Course of Raku / Functional, concurrent, reactive, and web programming / Concurrent programming / Hyper and race 🆕

Parallel maps with hyper

Call .hyper on a list before a map or grep, and the work is spread across multiple worker threads — while the results still come back in the original order:

say (1..5).hyper.map(* * 2); # (2 4 6 8 10)

This looks exactly like an ordinary map, and the result is identical; the only difference is that the doublings may have been computed on different cores at the same time. Because .hyper preserves order, you can use it as a drop-in replacement for a slow map without changing anything that depends on the order of the results:

say (1..10).hyper.grep(* %% 2); # (2 4 6 8 10)

The benefit only appears when each element’s work is large enough to outweigh the cost of coordinating threads — squaring a number is far too cheap to be worth parallelising in reality. For genuinely expensive per-element work over a big list, .hyper can turn a long wait into a short one for the price of a single method call.

Order is preserved, so reductions work as expected too:

say (1..100).hyper.map(* * 2).sum; # 10100

Course navigation

← Hyper and race 🆕   |   Quiz — hyper →

💪 Or jump directly to the exercises in this section.