I am working with fairly large S4 objects in R (~10GB). Often, I want to load them, do something to them, then save the results. Ideally, I would just overwrite the object using SaveRDS. For example:
a = readRDS(object_file)
a = do_something()
saveRDS(a, file=object_file)
However, this seems unsafe because if the file gets corrupted then I lose all of my work. For example, I work remotely and if my internet disconnects, I'm worried that it would only write part of the file. At the same time, I don't want to save multiple copies of the object, because then I need to manually organize them and delete old versions.
The approach I am considering is like this:
a = readRDS(object_file)
a = do_something()
saveRDS(a, file=temporary_file)
b = readRDS(temporary_file)
system(paste('mv', temporary_file, object_file))
Are there any problems with this? Is there a better way of validating a file and writing to it? I figure that the "mv" command is fast, so interrupting it is less of a worry. I've tried looking around, but couldn't find anything.
