Discussion:
categories for R packages
Brook Milligan
2014-09-05 16:15:32 UTC
Permalink
Some of you have noticed that I have committed a handful of R packages recently. As per the current pkgsrc state-of-the-art, these all went into math. It has been pointed out that perhaps some of these should have gone into other categories. I know there is policy against moving package categories, but perhaps since none of these are yet in a release it would be an appropriate time to discuss.

There is, however, a larger issue: existing R packages already cover a much wider range of categories (e.g., at least biology, geography, www, etc.) than implied by them all being in math. Thus, perhaps a broader policy discussion is relevant prior to the upcoming branch. Here are the two alternative views as I see them:

- Other language-specific packages (e.g., perl and python) have packages scattered across categories so why should not R packages be organized along category lines?
+ Pros: consistency with pkgsrc organization generally
+ Cons: implementing this consistently requires moving existing R packages to new categories

- All R packages are now in math, why not keep that organization?
+ Pros: consistent with current organization of R packages, single place (e.g., math) to look for R packages
+ Cons: inconsistent with rest of pkgsrc, perhaps surprising not to find R packages in an expected category

Can we identify a consensus for how to handle the categorization of R packages that covers long-standing, newly committed and future ones?

Cheers,
Brook
Hubert Feyrer
2014-09-05 16:23:21 UTC
Permalink
Post by Brook Milligan
- Other language-specific packages (e.g., perl and python) have packages scattered across categories so why should not R packages be organized along category lines?
+ Pros: consistency with pkgsrc organization generally
+ Cons: implementing this consistently requires moving existing R packages to new categories
There's enough prior art there to put the R packages where their
functionality lies, i.e. don't keep them in 'math'.


- Hubert
Brook Milligan
2014-09-05 16:47:12 UTC
Permalink
Post by Brook Milligan
- Other language-specific packages (e.g., perl and python) have packages scattered across categories so why should not R packages be organized along category lines?
+ Pros: consistency with pkgsrc organization generally
+ Cons: implementing this consistently requires moving existing R packages to new categories
There's enough prior art there to put the R packages where their functionality lies, i.e. don't keep them in 'math'.
Just to be clear: by this you mean that (i) existing R packages should be left as is, (ii) my newly committed ones should be moved, and (iii) future packages should be in appropriate categories? Or are you thinking that the entire set of R packages should be reorganized?

Cheers,
Brook

Loading...