Idiomatic Flux way to deal with first and last elements differently?

[spinicist]

As per #244, what is the idiomatic way to handle the first, middle and last elements of a sequence differently?

My specific use case is handling the output of flux::chunk(). I can process the middle elements with .drop(1).take(N - 2).for_each(), but am not sure this is best or how to handle the first and last elements.

Thanks in advance.

[tcbrindle]

As it happens, this exact question was posed by @elbeno on a recent episode of ADSP.

I don't think we have any particularly good way of handling this in Flux at the moment, other than either by stepping through manually with a cursor or using take and drop as you suggest.

I'd like to offer a better solution if we can. I'm open to the idea of adding a function along the lines of

void for_each_with_first_and_last(sequence, func1, func2, func3);

i.e. you'd pass it three unary callbacks, and it would call func1 with the first element, then func2 repeatedly for each element except the last, and then finally func3 with the last element. Obviously the function would need a better name though!

An alternative approach might be to define an enum like

enum class sequence_pos { front, middle, back }; 

and then introduce a sequence adaptor which yields a pair of (underlying_element, sequence_pos). For example, given std::vector vec = {1, 2, 3, 4, 5}, we could imagine something like

for (auto [i, pos] : with_position(vec)) {
     std::println("{} {}", i, to_string(pos));
}

// prints
// 1 front
// 2 middle
// 3 middle
// 4 middle
// 5 back

I'm open to ideas, so if anyone has any thoughts or suggestions please let me know.

[tcbrindle]

This is the ADSP episode I was thinking of: https://adspthepodcast.com/2025/04/11/Episode-229.html

[elbeno]

FWIW I have an algorithm (using C++98-style "classic" iterators) called initial_medial_final which builds on another algorithm, for_each_butlast. https://intel.github.io/cpp-std-extensions/#_for_each_butlast_for_each_butlastn

[spinicist]

I have listened to that episode!

I like the for_each_with_first_and_last idea, although I might bikeshed the name given a chance. I think it matches @elbeno 's initial_medial_final.