Hi!

I’d like to present streaming - a macro for creating streams much like the special form for.

streaming x <- 1..10, y <- 11..20, z <- 21..30 do
  {x, y, z}
end

creates a stream of all combinations of x, y, and z.

The main difference between streaming and for is that streaming is lazy and always returns a stream.

The reduce option is by design not supported since this cannot be done lazily. Just use for or Enum.reduce for this.

However, streaming instead provides convenient syntax for transforming a stream using the transform option.

streaming i <- 1..100, transform: StringIO.open("string") |> elem(1) do
  pid ->
    case IO.getn(pid, "", 1) do
      :eof -> {:halt, pid}
      char -> {[{i, char}], pid}
    end
after
  pid -> StringIO.close(pid)
end
|> Enum.to_list()

=> [{1, "s"}, {2, "t"}, {3, "r"}, {4, "i"}, {5, "n"}, {6, "g"}]

streaming also provides unfold, scan, resource options that works like corresponding functions in the Stream module.

Bitstring generators are also supported and can be combined with the other options.

The into option is based on Stream.into/3 and thus collects values into a collectable as a side-effect instead of returning the collectable.

Please see the GitHub repo for more examples:

https://github.com/flambard/streaming

Half-way through making this library I discovered lazy_for, but streaming is arguably more feature complete.