Nil error not catched on compile time

[alehander92]

type
  Style = ref object
    color: int

proc f*(style: Style) =
  if not style.isNil:
    echo style.color
  else:
    echo style.color

f(Style(color: 2))
f(nil)

I expect that f wouldn't be able to compile, as it's obvious that its fields are inaccessible in the else branch.

Nim can already prove a lot of nil related errors, so catching those should be a good step forward for type safety (nil runtime errors are obviously a very serious problem)

[lightness1024]

careful about asking for permission. I'd say warning with a flag --warningaserror for the jackasses.

[alehander92]

@lightness1024 Nope, it shouldn't be a flag in long term

In this particular case, the user already "asked for permission", so that should be a 100% error, as it's provably a bug.

if you don't want to ask for permission AND that's not actually a bug, you write

proc f*(style: Style not nil) =
  ..

[lightness1024]

maybe. to go your way, it would very certainly help to avoid dangerous liberties that C compilers are allowed to take on statically provable null dereferences.