I was looking at some of the modules such as ctype and math, and see that the "is" functions (i.e. isalpha, isnan) return a value of type int.  Given that this is a modern language having boolean types available, would it not be better to return a value of type boolean?  I'm thinking that code would be clearner.

Sure would, but we can't agree on what boolean would be!

I think it should be an integral type, whose size is the same as that of the architecture (for performance; the various supporting reasons for this are in my <self-serving promotion>forthcoming book "Imperfect C++"</self-serving promotion>), and which should *not* be implicitly convertible to/from any other integral type.

But Walter, and a handful of other people, seem content to have it be aliased to bit.

In the libraries I write, I typedef (remember D's typedefs are strong ones) boolean from int, and use the casts as appropriate to return them. Walter doesn't like this, but I'm immutable on this point.

"Peter" <Peter_member@pathlink.com> wrote in message news:c74i4i$l54$1@digitaldaemon.com...
> I was looking at some of the modules such as ctype and math, and see that the "is" functions (i.e. isalpha, isnan) return a value of type int.  Given that this is a modern language having boolean types available, would it not be
better
> to return a value of type boolean?  I'm thinking that code would be clearner.
>
>

Matthew wrote:

>Sure would, but we can't agree on what boolean would be!
>
>I think it should be an integral type, whose size is the same as that of the
>architecture (for performance)
>  
>
>But Walter, and a handful of other people, seem content to have it be aliased to
>bit.
>  
>
Won't bit be an integral type be a size that is optimal for the architecture.  I mean for the systems we use now byte is the most optimal choice for Boolean (due to caching and other factors).  If your using a single bit then it will be the size of a byte.

-- 
-Anderson: http://badmama.com.au/~anderson/

"Peter" <Peter_member@pathlink.com> wrote in message news:c74i4i$l54$1@digitaldaemon.com...
> I was looking at some of the modules such as ctype and math, and see that
the
> "is" functions (i.e. isalpha, isnan) return a value of type int.  Given
that
> this is a modern language having boolean types available, would it not be
better
> to return a value of type boolean?  I'm thinking that code would be
clearner.

Generally, a computation of 0 or !0 involves one or two less instructions than converting it to 0 or 1.

Walter wrote:
> "Peter" <Peter_member@pathlink.com> wrote in message
> news:c74i4i$l54$1@digitaldaemon.com...
> 
>>I was looking at some of the modules such as ctype and math, and see that
> 
> the
> 
>>"is" functions (i.e. isalpha, isnan) return a value of type int.  Given
> 
> that
> 
>>this is a modern language having boolean types available, would it not be
> 
> better
> 
>>to return a value of type boolean?  I'm thinking that code would be
> 
> clearner.
> 
> Generally, a computation of 0 or !0 involves one or two less instructions
> than converting it to 0 or 1.
> 
> 

Well, then maybe a boolean shouldn't be a bit (hint,hint), but its own type where true simply means !=0.

IMHO, your using int instead of bool for a situation where you really MEAN a boolean is a tell-tale sign that something is wrong with the current bool implementation.

Not that it hasn't been mentioned before ;).

Hauke

Matthew 提到:
> Sure would, but we can't agree on what boolean would be!
> 
> I think it should be an integral type, whose size is the same as that of the architecture (for performance; the various supporting reasons for this are in my <self-serving promotion>forthcoming book "Imperfect C++"</self-serving promotion>), and which should *not* be implicitly convertible to/from any other integral type.
> 
> But Walter, and a handful of other people, seem content to have it be aliased to bit.
> 
> In the libraries I write, I typedef (remember D's typedefs are strong ones) boolean from int, and use the casts as appropriate to return them. Walter doesn't like this, but I'm immutable on this point.
> 
> "Peter" <Peter_member@pathlink.com> wrote in message news:c74i4i$l54$1@digitaldaemon.com...
> 
>>I was looking at some of the modules such as ctype and math, and see that the "is" functions (i.e. isalpha, isnan) return a value of type int.  Given that this is a modern language having boolean types available, would it not be
> 
> better
> 
>>to return a value of type boolean?  I'm thinking that code would be clearner.
>>
It would be even better if you write a book called "Learning D" or "The D Programming Language" or "Thinking in D" or "Prefect D".

