Errm so what does "if (object)" and "Assert(object)" do??

Are they the same as (object !is null)?

cheers,

cw

"Chris Warwick" <sp@m.me.not> wrote in message news:est5n5$1krh$1@digitalmars.com...
> Errm so what does "if (object)" and "Assert(object)" do??

"if(object)" is the same as "if(object !is null)".

But "assert(object)" is a little tricky.  It's a little-known feature.  If you use "assert(object)", it will call any invariants defined for the object, i.e.

class C
{
    invariant
    {
        writefln("foo");
    }
}

...
scope c = new C();
assert(c); // prints foo since it calls the invariant

This means that if you use "assert(objectReference)" on a null reference, you will (like with '==' !) get an access violation, since it tries to look up the invariant from a null reference.

Therefore, when doing an assert on an object reference, you should always use "assert(c is null)" or "assert(c !is null)".

Jarrett Billingsley wrote:
> "Chris Warwick" <sp@m.me.not> wrote in message news:est5n5$1krh$1@digitalmars.com...
>> Errm so what does "if (object)" and "Assert(object)" do??
> 
> "if(object)" is the same as "if(object !is null)".
> 
> But "assert(object)" is a little tricky.  It's a little-known feature.  If you use "assert(object)", it will call any invariants defined for the object, i.e.
> 
> class C
> {
>     invariant
>     {
>         writefln("foo");
>     }
> }
> 
> ...
> scope c = new C();
> assert(c); // prints foo since it calls the invariant
> 
> This means that if you use "assert(objectReference)" on a null reference, you will (like with '==' !) get an access violation, since it tries to look up the invariant from a null reference.
> 
> Therefore, when doing an assert on an object reference, you should always use "assert(c is null)" or "assert(c !is null)". 
> 
> 

Really?
I thought assert(object) just throws an assert error if the reference is null!

Hasan Aljudy wrote:
> 
> Jarrett Billingsley wrote:
>> scope c = new C();
>> assert(c); // prints foo since it calls the invariant
>> 
>> This means that if you use "assert(objectReference)" on a null reference, you will (like with '==' !) get an access violation, since it tries to look up the invariant from a null reference.
>> 
>> Therefore, when doing an assert on an object reference, you should always use "assert(c is null)" or "assert(c !is null)". 
>> 
>> 
> 
> Really?
> I thought assert(object) just throws an assert error if the reference is null!

It's mentioned here:
http://www.digitalmars.com/d/class.html#invariants

The funny part is that assert(obj !is null && obj) does *not* call the invariant.  So assert(obj) is a special case.

I think assert(obj) being the same as assert(obj !is null), and then being able to do 'obj.invariant' if you really want to explicitly check the invariant would a lot clearer.  Or even 'invariant(obj)'.

Calling the invariant explicitly might be useful in the middle of methods.  Then I think this.invariant or invariant(this), is a lot more readable than assert(this).

You want both a null check and an invariant check, you could then do 'assert(obj && invariant(obj))' or assert(obj && obj.invariant)'.  Seems pretty obvious to me what that would do.

Maybe obj.invariant is the best way, since it doesn't try to hide that invariant really is a method?

On Sat, 10 Mar 2007 11:05:34 +0100, torhu wrote:

> So assert(obj) is a special case.
> 
> I think assert(obj) being the same as assert(obj !is null), and then being able to do 'obj.invariant' if you really want to explicitly check the invariant would a lot clearer.  Or even 'invariant(obj)'.

Yes. That would be a very good idea.

-- 
Derek Parnell
Melbourne, Australia
"Justice for David Hicks!"
skype: derek.j.parnell

"torhu" <fake@address.dude> wrote in message news:estvsu$2pd9$1@digitalmars.com...
> I think assert(obj) being the same as assert(obj !is null), and then being able to do 'obj.invariant' if you really want to explicitly check the invariant would a lot clearer.  Or even 'invariant(obj)'.

votes = votes + 1; // add one to votes

On Sat, 10 Mar 2007 12:43:22 -0500, "Jarrett Billingsley" <kb3ctd2@yahoo.com> wrote:

>"torhu" <fake@address.dude> wrote in message news:estvsu$2pd9$1@digitalmars.com...
>> I think assert(obj) being the same as assert(obj !is null), and then being able to do 'obj.invariant' if you really want to explicitly check the invariant would a lot clearer.  Or even 'invariant(obj)'.
>
>votes = votes + 1; // add one to votes
>
vote++ for obj.invariant