Bioconductor packages need to be re-detected for installation

[zkamvar]

In the transformation of https://github.com/carpentries-incubator/bioc-intro in https://github.com/fishtree-attempt/bioc-intro:

> manage_deps()                                                                                                                                                                                 
ℹ Consent to use package cache provided
→ Searching for and installing available dependencies
Done!ng R package dependencies ... 
Done!covering package dependencies ... 
* Copying packages into the cache ... Done!
* Resolving missing dependencies  ... 
The following package(s) were not installed successfully:

	[SummarizedExperiment]: package 'SummarizedExperiment' is not available
	[tidySummarizedExperiment]: package 'tidySummarizedExperiment' is not available

You may need to manually download and install these packages.

→ Restoring any dependency versions
* The library is already synchronized with the lockfile.
→ Recording changes in lockfile
* Lockfile written to '~/Documents/Carpentries/Git/carpentries/lesson-transition/sandpaper/carpentries-incubator/bioc-intro/renv/profiles/lesson-requirements/renv.lock'.

Solution

I believe the solution is to include bioconductor = TRUE in

sandpaper/R/utils-renv.R

Line 132 in d94acec

renv::init(project = path, bare = TRUE, restart = FALSE, profile = profile)

That being said, I will have to see if that's going to affect non-bioconductor lessons.

[zkamvar]

Honestly, I am so dang embarrassed about this

[zkamvar]

I belive the problem is coming from the fact that renv_available_packages() does not include BioConductor repositories if it's not part of the "repos" option.

[zkamvar]

Note, that the above function is called in the cascade from renv_retrieve_missing_record() starting with renv::hydrate()

[zkamvar]

N.B. the error is coming from sandpaper:::callr_manage_deps() calling renv::hydrate() in our codebase:

sandpaper/R/utils-renv.R

Lines 256 to 259 in d94acec

	if (needs_hydration) {
	hydra <- renv::hydrate(packages = pkgs, library = renv_lib, update = FALSE,
	project = path)
	}

[zkamvar]

N.B. I found that this is a bootstrapping problem. If we use a pattern like:

library(sandpaper)
manage_deps()
done <- work_with_cache()
renv::install("bioc::SummarizedExperiment")
renv::install("bioc::tidySummarizedExperiment")
done()
manage_deps()

Then things install correctly, so I might be able to catch those and try them with bioc, but there's no guarantee

[zkamvar]

Note that renv::install() may not be the right solution. I get the feeling that renv::install() is replacing CRAN with RSPM, but in this case it is not the right solution because of the lag (see carpentries/actions#69)