrader]

So I might have that bias towargs wanting this because I come from a python swe background. Thats why I went looking for rv in the first place.

For me the main impact is keeping environments clean and reproducible.
To separate the packages I actually need for my research, from packages that I only need for development, like IRkernel and languageserver for example.
Have not been building R packages for a while, so CI have not been on my mind with R.
Even though, in general I do I want fewer dependencies so runs are lighter and less brittle, while locally I need extra tooling for testing, analysis, and experimentation.
Especially to keep things not conflicting.

But also trying to consider what "dev" might mean for a data scientist where rv is managing a data science project

I think data scientists and science in general would greatly benefit from more "clean" environments. Like what kedro proposes in python world, where it tries to help organize a DS workflow and allow for reproducibility in experiments, analysis and pipelines.

[payam-delfi]

+1 this request.

One question for you as someone who has also identiifed this desire - meaningfully, how are the fact those dependencies exist or not impacting your workflow? is it wanting to pull in less dependencies for a CI run? or because they're not available in a certain environment? Is this for analysis based situations or apps?

For our workflow, it's the former. Not having to install heavy packages like devtools or testthat, or formatting packages like styler or lintr.