Building an renv environment in a Docker container for Nextflow

It seems like every time that I talk to somebody who uses our, they have an update, there are versions in a very long time, because every time that they do, a package breaks on them. This happens pretty often in my live group, for example. We have some projects that may not be able to be picked back up after they're not working properly, or after that computer dies on us. So, recently, I've been trying to put everything in a calm environment. But the issue that I've ran into, some helping a student turn a couple of our scripts into an excellent pipeline. And the problem is that those scripts, we've been running them, and they work in our studio, but then they don't work when we run them on the cluster. So we ran both of them, split up the scripts, ran it in our studio, runs just fine on our laptop, doesn't run on the cluster. I'm very excited about this, because this is always the fun problem that I love to solve of containerization of software packages. However, the problem was that she was using an old version of R because if she upgraded a package and filled in our local laptop, that should have been hit number one, that something was wrong. So then we went down the rabbit hole of trying out RIM. And RIM is a cool package that basically scrapes your entire R project, pulls in all of your libraries, independence, whether you include them or not, and then spits out the, like I just did JSON file, and as a lock file, this is configurable by the user, which I found interesting, and pulls the local versions that you have on your laptop and the R version as well. So you can get a very clean snapshot of that computer, and then just updates your gig or for you and everything, and you can commit it. Works kind of like a Python environment that just activates it via the R profile whenever you join, or change it in the directory. So that's cool.

Nice renv explaination

The real problem is nextflow's doesn't support for renv. Yeah, don't know if they ever will. Probably not needed because you know we are having a calm to support. The real issue is that now I'm trying to package up this RM into a Docker file and I was running into quite a few issues with it.

It would get about halfway through installing all of the packages and then it would hit a random error. Well, if you use the bioconductor image, maybe that's it. Then I would have thought, what if it's because our studio is uninstalled because I was using the R-A-R-R package instead of the rocker R-Studio package. (keyboard clicking)

- Installing BiocManager ...                    *** buffer overflow detected ***: terminated


A fellow traveler

No dice.

Eventually I ended up finding a GitHub repo through just searching GitHub for R&M and Docker. And then from those I found a reproducible template and in that it had a really customized Docker file that had a ton of caching built in. And I just kind of baked that into my image, added some more. The main thing as well was just through everything under the sun, edit on the install of the script for the base image. R just needs a ton of things to build. Used rocker instead as well, but the caching definitely sped up the local build times as well. Added another units and then it built finally pushed it up.

Compilation issueso

Package libtiff-4 was not found in the pkg-config search path.


Created: 2023-10-01 Sun 23:28