For your scenarios, as written, I agree with Benoit's #note-2 suggestions. ;) I also agree that core/stdlib #then
should only ever yield a single value to its block. However, it's worth noting that multi-parameter blocks will automatically deconstruct a single array arg. E.g:
irb(main):001:0> %i[hello world this_is_the_third].then {|x, y, z| puts "first: #{x}"; puts "second: #{y}"; puts "third: #{z}"; [x.to_s, y.to_s.upcase, z.to_s[-5..]] }
first: hello
second: world
third: this_is_the_third
=> ["hello", "WORLD", "third"]
irb(main):023:0> def foo(x) = puts "foo(%p)" % x
=> :foo
irb(main):024:0> def bar(y) = puts "bar(%p)" % y
=> :bar
irb(main):025:0> def baz(x) = puts "baz(%p)" % x
=> :baz
irb(main):026:0> def qux(x, y, z) = puts("qux(%p, %p, %p)" % [x, y, z]).then { :done_qux }
=> :qux
irb(main):027:0> honyarara, fugafugafuga, hogehogehoge = :honyarara, :fugafugafuga, :hogehogehoge
=> [:honyarara, :fugafugafuga, :hogehogehoge]
irb(main):028:1* [honyarara, fugafugafuga, hogehogehoge].then {|x,y,z| # blocks automatically destructure array args
irb(main):029:1* foo(x); bar(y); baz(x); qux(x, y, z)
irb(main):030:0> }
foo(:honyarara)
bar(:fugafugafuga)
baz(:honyarara)
qux(:honyarara, :fugafugafuga, :hogehogehoge)
=> :done_qux
irb(main):031:0> p = proc {|x,y,z| foo(x); bar(y); baz(x); qux(x, y, z) } # procs handle args like blocks
=> #<Proc:0x00007f418f800ed0 (irb):31>
irb(main):032:0> [honyarara, fugafugafuga, hogehogehoge].then(&p)
foo(:honyarara)
bar(:fugafugafuga)
baz(:honyarara)
qux(:honyarara, :fugafugafuga, :hogehogehoge)
=> :done_qux
irb(main):033:0> lunary = -> a { x,y,z = a; foo(x); bar(y); baz(x); qux(x, y, z) } # lambdas handle args like methods
=> #<Proc:0x00007f41900ff130 (irb):33 (lambda)>
irb(main):034:0> [honyarara, fugafugafuga, hogehogehoge].then(&lunary)
foo(:honyarara)
bar(:fugafugafuga)
baz(:honyarara)
qux(:honyarara, :fugafugafuga, :hogehogehoge)
=> :done_qux
irb(main):035:0> l3args = -> x, y, z { foo(x); bar(y); baz(x); qux(x, y, z) }
=> #<Proc:0x00007fb5a9025c40 (irb):35 (lambda)>
irb(main):036:0> [honyarara, fugafugafuga, hogehogehoge].then{l3args.(*_1)}
foo(:honyarara)
bar(:fugafugafuga)
baz(:honyarara)
qux(:honyarara, :fugafugafuga, :hogehogehoge)
=> :done_qux
irb(main):037:0> lwrapped = -> ((x, y, z)) { foo(x); bar(y); baz(x); qux(x, y, z) }
=> #<Proc:0x00007fb5a88402c8 (irb):37 (lambda)>
irb(main):038:0> [honyarara, fugafugafuga, hogehogehoge].then(&lwrapped)
foo(:honyarara)
bar(:fugafugafuga)
baz(:honyarara)
qux(:honyarara, :fugafugafuga, :hogehogehoge)
=> :done_qux
I most commonly do this when I'm iteratively constructing a one-shot data-munging query in irb/pry. Pipeline segments can output arrays which can be deconstructed by the next segment into multiple args. IMO, it's a great technique for the REPL, but positional parameters can get unwieldy fast. Except for when it's very simple and self-documenting, I prefer to refactor to something that's easier to read before committing or merging.