I don't see anything wrong with the "copy to a temporary directory" approach if your function actually does operate on files within a directory. In that scenari

I don't see anything wrong with the "copy to a temporary directory" approach if your function actually does operate on files within a directory. In that scenario, copying to a different directory is actually probably exactly what you want for a dry-run, since then the code that you're executing is the same exact code that would run were you to execute in non-dry-run mode (as opposed to virtual operations which are prone to bugs if that virtualization layer and your real operations ever fall out of sync).