parallel processing about missionary HOT 3 CLOSED

I see zero downside to solution 3 so far, it seems to be the way. Moreover, assuming a mechanism to limit concurrency, ?> and ?= actually unify because they're special cases for bounds 1 and ##Inf respectively.

In fact it's unclear to me if there's any valid use case for unbounded concurrent forking. If the input size has a known upper bound then you can specify it, otherwise you don't want to pull values indefinitely so you have to choose a bound anyway.

from missionary.

So here's a plan :

• ?> now accepts an optional argument before the flow, the concurrency bound. Must be a positive integer, the default value is 1, so the previous behavior is preserved.
• ?= desugars to ?> with an infinite bound. It is marked as deprecated and will eventually be removed, because it's a dangerous pattern that can blow up memory if not carefully used.
• amb= desugars to ?> with a bound equal to the form count.
• amb> is renamed to amb (both flavors of amb rely on ?> so the angle bracket has no meaning here).

The bad news is, we have a naming problem again. ReactiveX tricked me into thinking there was 3 forking operators, and I thought it was a good idea to steal their naming scheme - cf #44 . It turns out there's only 2, one giving priority to the parent and the other giving priority to the children.

from missionary.

Released in b.27

from missionary.