Re: [RFA]: BOOL coding standards (Was: Problem with +numberWithBool: ?)

[Pascal J . Bourguignon]

Helge Hess writes:
> On 02.02.2004, at 20:47, Nicola Pero wrote:
> >> I would again like to point out that if the intention was to limit 
> >> BOOL
> >> to YES and NO, it would have been defined as
> >>    typedef enum { YES, NO } BOOL;
> >
> > Wouldn't this have been a waste a space just to just store a YES/NO ?
> 
> Well, isn't that you who is suggesting that a "true boolean" is worth 
> major restrictions? ;-)
> 
> Your point is pretty mood because:
> a) BOOL is *NOT* specified to be char and indeed is not mapped to char 
> on some platforms


 
> Is int more wasteful than that:
> @interface MyFunkyObject
> {
>    BOOL a;
>    BOOL b;
> }
> I guess not, because a and b are probably aligned to 32 bits 
> boundaries. 

You  lucky guy!   you must  be  working on  a Cray.   That's the  only
architecture I  know where bytes are aligned  as integers.  Everywhere
else, bytes are ligned on byte boundaries.  That's the point of having
bytes.



> Is
> - (void)doIt:(BOOL)_a
> more wasteful than int? I also guess not, because the register _a is 
> being passed in is available in any case and probably not going to be 
> shared by multiple char variables.

???   What   are  you  speaking  about  here?    Argument  passing  is
different. Because of the way C  allows to pass any kind of arguments,
it require them to be all  aligned the same.  Register have nothing to
do here (but perhaps  on heavily optimizing compilers on architectures
that have a lot of register / register windows).
 

> In short: the practical gain of using char for BOOL is minimal, 
> especially in comparison to the already wasteful use of char instead of 
> a bit.

NeXTSTEP  was designed  on computers  with 8MB  RAM.  While  there was
virtual memory and  the constraints were not as strong  as a 64KB RAM,
it's obvious that  they did not was space  gratuituously: the used bit
fields for boolean attributes as much as possible in all classes.


> b) even char is as wasteful requiring 8 times the memory necessary
>     (which is why everyone is using bitfields anyway) - apparently it is
>     worth that

By the way, I'm sure we can find some BOOL value stored in a 1-bit field.


> PS: this discussion is a lot of fun - and I indeed learned a lot I 
> didn't knew before ;-)

-- 
__Pascal_Bourguignon__                     http://www.informatimago.com/
There is no worse tyranny than to force a man to pay for what he doesn't
want merely because you think it would be good for him.--Robert Heinlein
http://www.theadvocates.org/