Nim: deprecate typedesc in favor of type

[dawkot]

var x: int = 0
proc foo(T: type): var T = x
foo int

template/generic instantiation of `foo` from here
inim_1542371335.nim(4, 28) Error: type mismatch: got <int> but expected 'None'

[krux02]

please use correct markdown, otherwise I have to manually remove all those nim> directives in your snippet, like you have to do it now in my snippet, which btw solves your problem.

nim> var x: int = 0
nim> proc foo[T](_: typedesc[T]): var T = x
nim> foo(int) = 17
nim> echo x

This remains an issue though, because there should be an error message that forbids the use of type in the expression of the parameters.

[bluenote10]

According to the typedesc reforms PR, proc(T: type) is now the same as proc(T: typedesc), isn't it? Why should it be forbidden again?

[narimiran]

proc(T: type) is now the same as proc(T: typedesc), isn't it?

Yes, the original example throws the same error regardless if one uses type or typedesc. Analogously, @krux02's example works whichever version you use.

[krux02]

I wasn't aware of that PR. I guess then typedesc should be deprecated in favor of type. I don't like it when things can be referred by different names. It is very confusing to me when I stumble upon a project just just uses a different set of aliases than I do.

[mratsim]

This seems like a better test case, the original one is actually a closure which might brings yet other issues:

var x: int = 0

proc foo(T: type, x: var T): var T = x
foo(int, x)

Error: type mismatch: got <type int, int>
but expected one of:
proc foo(T: type; x: var type): var type
  first type mismatch at position: 2
  required type: var None
  but expression 'x' is of type: int

expression: foo(int, x)

[Araq]

No, we should deprecate type in favour of typedesc.