Add combinatorial iterators: product, permutations, combinations, etc.

[cjquines]

Related, but separate proposals:

• product, like xprod from Ramda, itertools.product from Python, std::ranges::cartesian_product_view from C++, itertools::Iterator::cartesian_product from Rust, requested in Lodash.
  • Would it be only for pairs, like zip, or variadic? It could be strictly typed too.
• permutations, like permutations from Deno std, itertools.permutations from Python, itertools::Iterator::permutations from Rust, similar to std::next_permutation from C++, requested in Lodash.
  • Slightly different APIs here: do we only want permutations of the whole list, or permutations of a given size?
• combinations, like itertools.combinations from Python, itertools::Iterator::combinations from Rust, requested in Lodash.

[cjquines]

i also definitely have a lot more proposals like this that i can write up if that's something people want 🤷

[eranhirsch]

These are all great and I don't mind adding them.

But be aware that their value in Python and other similar languages is that they rely heavily on the implementation using generators (yield) to make them efficient, as otherwise, they grow literally exponentially and could explode the memory in places where JS usually run which aren't optimized for memory-heavy computations like the browser and the edge.

Remeda has a notion of "iterator"-like API using our lazy evaluation mechanism, but it only works in pipes. This could trip users who use these methods naively in their dataFirst variant and we should avoid those situations:

for (const [a, b] of product(<BIG_ARRAY>, <BIG_ARRAY>)) {
  // The whole product array is first computed and loaded into memory
  // before an iterator over it is created by the engine...
  // Probably not what the user thought would happen
}

[eranhirsch]

@TkDodo - wdyt?