Please add maxElement and minElement in mir (trying to use std.algorithm gives opCmp not defined)

[prbuen]

To find the maximum value in an ndslice u, we can use mir's maxIndex:

u[u.maxIndex]

but it would be nice to have the functions maxElement and minElement directly implemented in mir. It's not possible to use these functions from std.algorithm (the complaint is that opCmd is not defined), unless doing u.flattened.maxElement.

It doesn't sound like much but it would allow to more easily write code that can apply to both standard D dynamic arrays and ndslices.

Here is my use case: in numerical computation (e.g. errors) we often need to find the maximum absolute value of an ndslice. This can be achieved both with dynamic arrays (were it not for this bug) and ndslices with

u[u.map!fabs.maxIndex].fabs

but it's pretty heavy and not as readable compared to

u.map!fabs.maxElement

Thanks!

[jmh530]

Something like:

/++
Returns the minimal(maximal) element of a multidimensional slice.

Params:
    pred = A predicate.

See_also:
    $(LREF minIndex),
    $(LREF maxElement),
    $(LREF maxIndex),
    $(LREF maxPos).
+/
template minElement(alias pred = "a < b")
{
    import mir.functional: naryFun;
    static if (__traits(isSame, naryFun!pred, pred))
    /++
    Params:
        slice = ndslice.
    Returns:
        Minimal(maximal) element of a multidimensional slice
    +/
    @fmamath size_t[N] minElement(Iterator, size_t N, SliceKind kind)(Slice!(Iterator, N, kind) slice)
    {
        return slice[slice.minIndex!pred];
    }
    else
        alias minElement = .minElement!(naryFun!pred);
}

/// ditto
template maxElement(alias pred = "a < b")
{
    import mir.functional: naryFun, reverseArgs;
    alias maxElement = minElement!(reverseArgs!(naryFun!pred));
}

[prbuen]

Thanks, that's great! Compilation fails with cannot implicitly convert expression 'slice.opIndex(minIndex(slice))' of type 'double' to 'ulong[<n>]' on the line

      return slice[slice.minIndex!pred];

but changing the return type of minElement from size_t[N] to auto makes it work.

Both u.maxElement and u.map!fabs.maxElement (and the minElement counterparts) work as expected. Would be great to merge :)
Thanks again!

[jmh530]

Yeah, I just kind of threw that together without testing. I'd prefer an actual return type instead of auto.
I don't have a lot of free time these days, but I can try to put together the PR soon-ish.

[prbuen]

That would be great. I agree with preferring an actual type but don't feel confident enough here to know how to write it. FWIW, std.algorithm.maxElement does return auto.

[jmh530]

@prbuen Implementation here: #473