Mirror of forum.nim-lang.org

1215 :: Named ranges

• Quikli (orginal) [2015-05-12T03:51:46+02:00] view original

  I was recently using a named range to give me bounds checking, when I hit a bug and realized I hadn't initialized the value and the default value was 0, instead of my lower bound of 1. I tested a few other things to prove that other than uninitialized variables, it did in fact bounds check it. Is this expected behaviour or a bug? Nim version 0.11.0?

  type
    Coordinate = 1..20
  
  var
     test: Coordinate
  
  echo test # prints 0, outside of the bounds of Coordinate
  test = 0 # compile Error: conversion from int literal(0)
  test = test - 1 # runtime unhandled exception: value out of range -1 [RangeError]
  Run

---

• Araq (orginal) [2015-05-12T10:34:46+02:00] view original

  It's a bug, the compiler should produce a "cannot prove 'test' is initialized" warning for this. Please report.

  EDIT Sorry, misread, that's not a bug at all. You have to use this syntax:

  type
    Coordinate = range[1..20]
  
  var
     test: Coordinate
  
  echo test
  Run

---

• Quikli (orginal) [2015-05-12T18:49:17+02:00] view original

  Thanks Araq! Using range[1..20] instead of just 1..20 does indeed produce the compiler warning. The generated C code appears to be the same with either syntax. Is there a reason the compiler accepts the code the way I wrote it and doesn't complain? I stumbled upon it while trying to reduce the number of times in my code I was expressing that range. The first thing I tried 1..20 was accepted and worked the way I expected (until I forgot to initialize a variable of that type).

---

• Araq (orginal) [2015-05-12T20:12:40+02:00] view original

  Yeah I'm wrong there certainly is a bug somewhere as either 'x..y' shouldn't be allowed or produce a proper range type.

---