ROBUSTNESS: Use defaults for unset ff* options

[HenrikBengtsson]

Issue

If some functions, own or third-party, resets all R options to what they were prior to loading ff, then some ff functions fail because querying those options result in NULL values. Imagine a function that reset all R options using on.exit() to protect against side effects from some rough function call setting, say, options(warn = 2).

Reproducible example

library(ff)

## ff options
names <- grep("^ff", names(options()), value = TRUE)
opts <- options()[names]

x <- ff(1:3)

## remove ff options
reset <- setNames(vector("list", length = length(opts)), names(opts))
options(reset)

y <- ff(1:3)
#> Error in splitted$path[nopath] <- getOption("fftempdir") : 
#>  replacement has length zero

Troubleshooting

The problem is that ff::fftempfile() assumes that R option fftempdir is always set;

ff/R/fileutil.R

Lines 181 to 187 in 46bc1c5

	fftempfile <- function(x){
	splitted <- splitPathFile(x)
	nopath <- splitted$path=="" & splitted$fsep==""
	if (any(nopath))
	splitted$path[nopath] <- getOption("fftempdir")
	tempPathFile(splitted, extension=getOption("ffextension"))
	}

Suggestion

I'd like to argue that code should not make assumptions about R options being set, even for "internal" R options. My suggestions is to add defaults to all getOption(), e.g.

fftempfile <- function(x){
  splitted <- splitPathFile(x)
  nopath <- splitted$path=="" & splitted$fsep==""
  if (any(nopath)) {
    path <- getOption("fftempdir")
    if (is.null(path)) path <- standardPathFile(file.path(tempdir(), "ff"))
    splitted$path[nopath] <- path
  }
  tempPathFile(splitted, extension=getOption("ffextension"))
}

Alternatively, give an informative error message:

path <- getOption("fftempdir", stop("Option 'fftempdir' has been removed. It is essential for the 'ff' package to function. Please restart your R session"))

Session info

> sessionInfo()
R version 4.2.0 (2022-04-22)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.4 LTS

Matrix products: default
BLAS:   /home/hb/shared/software/CBI/R-4.2.0-gcc9/lib/R/lib/libRblas.so
LAPACK: /home/hb/shared/software/CBI/R-4.2.0-gcc9/lib/R/lib/libRlapack.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] 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] ff_4.0.7  bit_4.0.4

loaded via a namespace (and not attached):
[1] compiler_4.2.0

[HenrikBengtsson]

A friendly nudge about this one. I don't think resetting R options should break a package this much. I think it wouldn't be too hard to recover from it, if it happens.