Re: Problem with +numberWithBool: ?

[Helge Hess]

On Jan 29, 2004, at 2:12 PM, Richard Frith-Macdonald wrote:
> Booleans in objc can be YES or NO ... this is not the same thing as 
> true/false in C/ObjC conditionals where 0 is false and anything 
> non-zero is true.

Such an assumption (which is probably not backed by the ObjC "spec") 
certainly breaks loads of existing code and is IMHO a completely 
unnecessary and highly error prone.

> So ...
>
> if (x)...
> is not the same as
> if (x == YES)
> where x is of type BOOL

Yes, this is not the same, but doesn't relate to BOOL. This is like 
stating that
  if (x)...
is not the same like
  if (x == 5)
which is true but doesn't relate to the actual question of the "truth" 
value of an expression.

Indeed BOOL is not an enumeration type of YES and NO (which would back 
your statement), but YES and NO are regular defines. If it would be 
defined as

  typedef enum {NO, YES } BOOL;

in objc.h, you would have a point ;-)

> which is why I prefer to write code which explicitly compares booleans 
> values with YES or NO.

Hopefully noone else does that. Just fixed a whole lot of OGo code 
regarding that issue.

Well, I actually try to return YES or NO explicitly do work around IMHO 
buggy code like the above, eg:
  return [self compare:obj] != 0 ? NO : YES;

Greets,
  Helge
--
OpenGroupware.org => http://www.opengroupware.org/