For a small number of keys, the nested loop isn't so bad. I've been burned enough times by changing requirements that if I have two algorithms of similar complexity to code and cost to run, I'll almost always favor the one that scales better, even if the ability to scale isn't strictly necessary.

The matching regex should probably include a quotemeta as well as the start and stop anchors: /^\Q$k\E$/. So key names like day* are safe. This SHOULD cover all the possible mixups. I think it does. I'm pretty sure...

All the above struggling with subtle bugs demonstrates something important. If you are testing for string equivalence, use the string equivalence operator, and not a regex.

So, if I decided to implement the code with a call to any I'd use:

next if any { $_ eq $k } @do_not_want;
[download]