Public API

myflatmap(f, x) = Iterators.flatten(map(f, x))
iteratorresult = @monadic map myflatmap begin
  x = 1:3
  y = [1, 6]
  @pure x + y
end
collect(iteratorresult)  # [2, 7, 3, 8, 4, 9]

The @monadic macro allows a syntax where containers and other contexts are treated rather as values, hiding the respective well-defined side-effects. Each line without @pure is regarded as a container, each line with @pure is treated as normal code which should be inlined.

For the example above you see that the side-effect semantics of iterables are the same as for nested for loops. With the crucial distinction, that the @monadic syntax has a return value.

@monadic can take a third argument wrapper in order to first apply a custom function before executing the @monadic code.

mywrapper(n::Int) = 1:n
mywrapper(any) = any
myflatmap(f, x) = Iterators.flatten(map(f, x))
iteratorresult = @monadic map myflatmap mywrapper begin
  x = 3
  y = [1, 6]
  @pure x + y
end
collect(iteratorresult)  # [2, 7, 3, 8, 4, 9]

Mark code to contain non-monadic code.

This can be thought of as generalized version of pure typeclass.