std.mem: add indexOfMin and indexOfMax

[zzyxyzz]

Find the index of the smallest/largest number in a slice, without traversing it twice.

closes #9891

[LemonBoy]

The comparison predicate should be specified by the user to allow for extra flexibility and custom non-scalar types.

[zzyxyzz]

I'm not sure about that. The precedent from std.math.min and std.mem.min is to only support scalar types. A custom predicate would be good to have, but maybe open an issue first?

[zzyxyzz]

Thanks for the review, @LemonBoy.

I've added the assert and more tests to the new functions, as well as to min and max, whence they were copied.

[zzyxyzz]

Drone failure unrelated?

[N00byEdge]

Hold on, there needs to be a indexOfMinMax too that gets both in a single pass, right?

[zzyxyzz]

@N00byEdge

Hold on, there needs to be a indexOfMinMax too that gets both in a single pass, right?

In principle, yes. But I'm not sure what the function signature would be. There doesn't seem to be an std.mem.minMax() to take inspiration from, and I don't know what the Zig-standard way of returning multiple values is in a situation like that. Any suggestions?

[InKryption]

How about something explicit like struct { min: usize, max: usize }? That's basically the current resolution reached so far in #498, as I understand it.
Edit: or maybe struct { index_min: usize, index_max: usize } would be more clear.

[zzyxyzz]

Added combined min-max functions in a separate commit.

[Snektron]

There is already an equivalent of the minIndex and maxIndex functions in std.sort:

zig/lib/std/sort.zig

Lines 1413 to 1418 in ee98d87

	pub fn argMax(
	comptime T: type,
	items: []const T,
	context: anytype,
	comptime lessThan: fn (context: @TypeOf(context), lhs: T, rhs: T) bool,
	) ?usize {

zig/lib/std/sort.zig

Lines 1361 to 1366 in ee98d87

	pub fn argMin(
	comptime T: type,
	items: []const T,
	context: anytype,
	comptime lessThan: fn (@TypeOf(context), lhs: T, rhs: T) bool,
	) ?usize {

[zzyxyzz]

Interesting. I wouldn't quite call it an equivalent, as these functions take an explicit comparator and corresponding context -- which I think is too verbose for general use. But if it's felt that argMin/argMax are good enough, then this PR can be closed.

[SpexGuy]

Going to close and reopen to kick off CI

[zzyxyzz]

Drone failure looks unrelated, but I'm not quite sure. Has it been seen before? Something about missing references to "ceill" and "florl".