This is the property example from http://digitalmars.com/d/1.0/property.html

int a;
int b = 1;
typedef int t = 2;
t c;
t d = cast(t)3;

int.init // is 0
a.init  // is 0
b.init  // is 1 // Well, I get 0
t.init  // is 2 // Well, I get 2
c.init  // is 2 // Well, I get 2
d.init  // is 3 // Well, I get 2

struct Foo
{
    int a;
    int b = 7;
}

Foo.a.init // is 0 // Well, I get 0
Foo.b.init // is 7 // Well, I get 0


And another:

enum E { V };
E.V.stringof; // I get int instead of E.V


And another:

float f = 1.0;
f.init; // This should be 1.0 according to the spec.. It's nan here.
const f2 = 1.0;
f2.init; // still nan... Only typedefs changes the init?

Ditto for other types.


And is it possible to check if a float is nan?
float f2;
f2 is float.nan; // false
f2.init is float.init; // false
float.nan is float.nan; // false...


I'm using dmd 1.028

"Simen Haugen" <simen.haugen@pandavre.com> wrote in message news:g2s4uk$2nt7$1@digitalmars.com...
> This is the property example from http://digitalmars.com/d/1.0/property.html
>
> ...
>
> Ditto for other types.

The behavior of .init was changed in 1.017.  Per-variable .inits now just get the .init of the type.  I don't know why the spec hasn't been updated to match.

>
> And is it possible to check if a float is nan?
> float f2;
> f2 is float.nan; // false
> f2.init is float.init; // false
> float.nan is float.nan; // false...

Either use std.math.isnan/tango.math.IEEE.isNaN, or a somewhat.. funnier looking method:

float f2;
f2 !<>= f2; // true

"Jarrett Billingsley" wrote in message
> Either use std.math.isnan/tango.math.IEEE.isNaN, or a somewhat.. funnier looking method:
>
> float f2;
> f2 !<>= f2; // true

Thanks a lot

If I had checked the expression chapter I would have seen that. I must say that at first I thought wtf! I had no idea what you were doing there: "f2 not lt gt eq f2"?!? Sure is a long strange looking operator.

"Simen Haugen" <simen@norstat.no> wrote in message news:g2tdtv$e76$1@digitalmars.com...
> "Jarrett Billingsley" wrote in message
>> Either use std.math.isnan/tango.math.IEEE.isNaN, or a somewhat.. funnier looking method:
>>
>> float f2;
>> f2 !<>= f2; // true
>
> Thanks a lot
>
> If I had checked the expression chapter I would have seen that. I must say that at first I thought wtf! I had no idea what you were doing there: "f2 not lt gt eq f2"?!? Sure is a long strange looking operator.

Heheh.  Well I don't think it actually mentions that that's a possible test for nan, but it just happens to work out because of the semantics of that operator.

Reply to Jarrett,

> "Simen Haugen" <simen@norstat.no> wrote in message
> news:g2tdtv$e76$1@digitalmars.com...
> 
>> "Jarrett Billingsley" wrote in message
>> 
>>> Either use std.math.isnan/tango.math.IEEE.isNaN, or a somewhat..
>>> funnier looking method:
>>> 
>>> float f2;
>>> f2 !<>= f2; // true
>> Thanks a lot
>> 
>> If I had checked the expression chapter I would have seen that. I
>> must say that at first I thought wtf! I had no idea what you were
>> doing there: "f2 not lt gt eq f2"?!? Sure is a long strange looking
>> operator.
>> 
> Heheh.  Well I don't think it actually mentions that that's a possible
> test for nan, but it just happens to work out because of the semantics
> of that operator.
> 

It's also the only thing it's useful for.

However I would like to see "f1 is f2" do an "incorrect" floating point comparison: (NaN is NaN) == true

I was planning on asking for it but Simen beat me to it.

On Thu, 12 Jun 2008 23:37:23 +0200, "Simen Haugen" <simen.haugen@pandavre.com> wrote:

>This is the property example from http://digitalmars.com/d/1.0/property.html
>
>int a;
>int b = 1;
>typedef int t = 2;
>t c;
>t d = cast(t)3;
>
>int.init // is 0
>a.init  // is 0
>b.init  // is 1 // Well, I get 0
>t.init  // is 2 // Well, I get 2
>c.init  // is 2 // Well, I get 2
>d.init  // is 3 // Well, I get 2
>
>struct Foo
>{
>    int a;
>    int b = 7;
>}
>
>Foo.a.init // is 0 // Well, I get 0
>Foo.b.init // is 7 // Well, I get 0
> ...

The .init error in spec is already in bugzilla. http://d.puremagic.com/issues/show_bug.cgi?id=2045

Gide