adapters::cartesian_product does a bit too much

[reuvenpo]

Looking at how pub fn cartesian_product<I, J>(mut i: I, j: J) -> Product<I, J> initialises Product, one can see that it yields the first item in i.
Wouldn't it be better if Product.a_cur was an Option<Option<I::Item>> and initialized as None, rather than initializing it by starting the consumption of i?

This may cause problems if iteration over i has side effects.

Same with Product.b too. No reason to clone it ahead of time.

I haven't looked deep into at the implementation of multi_cartesian_product but it may be a similar situation there too.

[Philippe-Cholet]

I agree, I came to the same conclusion and same "nested option" solution, implemented in #800. More about laziness in #791. It took a while but it's solved.

[reuvenpo]

Wow, I have no idea what i could've needed this for back in 2018, but i might actually need it soon for something so... thanks for fixing it - good timing! 😆