I use the checkpoint package for reproducible data analysis. Some of the computations take a long time to compute, so I want to run those in parallel. When run in parallel however the checkpoint is not set on the workers, so I get an error message "there is no package called xy" (because it is not installed in my default library directory).
How can I make sure, that each worker uses the package versions in the checkpoint folder? I tried to set .libPaths in the foreach code but this does not seem to work. I would also prefer to set the checkpoint/libPaths once globally and not in every foreach call.
Another option could be to change the .Rprofile file, but I do not want to do this.
checkpoint::checkpoint("2018-06-01")
library(foreach)
library(doFuture)
library(future)
doFuture::registerDoFuture()
future::plan("multisession")
l <- .libPaths()
# Code to run in parallel does not make much sense of course but I wanted to keep it simple.
res <- foreach::foreach(
x = unique(iris$Species),
lib.path = l
) %dopar% {
.libPaths(lib.path)
stringr::str_c(x, "_")
}
Error in { : task 2 failed - "there is no package called 'stringr'"
Author of the future package here.
UPDATE 2022-05-25: As of future 1.20.0 (2021-11-03), multisession parallel workers automatically inherit the R library path (=.libPaths()) from the main R session. Because of this, the below workaround is no longer needed. However, it may still be needed for other future backends.
Passing the library path of the master R process as a global variable libs and set it for each worker using .libPaths(libs) should be enough;
## Use CRAN checkpoint from 2018-07-24 to get future (>= 1.9.0) [1],
## otherwise the below stdout won't be relayed back to the master
## R process, but settings .libPaths() does also work in older
## versions of the future package.
## [1] https://cran.microsoft.com/snapshot/2018-07-24/web/packages/future
checkpoint::checkpoint("2018-07-24")
stopifnot(packageVersion("future") >= "1.9.0")
libs <- .libPaths()
print(libs)
### [1] "/home/hb/.checkpoint/2018-07-24/lib/x86_64-pc-linux-gnu/3.5.1"
### [2] "/home/hb/.checkpoint/R-3.5.1"
### [3] "/usr/lib/R/library"
library(foreach)
doFuture::registerDoFuture()
future::plan("multisession")
res <- foreach::foreach(x = unique(iris$Species)) %dopar% {
## Use the same library paths as the master R session
.libPaths(libs)
cat(sprintf("Library paths used by worker (PID %d):\n", Sys.getpid()))
cat(sprintf(" - %s\n", sQuote(.libPaths())))
stringr::str_c(x, "_")
}
### - ‘/home/hb/.checkpoint/2018-07-24/lib/x86_64-pc-linux-gnu/3.5.1’
### - ‘/home/hb/.checkpoint/R-3.5.1’
### - ‘/usr/lib/R/library’
### Library paths used by worker (PID 9394):
### - ‘/home/hb/.checkpoint/2018-07-24/lib/x86_64-pc-linux-gnu/3.5.1’
### - ‘/home/hb/.checkpoint/R-3.5.1’
### - ‘/usr/lib/R/library’
### Library paths used by worker (PID 9412):
### - ‘/home/hb/.checkpoint/2018-07-24/lib/x86_64-pc-linux-gnu/3.5.1’
### - ‘/home/hb/.checkpoint/R-3.5.1’
### - ‘/usr/lib/R/library’
str(res)
### List of 3
### $ : chr "setosa_"
### $ : chr "versicolor_"
### $ : chr "virginica_"
FYI, it is on future's roadmap to make it easier to pass down the library path(s) to workers.
My details:
> sessionInfo()
R version 3.5.1 (2018-07-02)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 18.04.1 LTS
Matrix products: default
BLAS: /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.7.1
LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.7.1
locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8 LC_MONETARY=en_US.UTF-8
[6] LC_MESSAGES=en_US.UTF-8 LC_PAPER=en_US.UTF-8 LC_NAME=C LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] foreach_1.4.4
loaded via a namespace (and not attached):
[1] drat_0.1.4 compiler_3.5.1 BiocManager_1.30.2 parallel_3.5.1 tools_3.5.1 listenv_0.7.0 doFuture_0.6.0
[8] codetools_0.2-15 iterators_1.0.10 digest_0.6.15 globals_0.12.1 checkpoint_0.4.5 future_1.9.0