GENERAL R QUESTION: Accessing package versions of an old R object/project?

Question

[Natalia]

How do I access the packageVersions() (or similar command, like sessionInfo()) of an rds object or .Rdata object from MONTHS ago? I simply would like to know: what package versions was I using at THAT time?

For context: I am running single cell analyses with Seurat. I try restarting the session and simply loading the object to my environment-- but when I run any of the above commands, only the ambient/'current' package versions pop up (I presume they are automatically loaded when the object is), and I know those are certainly not the full extent of the packages used and I can tell they are newer versions. Therefore, I load the rest of the packages, but as expected, they have all updated to newer versions (I had highly likely done so in other single cell experiment projects I have been running at the same time). I am having dependency issues in not being able to run a function that previously worked months ago, so I want to know what package version of that function I used to use which worked. Makes sense?

If it cannot be retrieved, what is your preferred/favorite way to document what package versions you are using in your projects?

Thanks in advance!!

Answer 1

[jared.andrews07]

This is the use case for renv. There is no way to determine what version of a package a given object was generated with unless that info is stored in the object itself somehow.

If I remember correctly, Seurat does indeed store this info to some degree, which can be accessed with Version(seurat_object). I believe that will tell you the version of the data structure, which should at least correspond to the major Seurat version (e.g. 3, 4, 5).

Answer 2

[ATpoint]

If it cannot be retrieved, what is your preferred/favorite way to document what package versions you are using in your projects?

I use Docker for all of my projects. Usually I build with the Bioconductor Docker image as a basis. DockerHub allows to store the images and GitHub takes care of version-controlling code and any notes/READMEs of the analysis.

Honestly, just re-do your analysis with freshly installed packages and from now on track things properly.