Feature Request: Support disk directive with local executor

[schorlton-bugseq]

New feature

Thanks for all of your work on nextflow! As best I can tell from the docs, the disk directive is not implemented with the local executor. Is this accurate? I very naively thought local would have been the first executor for it to be implemented only to realize later that it wasn't having any impact on my workflow (as far as I can tell). This directive would likely be very useful to the community for processing big data.

Without a full appreciation for how the directive is implemented on other platforms, I would think we could call something similar to https://unix.stackexchange.com/a/6014 in .command.run and error if there is not sufficient disk space? Thanks as always for your consideration!

[bentsherman]

I think we could use df to get the available disk space at the start of a run. The local executor already does a similar thing with memory. But local doesn't actually limit the memory usage of each task, it just keeps track of how much memory it has allocated against the total and blocks pending tasks accordingly. This seems like a reasonable heuristic to use for disk space

[bentsherman]

Even better, File::getUsableSpace() can give us a disk space estimate in a Java-native way. Some basic tests suggest it is very close to df.