Hi,

floats are stored on 32 bits using ieee754...and I would like (for some obscure reason) to reinterpret a such float into a 32 bits uint (i.e without altering the memory). A simple :

import std.stdio;

void main()
{
    float f = 0.5;
    uint i = cast(uint)(f);
    writeln(i);
}

doesn't work since it just round the float.

In C++, I do : reinterpret_cast<std::size_t&>(my_float). How could I do this in D ?

Thanks in advance for your help !

On Sunday, 29 March 2015 at 13:39:47 UTC, matovitch wrote:
> Hi,
>
> floats are stored on 32 bits using ieee754...and I would like (for some obscure reason) to reinterpret a such float into a 32 bits uint (i.e without altering the memory). A simple :
>
> import std.stdio;
>
> void main()
> {
>     float f = 0.5;
>     uint i = cast(uint)(f);
>     writeln(i);
> }
>
> doesn't work since it just round the float.
>
> In C++, I do : reinterpret_cast<std::size_t&>(my_float). How could I do this in D ?
>
> Thanks in advance for your help !

Ok this works :

import std.stdio;

void main()
{
    float f = 0.5;
    uint i = *cast(uint*)(&f);
    writeln(i);
}

It is kind of logical...I feel a bit dumb. :D

On Sun, 29 Mar 2015 13:45:10 +0000, matovitch wrote:

you can also use unions.

On Sunday, 29 March 2015 at 14:50:24 UTC, ketmar wrote:
> On Sun, 29 Mar 2015 13:45:10 +0000, matovitch wrote:
>
> you can also use unions.

Good idea ! In my case I think it was better to cast, but this could be helpful another time thanks ! :)

On Sun, 29 Mar 2015 16:00:05 +0000, matovitch wrote:

> On Sunday, 29 March 2015 at 14:50:24 UTC, ketmar wrote:
>> On Sun, 29 Mar 2015 13:45:10 +0000, matovitch wrote:
>>
>> you can also use unions.
> 
> Good idea ! In my case I think it was better to cast, but this could be helpful another time thanks ! :)

unions also looks better than pointers, and they are easier to read, i think. ;-)

union FU {
  float f;
  uint u;
}

void main () pure {
  float t = 42.0;
  assert((cast(FU)t).u == 0x42280000);
}

On Sunday, 29 March 2015 at 16:29:40 UTC, ketmar wrote:
> On Sun, 29 Mar 2015 16:00:05 +0000, matovitch wrote:
>
>> On Sunday, 29 March 2015 at 14:50:24 UTC, ketmar wrote:
>>> On Sun, 29 Mar 2015 13:45:10 +0000, matovitch wrote:
>>>
>>> you can also use unions.
>> 
>> Good idea ! In my case I think it was better to cast, but this could be
>> helpful another time thanks ! :)
>
> unions also looks better than pointers, and they are easier to read, i
> think. ;-)
>
> union FU {
>   float f;
>   uint u;
> }
>
> void main () pure {
>   float t = 42.0;
>   assert((cast(FU)t).u == 0x42280000);
> }

AFAIK this would be undefined behaviour in C++, right?

On Sun, 29 Mar 2015 17:33:02 +0000, Namespace wrote:

> On Sunday, 29 March 2015 at 16:29:40 UTC, ketmar wrote:
>> On Sun, 29 Mar 2015 16:00:05 +0000, matovitch wrote:
>>
>>> On Sunday, 29 March 2015 at 14:50:24 UTC, ketmar wrote:
>>>> On Sun, 29 Mar 2015 13:45:10 +0000, matovitch wrote:
>>>>
>>>> you can also use unions.
>>> 
>>> Good idea ! In my case I think it was better to cast, but this could be helpful another time thanks ! :)
>>
>> unions also looks better than pointers, and they are easier to read, i think. ;-)
>>
>> union FU {
>>   float f;
>>   uint u;
>> }
>>
>> void main () pure {
>>   float t = 42.0; assert((cast(FU)t).u == 0x42280000);
>> }
> 
> AFAIK this would be undefined behaviour in C++, right?

i honestly don't remember. C++ itself is UB. ;-)