BiocVersion recorded in lockfile but not used in this project

[lucygarner]

Hi,

I am getting the message that BiocVersion "recorded in the lockfile, but do not appear to be used in this project".
"Use renv::snapshot() if you'd like to remove these packages from the lockfile."

However, when I run renv::snapshot(), it says "The lockfile is already up to date". Is this something specific about the BiocVersion package?

Best wishes,
Lucy

sessionInfo()

R version 4.3.0 (2023-04-21)
Platform: x86_64-apple-darwin20 (64-bit)
Running under: macOS Ventura 13.3.1

Matrix products: default
BLAS:   /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-x86_64/Resources/lib/libRlapack.dylib;  LAPACK version 3.11.0

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

time zone: Europe/London
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices datasets  utils     methods   base     

loaded via a namespace (and not attached):
[1] compiler_4.3.0      BiocManager_1.30.20 tools_4.3.0         rstudioapi_0.14    
[5] yaml_2.3.7          renv_0.17.3   

[hadley]

Can you provide at least a rough reprex of your current package state?

[kevinushey]

I think we need to pick up BiocVersion as an "implicit" dependency of projects using Bioconductor, because no packages will explicitly depend on it but users can install a specific version of the BiocVersion package to implicitly depend on a specific Bioconductor release. At least, that's my understanding based on https://bioconductor.org/packages/release/bioc/html/BiocVersion.html.

[kevinushey]

BiocManager is a similar type of dependency here.

[lucygarner]

Can you provide at least a rough reprex of your current package state?

What would be the most useful thing to have? A copy of the renv.lock file?

renv.lock (contents of the renv.lock file as a txt file)

I am trying to find a simple way to recreate the issue, but haven't yet managed. I am getting the same message for another one of my projects though. It only seems to happen since I've upgraded to R 4.3.0 and therefore also upgraded Bioconductor versions (3.17).

[lucygarner]

I have managed to recreate the issue with the following code:

renv::init()

renv::install("BiocManager")
BiocManager::install("scuttle")

renv::snapshot()

BiocManager::install("clusterProfiler")

renv::snapshot()

The issue only appears after I then restart my R session. I have a separate packages.R file which loads the scuttle and clusterProfiler packages.

[lucygarner]

I think we need to pick up BiocVersion as an "implicit" dependency of projects using Bioconductor, because no packages will explicitly depend on it but users can install a specific version of the BiocVersion package to implicitly depend on a specific Bioconductor release. At least, that's my understanding based on https://bioconductor.org/packages/release/bioc/html/BiocVersion.html.

Just to say that BiocVersion was installed as a dependency of another package, rather than me explicitly installing it.

[kevinushey]

For reference:

> tools::package_dependencies("BiocVersion", reverse = TRUE)
$BiocVersion
[1] "AnnotationHub" "pkgndep"      

But, at least from the lockfile, it doesn't look like your project depends on AnnotationHub?

[hadley]

I can't reproduce the problem with this reprex script:

renv::init()

renv::install("bioc::scuttle")
writeLines("library(scuttle)", "dep-1.R")
renv::snapshot()

renv::install("bioc::clusterProfiler")
writeLines("library(clusterProfiler)", "dep-2.R")
renv::snapshot()

renv:::renv_restart_request(".")
renv::snapshot()

I switch to using renv::install() since there are SO many packages to install and the cache makes a big difference.

[hadley]

Oh because your reprex doesn't include status() which is the actual source of the problem.

[hadley]

Ok, here's a reprex that shows the problem — it appears to be caused by taking and then removing a BioC dependency:

renv::init()

renv::install("bioc::BiocGenerics", prompt = FALSE)
writeLines("library(BiocGenerics)", "deps.R")
renv::snapshot(prompt = FALSE)

unlink("deps.R")
renv::snapshot(prompt = FALSE)
renv::status()

The following package(s) are installed, but not recorded in the lockfile:

- BiocManager [1.30.20]

Use `renv::snapshot()` to add these packages to the lockfile.

It's not exactly the same package, but the problem seems like it's in the same ballpark.

[hadley]

And that arises because of this special case:

renv/R/status.R

Lines 186 to 189 in e1953d0

	sources <- extract_chr(library, "Source")
	if ("Bioconductor" %in% sources)
	packages <- unique(c(packages, "BiocManager"))

It feels like that logic should involve packages (what's used), not just library (what's installed).

And possibly this too:

renv/R/snapshot.R

Lines 925 to 931 in e1953d0

	sources <- extract_chr(kept, "Source")
	if ("Bioconductor" %in% sources) {
	packages <- c("BiocManager", "BiocInstaller", "BiocVersion")
	for (package in packages)
	kept[[package]] <- records[[package]]
	}

[lucygarner]

Oh because your reprex doesn't include status() which is the actual source of the problem.

Interesting. I didn't have to run status() for the message to show up.