Mirror of forum.nim-lang.org

1298 :: Why doesn't system.nim use ``not nil`` annotations?

• judofyr (orginal) [2015-06-07T11:58:29+02:00] view original

  For instance, proc len(x: string): int requires that the string is not nil. Is there a specific reason why this isn't annotated as proc len(x: string not nil): int?

---

• Araq (orginal) [2015-06-07T13:18:19+02:00] view original

  system.len can deal with nil. Since 0.11.0.

  The stdlib doesn't use it since for now as it would break code and I'm still not entirely convinced that it's not too tedious to work with in practice.

---

• judofyr (orginal) [2015-06-07T14:59:54+02:00] view original

  I see. It feels difficult to justify "you should use not nil annotations for safety" when we're not able to use it properly in the stdlib ourself. If it's too tedious to work with, maybe it means we need to improve nil-prover to make it less tedious? What makes it tedious today, by the way? It would be great if we could say that every proc in stdlib that can't handle nil is annotated with not nil.

---