search

tidyomics / plyranges

A grammar of genomic data transformation
https://tidyomics.github.io/plyranges/
140 stars 18 forks source link

R>3.5.0 requirements #25

Closed mschubert closed 6 years ago

mschubert commented 6 years ago

Hi,

Your DESCRIPTION file specifies that R>3.5.0 is required which has not been released yet. This makes your Github HEAD not installable with the current R release (which is, I guess, roughly 100% of your users).

I found a warning on the Bioconductor build process, which I think is the reason why you changed it, or is there any other reason as well?

If not, you might want to have a separate branch for the Bioconductor submission for now so that people can actually use your package with the current stable R.

sa-lee commented 6 years ago

Hi Michael,

Yes I'm trying to get this package on the next release of Bioconductor which is why I modified the description file on master but you're right I should have a way of letting people use it on the current R release (unfortunately the Bioconductor submission process needs to happen on master). I've created a new branch https://github.com/sa-lee/plyranges/tree/current-r-release until then.

Thanks for the suggestion!

lawremi commented 6 years ago

I've never really understood why they require a dependency on the current version of R. We should trust package authors to get that right.

It's enouraging that people are already interested in the package!

mschubert commented 6 years ago

It's not even the current version, but rather the version in active development.

There's some discussion on the Bioc-devel mailing list and we just had a long thread on Twitter.

Bottom line: Bioc-devel thinks this will make pkg developers download R-devel, Bioc-devel, and check their package against these.

Of course, this will break all packages currently being submitted to Bioconductor for use with the current stable version of R, leaving individual package developers to find a workaround for this mess (until the next Bioc release, where it will be available there).

Quite frankly, I think it would be better to rely on automated tests and not abuse a package field in a way that was not intended (the R dependency should indicate minimum version for the package to function, not a convenience to nudge users to Bioc-devel). The Bioc releases themselves could impose whatever requirement on R as they like of course, but that could also be enforced using e.g. an R dependency of Biobase.

The package looks really useful, I just included it in one of my analyses! :+1:

lawremi commented 6 years ago

Agreed. Actually, BiocInstaller already enforces the correct R version.