Explain why to use :> instead of Cons

[tomjaguarpaw]

There is a warning to use :> instead of Cons. Could you also explain in the warning the reason for it?

[ggreif]

I guess (:>) is an abstract interface, Cons the concrete one. Possibly later the concrete interface won't be exposed to the module users.

As an aside, I have a little project underway to do run-length-encoded Vectors for simulation, which would not have Cons any more.

[tomjaguarpaw]

If that's the case it would be helpful to have "Cons will be deprecated in future" in the docs.

[christiaanb]

I'll update the documentation.

What @ggreif said is correct, :> is the abstract interface, while Cons is the concrete one.

The reason is a bit silly/sad: While the clash compiler happily supports the constructor aspect of GADTs, it (currently) does not support the pattern matching aspect (see the end of chapter 4 of my thesis if you want to know why). So I added the pattern synonym (:>), which is based on head and tail for the pattern-matching side, which are supported by the compiler.

Hopefully, if/when clash supports GADT patterns, the (:>) will become the concrete interface.

[martijnbastiaan]

GADT support has recently landed in Clash (#465). We can now use Cons :). We should remove the warning and remove the pattern synonyms currently present.

[christiaanb]

The warning has been removed. We keep the pattern synonym for the reasons described in: #966