Missing Virtuousness

According to Larry, laziness is a programmers virtue. The best way to be lazy is having somebody else do it. By my request, SmokeMachine kindly did so. This is not fair. We both should have been lazy and offload the burden to the CORE-team.

Please consider the following code.

my @many-things = (1..10).List;
sub doing-one-thing-at-a-time($foo) { ... }
say doing-one-thing-at-a-time(@many-things.all);

Rakudo goes out of it’s way to create the illusion that sub doing-one-thing-at-a-time can deal with a Junction. It can’t, the dispatcher does all the work of running code in parallel. There are tricks we can play to untangle a Junction, but there is no guarantee that all values are produced. Junctions are allowed to short-circuit.

This was bouncing around in my head for quite some time, until it collided with my thoughts about Range. We may be handling HyperSeq and RaceSeq wrong.

my @many-things = (1..10).List;
sub doing-one-thing-at-a-time($foo) { ... }
say doing-one-thing-at-a-time(@many-tings.hyper(:degree<10>));

As with Junctions doing dispatch-magic to make hyper/race just work, moving the handling to the dispatcher would move the decision from the callee to the caller and, as such, from the author of a module to the user. We can do that by hand already with .hyper.grep(*.foo) or other forms of boilerplate. In Raku-land we should be able to do better and provide a generalisation of transforming calls with the help of the dispatcher.

I now know what to ask Santa for this year.