Bump to R 4.5 & Bioconductor 3.22

[sjspielman]

Closes #869

This PR updates the renv environment to 4.5 with associated bioc 3.22. For this, I removed the last renv environment and snapshotted in a new one for the versions of interest. Matching our current renv.lock file, I ensured that we're pointing to the ppm link for the CRAN repo.

After snapshotting, I compared with setdiff packages that were in 4.4 vs now in 4.5 lock files. I made a little spreadsheet here to confirm that it all makes sense, and it does! https://docs.google.com/spreadsheets/d/1JAKuYGS6_mdzga06urdE9RsrOSsjO-2zMxdoAWHQtOk/edit?usp=sharing

The first sheet contains packages in 4.4 but not the 4.5 lockfile. It might be that we want some of these in the dependencies.R file, but we can't say for sure at this time. We'll probably want to revisit when we journey into how the version bump affected module notebooks. Note that otherwise, dependencies.R looks fine to stay as is for now.

That said, note this new aspect of renv's settings: https://rstudio.github.io/renv/reference/settings.html?q=snapshot.dev#snapshot-dev.

Finally, this is kind of annoying. Following the code in renv/activate.R that is throwing this, I don't immediately see why this would happen because versions, indeed, match. Do you see the "culprit"?

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

renv 1.1.6 was loaded from project library, but this project is configured to use renv 1.1.6.
- Use `renv::record("renv@1.1.6")` to record renv 1.1.6 in the lockfile.
- Use `renv::restore(packages = "renv")` to install renv 1.1.6 into the project library.
- Project '~/ALSF/training/training-modules' loaded. [renv 1.1.6]
> renv::status()
No issues found -- the project is in a consistent state.

[sjspielman]

Ah, well this Docker build fail checks out. Going to be hard to build for 4.5/3.22 when the Dockerfile isn't getting that version to begin with!

[jashapiro]

Finally, this is kind of annoying. Following the code in renv/activate.R that is throwing this, I don't immediately see why this would happen because versions, indeed, match. Do you see the "culprit"?

known bug: https://github.com/rstudio/renv/blob/main/NEWS.md

[sjspielman]

known bug: https://github.com/rstudio/renv/blob/main/NEWS.md

thanks! i had only looked at their issues, not their release notes. so that's good.

Noting also I tried to build locally before pushing the last 2 commits, but kept hitting a python error. I sadly see that the GHA hits the same error: https://github.com/AlexsLemonade/training-modules/actions/runs/21218044606/job/61044746608?pr=884#step:8:11468

edit - seems to be a bioc 3.22 base image situation, e.g. found this https://pythonspeed.com/articles/externally-managed-environment-pep-668/

[jashapiro]

The python error has an easy and a hard solution. For now I think we can try the easy but dangerous one and then we can consider the best way to implement the hard but safe one.

Basically, we can force install the Python packages at the system level, which is what we are doing, but has always been a not-great idea.

A longer term solution will be to install the packages in a separate environment and then update the skel files to direct all users to that environment. Let's file a separate issue to come back to that.

I also suggested here updating the base image to match the bioconductor version

[sjspielman]

The python error has an easy and a hard solution. For now I think we can try the easy but dangerous one and then we can consider the best way to implement the hard but safe one.

Basically, we can force install the Python packages at the system level, which is what we are doing, but has always been a not-great idea.

A longer term solution will be to install the packages in a separate environment and then update the skel files to direct all users to that environment. Let's file a separate issue to come back to that.

Thanks for weighing in! I uncovered this right before hopping over to a meeting, so just now getting back to it. Your comment is happily along the lines of what I was thinking - we want a more robust solution here, but that's a whole thing™️ that seems beyond the scope of bumping R versions. We'll live on the edge for now, and come back from the edge later. (is this a pun re-edge versioning? possibly)

[sjspielman]

no replies to this, but possibly relevant COMBINE-lab/salmon#984

[jashapiro]

no replies to this, but possibly relevant COMBINE-lab/salmon#984

Good find. It actually looks like this was updated in later point releases (that are tagged but not "releases"), so I suggested the version bump here. It also seems to require libzstd, so I added that as well.

[sjspielman]

Good find. It actually looks like this was updated in later point releases (that are tagged but not "releases"), so I suggested the version bump here.

Good find to everyone.

[jashapiro]

LGTM. I admit I did not go through all of the renv changes, but I am trusting that there is nothing too crazy in there.

[jashapiro]

Nice that this works again