"School" <itsnotvalid-news@yahoo.com.hk> wrote in message news:c75hcv$23rj$1@digitaldaemon.com...
> Matthew ´£¨ì:
> > Sure would, but we can't agree on what boolean would be!
> >
> > I think it should be an integral type, whose size is the same as that of the architecture (for performance; the various supporting reasons for this are in
my
> > <self-serving promotion>forthcoming book "Imperfect C++"</self-serving promotion>), and which should *not* be implicitly convertible to/from any
other
> > integral type.
> >
> > But Walter, and a handful of other people, seem content to have it be aliased
to
> > bit.
> >
> > In the libraries I write, I typedef (remember D's typedefs are strong ones) boolean from int, and use the casts as appropriate to return them. Walter
doesn't
> > like this, but I'm immutable on this point.
> >
> > "Peter" <Peter_member@pathlink.com> wrote in message news:c74i4i$l54$1@digitaldaemon.com...
> >
> >>I was looking at some of the modules such as ctype and math, and see that the "is" functions (i.e. isalpha, isnan) return a value of type int.  Given that this is a modern language having boolean types available, would it not be
> >
> > better
> >
> >>to return a value of type boolean?  I'm thinking that code would be clearner.
> >>
> It would be even better if you write a book called "Learning D" or "The D Programming Language" or "Thinking in D" or "Prefect D".

Well, you may be onto something there. ;)

"Hauke Duden" <H.NS.Duden@gmx.net> wrote in message news:c752e6$1e7a$1@digitaldaemon.com...
> Walter wrote:
> > Generally, a computation of 0 or !0 involves one or two less
instructions
> > than converting it to 0 or 1.
> Well, then maybe a boolean shouldn't be a bit (hint,hint), but its own type where true simply means !=0.

I'm just one of those people, probably very few in number, who think that C got it right by using 'int' for it and testing for 0 or !0. It's very efficient, and I am loathe to give that up for anything less than a thoroughly compelling reason. It's served me well for 20+ years of C coding.

There are 3 scenarios for a boolean:

1) need it to be fast
2) need a single value to be small
3) need a packed array of boolean values

'int' serves (1) well. D's 'bit[]' serves (3), and 'bit' does double duty by
also being a byte wide value when used outside of an array, serving (2).

> IMHO, your using int instead of bool for a situation where you really MEAN a boolean is a tell-tale sign that something is wrong with the current bool implementation.

While I see your point, I don't view it that way. I view it as getting comfortable using an idiom that maps well onto efficient execution on the hardware. Perhaps one reason I am so comfortable with it is all the assembler programming I've done.

> Not that it hasn't been mentioned before ;).

Indeed. I recall the usenet flame wars on this back in the 1980's. The various camps have all long since dug in and well fortified their positions, so I don't think anyone is going to change their mind <g>.

http://www.thedprogramminglanguage.com/

Wonder who owns that domain name...

Matthew wrote:
> "School" <itsnotvalid-news@yahoo.com.hk> wrote in message
> news:c75hcv$23rj$1@digitaldaemon.com...
> 
>>Matthew ´£¨ì:
>>
>>>Sure would, but we can't agree on what boolean would be!
>>>
>>>I think it should be an integral type, whose size is the same as that of the
>>>architecture (for performance; the various supporting reasons for this are in
> 
> my
> 
>>><self-serving promotion>forthcoming book "Imperfect C++"</self-serving
>>>promotion>), and which should *not* be implicitly convertible to/from any
> 
> other
> 
>>>integral type.
>>>
>>>But Walter, and a handful of other people, seem content to have it be aliased
> 
> to
> 
>>>bit.
>>>
>>>In the libraries I write, I typedef (remember D's typedefs are strong ones)
>>>boolean from int, and use the casts as appropriate to return them. Walter
> 
> doesn't
> 
>>>like this, but I'm immutable on this point.
>>>
>>>"Peter" <Peter_member@pathlink.com> wrote in message
>>>news:c74i4i$l54$1@digitaldaemon.com...
>>>
>>>
>>>>I was looking at some of the modules such as ctype and math, and see that the
>>>>"is" functions (i.e. isalpha, isnan) return a value of type int.  Given that
>>>>this is a modern language having boolean types available, would it not be
>>>
>>>better
>>>
>>>
>>>>to return a value of type boolean?  I'm thinking that code would be clearner.
>>>>
>>
>>It would be even better if you write a book called "Learning D" or "The
>>D Programming Language" or "Thinking in D" or "Prefect D".
> 
> 
> Well, you may be onto something there. ;)
> 
> 

If you look at the whois entry, you will see that the owner is Synesis Software and the administrative contact is
Dr Matthew D. Wilson

Brad Anderson wrote:
> http://www.thedprogramminglanguage.com/
> 
> Wonder who owns that domain name...
>