proposal: x/exp/slices: add Pop

[thatradius]

// Pop removes the elements s[i] from s, returning the modified slice
// and the removed element. Pop panics if s[i] is not a valid slice of s.
func Pop[S ~[]E, E any](s S, i int) (S, E) {
	e := s[i]
	return append(s[:i], s[i+1:]...), e
}

[ianlancetaylor]

Can you write a doc comment for the proposed function? Specifically, what happens when the slice is empty.

Seems odd to have Pop without Push.

[thatradius]

@ianlancetaylor Like Delete but returns the removed element.

[thatradius]

About Push maybe just like this.

// Push appends the element at s[:i], returning the modified slice.
func Push[S ~[]E, E any](s S, i int, e E) S {
	return append(append(s[:i], e), s[i+1:]...)
}

[seankhliao]

That Push is the same as Insert

Push/Pop seems like a strange name for the given actions, usually they mean operating at the ends of a queue

[thatradius]

Maybe change the method name to Remove like Rust's vector?

[DeedleFake]

I agree that Pop is a strange name for it, but Remove for parity with Insert makes sense to me.

[thatradius]

What about this?

// Pop removes the element s[i] from s, returning the modified slice
// and the removed element. Pop panics if i is not valid index of s.
func Pop[S ~[]E, E any](s S, i int) (S, E) {
	switch i {
	case 0:
		return s[1:], s[0]
	case len(s) - 1:
		return s[:len(s)-1], s[len(s)-1]
	}
	return append(s[:i], s[i+1:]...), s[i]
}

// Remove removes the elements s[i:j] from s, returning the modified slice
// and the removed elements. Remove panics if s[i:j] is not a valid slice of s.
func Remove[S ~[]E, E any](s S, i, j int) (S, S) {
	var s2 S
	if n := j - i; cap(s) >= len(s)+n {
		// Use the free space to avoid an allocation.
		s2 = s[len(s) : len(s)+n]
	} else {
		s2 = make(S, n)
	}
	copy(s2, s[i:j])
	return append(s[:i], s[j:]...), s2
}

[earthboundkid]

Python has dict.pop which works like this. I don't think this behavior really makes sense for a slice though. It's not a very substantial savings versus just doing:

e := s[i]
s = slices.Delete(s, i, i+1)

It saves you a line, but it adds noise to the docs and open questions about behavior, when really we should be discouraging using O(N) operations more than necessary.

[rsc]

Pop should not be an O(N) operation. Saving s[i] and calling Delete seems fine for this use case. Delete is a bit problematic already; better not to have two.

[rsc]

This proposal has been added to the active column of the proposals project
and will now be reviewed at the weekly proposal review meetings.
— rsc for the proposal review group

[nightlyone]

Pop could also return a book whether the element was found.

Then the panic on empty slice would be handled in a way that is consistent with e.g. the map check

Adding more ways to panic in Go doesn't sound particularly Go like to me, as Go programs are supposed to handle such conditions explicitly instead of checking preconditions and react on failure to do so.

Slice access is one of the notable exceptions.

[earthboundkid]

Go programs are supposed to handle such conditions explicitly instead of checking preconditions and react on failure to do so.

I agree "Go programs are supposed to handle such conditions explicitly" but I don't see how it follows that being explicit is "instead of checking preconditions". For map, if you couldn't do _, ok := m[k] how would you explicitly check that the map has a key? Comma ok is checking the precondition. For a slice, the equivalent check is if len(s) > i. There doesn't need to be comma ok version of slice access because we have if len(s) > i.

Anyway, Pop is O(N) and really should not be encouraged.

[rsc]

Based on the discussion above, this proposal seems like a likely decline.
— rsc for the proposal review group

[rsc]

No change in consensus, so declined.
— rsc for the proposal review group