Pair sample class

[Charles]

Re: Pair inheriting Pair:

Right now Pair<of T> has a strongly typed indexer and enumerable interface:

class Pair<of T> implements IEnumerable<of T>

    get [i as int] as T

class Pair<of TA, TB> implements System.Collections.IEnumerable 

    get [i as int] as dynamic?

I think the strongly typed enumerable could be saved, but not the indexer. Maybe that's okay, but I wanted to point it out. Too bad .NET does not have a concept of "greatest common denominator" for types:

class Pair<of TA, TB> implements System.Collections.IEnumerable 

    get [i as int] as gcd(TA, TB)

Cobra has such a concept in certain places like when it infers the type of collection literals such as [Square(), Circle()] which would be List<of Shape>. However, using it in return types would be fairly meaningless because run-time construction of .NET generics wouldn't implement it.

[jaegs]

Seems like a worthwhile sacrifice.

[Charles]

I've made some progress on this, but when I try it out for the intended purpose, the compiler falsely complains that Pair<of int> cannot be passed where Pair<of int, int> is expected. I'll have to work that out.

[Charles]

False alarm on the type compatibility.

This is done now in changeset:2790.

Also, I noticed Pair<of TA, TB>'s .init method was still taking dynamically typed params, no doubt a left over from the original Pair class which was all dynamic. This has been improved as appropriate for generics:

-       cue init(a, b)
+       cue init(a as TA, b as TB)

One notable change is that the indexer on Pair<of TA, TB> must necessarily return "dynamic?" since nothing is known about what TA and TB have in common. Hence there were several changes in MultiList such as:

-               _shape = for range in ranges get range[1] - range[0]
+               _shape = for range in ranges get range.b - range.a

-                   addr += (index + _ranges[i][0]) * _strides[i]
+                   addr += (index + _ranges[i].a) * _strides[i]

...because .a and .b can be strongly typed since they are separate properties.

Please take this for a test spin and let me know if you have any issues.