small change to attempt to stop tempfile() problems in larger documents.

[abelew]

Greetings,
When knitting larger documents with many images I sometimes get tempfile errors saying it ran out of files. When I looked more closely, it seemed to me that tempfile() was just not trying very hard; so I hacked a quick md5 tempfile generator which in theory takes all the same arguments as base::tempfile(). I was thinking to make it more robust and use some actual pseudorandomly generated material, but this seems to have worked fine for all of my largest and most troublesome documents.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[yihui]

I wish to understand the problem before fixing it. I don't see the limitation mentioned on the help page ?tempfile. I tested a million temp files and it seems to work fine for me:

for (i in 1:1e6) {
  f = tempfile(); file.create(f); file.remove(f)
}

Do you have more than a million plots in your document?

> xfun::session_info()
R version 4.3.1 (2023-06-16)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Ventura 13.5.2, RStudio 2023.9.1.494

Locale: en_US.UTF-8 / en_US.UTF-8 / en_US.UTF-8 / C / en_US.UTF-8 / en_US.UTF-8

Package version:
  compiler_4.3.1    graphics_4.3.1    grDevices_4.3.1   rstudioapi_0.15.0 stats_4.3.1       tools_4.3.1       utils_4.3.1      
  xfun_0.40.2 

[abelew]

I think I replied via email, but the following is an example more similar to what happens in my documents and shows that you only need ~ 30 plots (at least in my hands) to hit a duplicate tmpnam:

library(parallel)
library(doParallel)
library(iterators)
library(foreach)
iterators <- 10
returns <- c()
## I haven't used parallel in a while, I can't remember if it is like make -j and I can't remember if returns or res gets the result back...
cl <- parallel::makeCluster(iterators + 1)
registered <- doParallel::registerDoParallel(cl)
res <- foreach(c = 1:(iterators * 10)) %dopar% {
  filename <- tempfile()
  returns <- c(returns, filename)
}
stopped <- parallel::stopCluster(cl)
table(duplicated(unlist(res)))

[yihui]

Sorry but I don't quite understand the code since I'm not familiar with these packages. In particular, I don't understand what returns <- c(returns, filename) is trying to do. sapply(res, duplicated) is all FALSE, and I don't know what duplicated(unlist(res)) implies.

Anyway, it will be clearer if you can show a minimal reproducible example that actually uses knitr.