Mirror of forum.nim-lang.org

843 :: Difference between static[T] and T{lit} parameter constraint?

• jboy (orginal) [2015-02-08T07:58:22+01:00] view original

  Hi all,

  I'm just wondering if anyone can explain / clarify / elaborate upon the difference between static[T] (for example, static[int] or static[string]: http://nim-lang.org/manual.html#static-t ) and the T{lit} parameter constraint (for example, int{lit} or string{lit}: http://nim-lang.org/manual.html#parameter-constraints , http://nim-lang.org/manual.html#typedesc )?

  So far, my best understanding is this:

  • static[T] is used for the parameters of procs (rather than templates or macros) or the generic parameters of generic procs or types. A static[T] proc parameter causes the proc to become implicitly generic, and be "compiled separately for each unique supplied value".
  • The T{lit} parameter constraint is used for the parameters of templates and macros (rather than procs or types).

  Is this understanding correct? Is there any more to it than this?

  Thanks for any clarification.

---

• LeuGim (orginal) [2015-02-08T09:22:55+01:00] view original

  Is there any more to it than this?

  T{lit} can be used for procs, e.g. for overloading:

  proc p(n: int{lit}): int = n*10
  proc p(n: int): int = n*100
  echo 5.p      # 50, 1st proc called
  let five = 5
  echo five.p   # 500, 2nd proc called
  Run

  ---

  In tests int{lit} matches also consts (const five = 5) and compile-time computed expressions (like 2 + 3), though manual says {lit} is only for literals (like 5). I don't know which behavior is intended.

---

• Araq (orginal) [2015-02-08T09:49:21+01:00] view original

  In theory T{lit} is purely a syntactic match, static[T] is a semantic match. In practice lit matches more than it should and static[T] is buggy and these two things should be unified, somehow.

---

• jboy (orginal) [2015-02-09T02:20:06+01:00] view original

  @LeuGim, thanks for that clarification that T{lit} can also be used for procs! And the int{lit} vs int proc parameter overloading effect is very interesting -- I hadn't encountered that before.

  @Araq, thanks for the clarification on the theoretical differences. The intention of differentiating syntactic vs semantic matching is an intriguing distinction. Now that I understand the intention of syntactic matching, that language feature appeals to me.

  I think I also understand what is intended by semantic matching using static[T], but it's more subtle. And as a beginner with the language, I agree that it would be a little easier on beginners if those two different concepts (syntactic vs semantic matching) could be unified somehow into a single language feature. :)

  In the meantime, I'll read a bit about the Nim compiler architecture: http://nim-lang.org/intern.html#the-compiler-s-architecture

  By the way, I just skimmed that "Compiler Internals" page, and I saw a mention of a non-existent ast module: http://nim-lang.org/intern.html#the-syntax-tree

  I assume that should be updated to point to the macros module?

---

• filwit (orginal) [2015-02-09T03:49:49+01:00] view original

  jboy: I saw a mention of a non-existent ast module

  The ast module is part of the compiler, not the standard libs.

---

• jboy (orginal) [2015-02-09T04:42:12+01:00] view original

  Ah, thanks.

---