RogerBW's Blog

I've found myself thinking lately about loops.

Specifically, how do you terminate a loop? I see two basic sorts of thing here:

• There is a reasonably knowable number of iterations when you start. For example, "given this list of items, do the following once for each item". The for loop can count as a special case of this; BASIC's FOR I = 1 TO 10 can be expressed as foreach my $i (1 .. 10) in Perl.

• You don't know how many iterations you have: you want to loop until a condition changes. This takes various forms: in BBC BASIC it was REPEAT...UNTIL, and in Perl it's usually while (condition) { block }; languages may allow a test at the beginning or the end of the loop, depending on whether you want it to be executed at least once or not.

But while fiddling with PostScript, which doesn't have a while operator, I realised that start and end tests are just special cases: the test can be anywhere, taking the form "if (terminating condition), leave the loop and go on to the next thing". And indeed you can have multiple tests at different places.

You can extend this thinking to more sophisticated languages (Perl has last to terminate a loop, C-like languages have break). (Generally there's also some way to say "abandon this iteration of the loop and carry on with the next one".) My Rust these days often has an infinite loop { } with explicit termination conditions where I want them.

Except that functional languages don't want you to do that. In Scala you can by design not break out of a loop. So I at least find myself constructing scaffolding around it: have an exit-flag as part of the while test, and if I set it mid-way through loop processing I must explicitly skip over the rest of the loop body to go straight to the test point. I'm sure that people who understand functional programming can explain why it's important to have to do this.