>In theory (extrapolating here) you could get caught in a
>infinite loop if the random assignment returned yourself every time

Well, *in theory* yes, but considering that perl can generate well over 100,000 random numbers per second on even a slow machine, the odds of an infinite loop are, well, infinitely small. The odds of it taking over 1 second are pretty slim, too. Nothing to worry about here.

>This script is interesting, I think, as it is typical of
>scripts that must work first time, and you cannot really
>test it over and over with real data as you don't want to
>bug people with junk email. I normally approach this type
>of problem with writing to files and the writing another
>Perl program to test.

Actually, I tested mine extensively - that's how I found the last person "gotcha". A simple rewrite of one line to

open (M, "|more") || die "Could not open more!\n";
[download]

takes care of that.

>Anyway, just a note to say that my solution, or code
>rather, doesn't have the problem you mention at the end,
>but it was good to point it out. I've never had the
>problem after substantial testing.

Your code *does* have that problem - if everyone else is used up, it assigns a null Kris Kringle. Try this out:

chomp,@_=split(/::/),($e{$_[0]},$l{$_[0]})=($_[1],0)while(<>);

for (keys %l) {
  $l{$_}++;
  @L=grep{($l{$_}==0)}keys%l;
  $x=int rand(@L);
  $l{$L[$x]}++;
  $l{$_}--;
  print "$e{$_} has a KK of $L[$x]\n";
}
[download]

Now run that with a data set like this:

A::aaron
B::bob
C::cindy
[download]

When aaron and bob get each other, cindy gets nothing:

%type test.txt | perl kk2.pl

aaron has a KK of B
bob has a KK of A
cindy has a KK of
[download]

The chance of it happening decreases with the total number of people, but it can still happen....