embed and deps when global cache not found

[r2evans]

In the docs, I see "inspects the dependencies" and think that it will identify packages that are needed for that script. However, it seems to not do much in this case.

~/tmp/quux.R

library(data.table)
library(dplyr)
library(collapse)

data.table(a = pi) |>
  mutate(b = a * 2)

renv::embed("~/tmp/quux.R")
# [1] TRUE

This results in the following being prepended to the file.

renv::use(
  renv = "renv@1.1.4"
)

Given that renv otherwise can determine deps fairly easily with

renv::dependencies("~/tmp/quux.R")
# Finding R package dependencies ... Done!
#                 Source    Package Require Version   Dev
# 1 /Users/r2/tmp/quux.R   collapse                 FALSE
# 2 /Users/r2/tmp/quux.R data.table                 FALSE
# 3 /Users/r2/tmp/quux.R      dplyr                 FALSE
# 4 /Users/r2/tmp/quux.R       renv                 FALSE

I expected something like either of the following:

renv::use(
  renv = "renv@1.1.4",
  data.table = "data.table@1.17.8",
  dplyr = "dplyr@1.1.4",
  collapse = "collapse@2.1.2"
)
### or ###
renv::use(
  renv = "renv@1.1.4",
  data.table = "data.table",
  dplyr = "dplyr",
  collapse = "collapse"
)

Granted, it's not difficult for me to manually edit that portion of the file, but it seems like the utility/purpose of embed() is not working for me here. What am I missing and/or doing wrong?

[kevinushey]

Do you have those packages installed? renv tries to look at the versions of those packages (and their dependencies) that you currently have installed, and uses those when creating the renv::use() statement. The fact that renv isn't warning you about this situation is definitely a bug, though.

In my case, I see:

renv::use(
  cli        = "cli@3.6.5",
  collapse   = "collapse@2.1.2",
  data.table = "data.table@1.17.4",
  dplyr      = "dplyr@1.1.4",
  generics   = "generics@0.1.4",
  glue       = "glue@1.8.0",
  lifecycle  = "lifecycle@1.0.4",
  magrittr   = "magrittr@2.0.3",
  pillar     = "pillar@1.10.2",
  pkgconfig  = "pkgconfig@2.0.3",
  R6         = "R6@2.6.1",
  Rcpp       = "Rcpp@1.0.14",
  renv       = "renv@1.1.5",
  rlang      = "rlang@1.1.6",
  tibble     = "tibble@3.3.0",
  tidyselect = "tidyselect@1.2.1",
  utf8       = "utf8@1.2.6",
  vctrs      = "vctrs@0.6.5",
  withr      = "withr@3.0.2"
)

(note that some packages are out-of-date as I hadn't updated packages on this machine recently)

[r2evans]

Yes, the packages are installed:

xfun::session_info(c("renv", "data.table", "dplyr", "collapse"))
# R version 4.4.3 (2025-02-28)
# Platform: aarch64-apple-darwin20
# Running under: macOS Sequoia 15.6.1
#
# Locale: en_US.UTF-8 / en_US.UTF-8 / en_US.UTF-8 / C / en_US.UTF-8 / en_US.UTF-8
#
# Package version:
#   cli_3.6.5         collapse_2.1.2    data.table_1.17.0 dplyr_1.1.4       generics_0.1.3    glue_1.8.0       
#   graphics_4.4.3    grDevices_4.4.3   lifecycle_1.0.4   magrittr_2.0.4    methods_4.4.3     pillar_1.10.2    
#   pkgconfig_2.0.3   R6_2.6.0          Rcpp_1.0.14       renv_1.1.5        rlang_1.1.6       stats_4.4.3      
#   tibble_3.3.0      tidyselect_1.2.1  utf8_1.2.6        utils_4.4.3       vctrs_0.6.5       withr_3.0.2      

.libPaths()
# [1] "/Users/r2/Library/R/4.4"                                             
# [2] "/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library"
sapply(c("renv", "data.table", "dplyr", "collapse"), \(pkg) attr(packageDescription(pkg), "file"))
#                                                  renv                                            data.table 
#       "/Users/r2/Library/R/4.4/renv/Meta/package.rds" "/Users/r2/Library/R/4.4/data.table/Meta/package.rds" 
#                                                 dplyr                                              collapse 
#      "/Users/r2/Library/R/4.4/dplyr/Meta/package.rds"   "/Users/r2/Library/R/4.4/collapse/Meta/package.rds" 

I updated to renv_1.1.5 just to be sure, same results as above.

[kevinushey]

Thanks -- I'll dig in some more this week and see if I can figure out what's going on.

[r2evans]

Found it (I think). The system I was testing this on does not have a global renv package cache (well, there's the directory, but it's empty). When I run it on a system where the cache is found and filled, it returns what you showed in your test.

Perhaps temporarily move your cache out of the way to reproduce the bug? @kevinushey

[kevinushey]

We've made some improvements in the latest release of renv -- let me know if these are sufficient. Thanks!