Add SliceType and ArrayPointerType to std.meta

[marler8997]

Provide a couple functions in std.meta to transform pointer/array types to the slice/array pointer variations. Note that the key feature here is that it preserves all pointer attributes (i.e. const, volatile, etc)

    testing.expect(SliceType([*]u8) == []u8);
    testing.expect(SliceType([*]const u8) == []const u8);

    testing.expect(ArrayPointerType([]u8) == [*]u8);
    testing.expect(ArrayPointerType([]const u8) == [*]const u8);

I've converted 3 projects to zig and have already made use of these functions in all 3 projects.

[andrewrk]

This should be unblocked now.

[marler8997]

Not sure how much testing I should do here...

[andrewrk]

Basic functionality is fine.

[marler8997]

Is Type redundant? Should it just be Slice and ArrayPointer rather than SliceType and ArrayPointerType?

[andrewrk]

I prefer without the Type suffix since we have a pretty consistent formatting style that types are capitalized.

[andrewrk]

Let's have this either be a real test case or deleted. It's not clear what this comment is communicating here.

How about this test case?
testing.expect(Slice(*[10]u8) == []const u8);

Same comment for the other commented out test case below.

[marler8997]

I was thinking we would integrate #3173 first, then uncomment this.

[andrewrk]

I think it's a smell that the function would accept a non-reference type and return a reference type. I think the test case I provided is the one you're looking for here.

[marler8997]

Oh sorry I misunderstood your point. Yes I think *[N]T makes more sense than [N]T. Will change

[marler8997]

So the question is, should Slice(*[N]T) become [][N]T or []T?

[andrewrk]

Slice(*[N]T) should become []T to be consistent with how type coercion works in the rest of the language.

[marler8997]

Hmmm, given that answer, I'm not sure what the rules are. It sounds like we have one or more special cases for single-value pointers? What would these cases be?

if (single_value_pointer)
    if (pointer_to_array) return []T.child.child
    else return []T.child

Are there any other cases where we would return []T.child.child?

[andrewrk]

Are there any other cases where we would return []T.child.child?

Single item pointer to array is the only case that matches type coercion semantics

[andrewrk]

Feel free to re-open when this is ready for review

[marler8997]

@andrewrk I had an outstanding question I was waiting for you to answer.