Generators and deliberate materialization

The idea

A generator describes values to compute as iteration proceeds. It can feed sum, any, or another consumer without first allocating a complete array of results. A comprehension with square brackets deliberately materializes an array instead. Both forms can express the same transformation, but their memory use and the timing of computation differ, especially when producing each item has a side effect.

Iterator composition also lets a finite consumer work with a potentially unbounded source. Bound that source with Iterators.take or an explicit stopping rule before a consumer that needs every item. Laziness does not imply caching or purity: a generator may calculate again when traversed again, and changing captured data can change later results. Make repeated traversal and ownership assumptions explicit.

g = (n * n for n in 1:4 if iseven(n))
println(sum(g))
first_three = Iterators.take(Iterators.countfrom(1), 3)
println(join(collect(first_three), ","))

20
1,2,3

The filtered generator contributes 4 and 16. take bounds the increasing source to three values before collection.