ractors

[grosser]

• in_ractors kinda works
• need to fix map/each etc
• need to add tests (exclude from ruby <3)

see gotchas from https://github.com/grosser/parallel/issues/298

[enumag]

Hello, is there some documentation for the ractors feature? In readme you have just this:

results = Parallel.map(['a','b','c'], in_ractors: 3, ractor: [SomeClass, :expensive_calculation])

Which is a bit confusing:

• What is that ractor: [SomeClass, :expensive_calculation]? Is expensive_calculation a method name on SomeClass or something else?
• Is there a way to propagate global variables into the ractor? Or is it done automatically?
• If not does the ractor at least get the ARGV array of the original process?
• Can I instead of in_ractors: 3 say "use as many ractors as possible for the current hardware" (such as the total amount of CPU cores)?

[grosser]

• yes
• ractors cannot use global variables
• ractors cannot use ARGV
• no, either make a PR or use in_ractors: Etc.nprocessors / Parallel.physical_processor_count

[enumag]

Is there a way to pass some additional parameters to the ractor? When I use the Parallel.map call the SomeClass, :expensive_calculation method will receive one of the items from the array as parameter. But of course I need to pass some other values (same for all ractors). How can I do that?

[grosser]

actual = [1,2,3]
input = actual.map { |x| [x, ARGV, $whatever] }
Parallel.map input, ...

if it still complains then wrap the object in Ractor.make_shareable

[enumag]

Not very nice API in my opinion but should do the job. Thanks!

[grosser]

yeah, best I could do, complain to the Ractor implementors :D