In the L2R thread, I made a comment where I contrasted a for loop
with a junction. On reflection, I think we could go further.
If we defined
my $prev = 0;
sub mangle($value) { $value + $prev; $prev=$value }
and then used it as:
my $x = any(1,2,3);
$y = mangle($x);
then the semantics of junctions are defined to fork 3 threads, to
evaluate C<mangle> for each member of the junction in an
arbitrary order. So there are now many possible values for
$y. For example: any(2,5,6) if the order of evaluation is 2,3,1.
Let us now imagine that C<for> is a junction operator, instead of a
looping statement. Now we could write:
my $x = for(1,2,3);
my $y = mangle($x);
and we can be certain that $y == (1,3,6), because a "for-junction"
will distribute that evaluations sequentially.
This ties in nicely with the pipeline syntax discussion, because
for 1,2,3 ~> print;
and
for 1,2,3 -> { print };
would naturally behave correctly. But by defining a special type of
sequentially distributed junction, we can do all sorts of clever (and
therefore nasty) things.
Dave.
--
mailto:[EMAIL PROTECTED]; http://dave.whipp.name
