Lazy enumeration and bounded demand

The idea

Enum operations consume their input immediately and return results, while many Stream operations describe a lazy traversal. The lazy pipeline runs when a consumer enumerates it. This lets a program process large inputs incrementally or take a finite prefix from an unbounded source. Laziness does not mean the results are cached for future traversals.

Place limits before eager materialization. A pipeline that converts an infinite stream into a list before taking ten values never reaches the limiting step. Repeated enumeration can repeat callbacks and side effects, so keep transformations pure when the stream is intended to be reusable. Resource-backed streams also need a cleanup policy for normal completion and early termination.

result = Stream.iterate(1, &(&1 + 1))
  |> Stream.filter(&(rem(&1, 2) == 0))
  |> Stream.map(&(&1 * 3))
  |> Enum.take(3)
IO.inspect(result)

[6, 12, 18]

The consumer demands only three transformed even values, keeping the otherwise infinite producer bounded.