typedesc[T] can't be used at compile time, unlike generics

[timotheecour]

typedesc[T] can't be used at compile time, unlike generics, as shown below:

when defined(case_default_old):
  template default*[T](t: typedesc[T]): T =
    # Error: cannot evaluate at compile time: v116407 var v: T
    var v: T
    v
else:
  proc default*[T]():T=
    var a: T
    return a

proc len(T: typedesc[tuple]):auto=
  when defined(case_default_old):
    const t=default(T)
  else:
    const t=default[T]()

  var n=0
  for _ in t.fields:
    inc n
  return n

proc test()=
  var a=(1,2,3)
  const length = a.type.len
  echo length

test()

Can that limitation be lifted?

NOTE: this affect the debate here: [RFC] guidelines for when to use typedesc vs generics #7517
NOTE: the motivating example was from default should be public instead of redefined everywhere; should use default[T] instead of default(T) #8485 ; if this limitation were lifted, we'd be able to use default(T) (or int.default) at compile time, instead of having to resort to default[T]()

EDIT /cc @yglukhov NOTE: see workaround below via intermediate defaultImpl[T]():T

proc defaultImpl[T]():T=
  var a: T
  return a
template default*[T](t: typedesc[T]): T = defaultImpl[T]()

[yglukhov]

Works with devel

template default*[T](t: typedesc[T]): T =
    var v: T
    v
const a = default(int)

[timotheecour]

@yglukhov I know, I'm using devel; it works in your example, but not the example I mentioned above:

rnim -d:case_default_old bugs/t74_typedesc_CT.nim
nim c --nimcache:/tmp/nim//nimcache/ -o:/tmp/nim//app -r -d:case_default_old bugs/t74_typedesc_CT.nim
/Users/timothee/git_clone/nim/timn/bugs/t74_typedesc_CT.nim(26, 24) template/generic instantiation from here
/Users/timothee/git_clone/nim/timn/bugs/t74_typedesc_CT.nim(6, 9) Error: cannot evaluate at compile time: v116407
      var v: T
          ^


rnim bugs/t74_typedesc_CT.nim
nim c --nimcache:/tmp/nim//nimcache/ -o:/tmp/nim//app -r bugs/t74_typedesc_CT.nim
CC: timn_t74_typedesc_CT
Hint: operation successful (12210 lines compiled; 0.614 sec total; 16.25MiB peakmem; Debug Build) [SuccessX]
3

[yglukhov]

My bad, sorry.

[dom96]

Duplicate of #7596. (Please consider grouping related issues together).

[timotheecour]

@dom96 how about adding a github label duplicate? (I can't as I don't have the required priviledge in this repo); so this can be marked as closed and duplicate

[dom96]

Why would it be useful to find duplicated issues?

[andreaferretti]

https://github.com/nim-lang/Nim/issues?utf8=%E2%9C%93&q=is%3Aissue+is%3Aclosed+duplicate

[timotheecour]

Why would it be useful to find duplicated issues?

=> moved discussion here #8507