The performance win would come from ignoring directories. My tree has a website subdir with > 33K files that I do not need to search.
Hmm, then maybe glob is indeed a better choice, especially if you don’t need the Info struct on most or all of the results?
Ah, ok, I didn’t realize that glob reads the filesystem, I assumed it was some kind of pattern thingie. I’ll try to find some time to fiddle with it this week, thanks.,