r - 在 R 中运行并行计算时如何在 worker 上设置 .libPaths(检查点)

Question

我使用检查点包进行可重现的数据分析。有些计算需要很长时间才能计算，所以我想并行运行这些计算。但是当并行运行时，检查点没有设置在 worker 上，所以我收到一条错误消息 “没有名为 xy 的包”(因为它没有安装在我的默认库目录中)。

我如何确保每个工作人员都使用检查点文件夹中的包版本？我试图在 foreach 代码中设置 .libPaths 但这似乎不起作用。我还希望在全局范围内而不是在每个 foreach 调用中设置一次检查点/libPaths。

另一种选择是更改 .Rprofile 文件，但我不想这样做。

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'"

Answer 1

future 的作者在这里打包。

2022-05-25 更新:自 future 1.20.0 (2021-11-03) 起，多 session 并行工作器自动继承 R 库路径(=.libPaths()) 来自主 R session 。因此，不再需要以下解决方法。但是， future 的其他后端可能仍需要它。

---

将 master R 进程的库路径作为全局变量 libs 传递，并使用 .libPaths(libs) 为每个 worker 设置它就足够了；

## 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_"

仅供引用，它在未来的路线图上 make it easier to pass down the library path(s) to workers .

我的详细信息:

> 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