Rakudo any/all/none junctions now collapse

Originally published on 13 November 2008 by Patrick Michaud.

For those who are wondering what happened to my journal posts over the past couple of months – I’ve been focusing on some underlying Rakudo/Parrot design issues and supporting others more than writing code or prose. Also, I think I needed a bit of a break.

I’ve decided that it’ll be easier (and thus more likely) for me to increase visibility on Rakudo progress if I blog about new features and decisions as they happen rather than try to batch them up for a post every few days. I do still plan to come up with “weekly summaries” as well, but I figure the immediate posts will be more satisfying for me and more helpful for others.

So, here’s the first such post – earlier this week I refactored Rakudo’s Junction implementation to be cleaner and more correct, and bacek++ provided a patch to collapse duplicate values of one, any, and none Junctions. We had to wait for confirmation from rakuanguage for this, which we just got today. So, we now have:

> say any(1,2,2,1,3,2).raku;
any(1, 2, 3);
> say (any(1,2,3) % 2).raku;
any(1, 0);

Although any/all/none junctions collapse, one junctions do not:

> say (one(1,2,2,1,3,2) % 2).raku
one(1, 0, 0, 1, 1, 0)

This is because duplicate values change the boolean meaning of a one junction, so we can’t simply collapse them. As Larry says, “… one junctions produce bags rather than sets.”

As a result we’ve been able to unfudge more spectest for rakudo – we’re currently (r32625) passing 4598 spectests.