Blitz++ Library

There are only two operator= for blitz::Array, which

a) Copy Array of the same type, or
b) assign the result of an expression.

So you can use b) via

  blitz::Array< float, 2 > x;
  blitz::Array< int, 2 > y;
  x = y(blitz::tensor::i, blitz::tensor::j);

or 

  blitz::Array< float, 2 > x;
  blitz::Array< int, 2 > y;
  blitz::firstIndex i;  blitz::secondIndex j;
  x = y(i,j);

Arguably the type conversion operator= is missing. On the other hand,
making the type conversion more explicit might save you from
accidentally losing precision.

Volker


On Fri, 2008-01-04 at 18:41 +0100, Johannes Ballé wrote: 
> Dear Blitz++ gurus,
> 
> I've been getting strange compilation errors involving implicit type casts 
> with Blitz++ 0.9. It boils down to the following problem:
> 
> #include <blitz/array.h>
> 
> int main( int argc, char *argv[] ) {
>   blitz::Array< float, 2 > x;
>   blitz::Array< int, 2 > y;
>   x = y;
>   return 0;
> }
> 
> This code doesn't compile with g++, although I think it should be able to cast 
> each int element to a float element ...
> 
> Maybe some compiler issue? Or am I just blind? ;)

Dear Volker,

thanks for your help. I tried your first suggestion and it works!

About the design: My intuition suggests that an array _is_ a (trivial) arra=
y=20
expression. That is probably why I couldn't find the solution by myself. I=
=20
just expected the problem to be somewhere else.

=46rom my point of view, a type casting operator=3D would be logical. IMHO,=
 the=20
code you suggested looks a bit clumsy, considering it implements just a=20
simple array cast. However, there might be reasons against it, as you said.

Previously, I have also tried

x =3D blitz::cast( y, float() );

(to no avail, of course) but now I start to realize that this is probably n=
ot=20
the way the blitz::cast function was meant to be used.

Thanks,

Johannes.

On Saturday 05 January 2008 16:39:16 you wrote:
> There are only two operator=3D for blitz::Array, which
>
> a) Copy Array of the same type, or
> b) assign the result of an expression.
>
> So you can use b) via
>
>   blitz::Array< float, 2 > x;
>   blitz::Array< int, 2 > y;
>   x =3D y(blitz::tensor::i, blitz::tensor::j);
>
> or
>
>   blitz::Array< float, 2 > x;
>   blitz::Array< int, 2 > y;
>   blitz::firstIndex i;  blitz::secondIndex j;
>   x =3D y(i,j);
>
> Arguably the type conversion operator=3D is missing. On the other hand,
> making the type conversion more explicit might save you from
> accidentally losing precision.
>
> Volker
>
> On Fri, 2008-01-04 at 18:41 +0100, Johannes Ball=C3=A9 wrote:
> > Dear Blitz++ gurus,
> >
> > I've been getting strange compilation errors involving implicit type
> > casts with Blitz++ 0.9. It boils down to the following problem:
> >
> > #include <blitz/array.h>
> >
> > int main( int argc, char *argv[] ) {
> >   blitz::Array< float, 2 > x;
> >   blitz::Array< int, 2 > y;
> >   x =3D y;
> >   return 0;
> > }
> >
> > This code doesn't compile with g++, although I think it should be able =
to
> > cast each int element to a float element ...
> >
> > Maybe some compiler issue? Or am I just blind? ;)

=2D-=20
Johannes Ball=C3=A9
Institut f=C3=BCr Nachrichtentechnik, RWTH Aachen

phone: +49 241 80-27678
jabber: balle.ient@...

http://www.ient.rwth-aachen.de/

Hi Johannes,

Apparently you prefer this, which works as well:

  blitz::Array< float, 2 > x;
  blitz::Array< int, 2 > y;
  x = blitz::cast<float>(y);

I think the documentation of the cast function needs to be updated,
though. 3.7 says this is is not yet supported :-)

Volker


On Sun, 2008-01-06 at 15:16 +0100, Johannes Ballé wrote:
> x = blitz::cast( y, float() );

Ah yes!

That was what I was looking for. Strange then, that

x =3D blitz::cast( y, float() );

doesn't compile. I thought if this doesn't work in the first place, then th=
e=20
newer syntax won't work either :(

Regards,

Johannes.

On Sunday 06 January 2008 17:07:34 Volker Braun wrote:
> Hi Johannes,
>
> Apparently you prefer this, which works as well:
>
>   blitz::Array< float, 2 > x;
>   blitz::Array< int, 2 > y;
>   x =3D blitz::cast<float>(y);
>
> I think the documentation of the cast function needs to be updated,
> though. 3.7 says this is is not yet supported :-)
>
> Volker
>
> On Sun, 2008-01-06 at 15:16 +0100, Johannes Ball=C3=A9 wrote:
> > x =3D blitz::cast( y, float() );

=2D-=20
Johannes Ball=C3=A9
Institut f=C3=BCr Nachrichtentechnik, RWTH Aachen

phone: +49 241 80-27678
mobile: +49 176 64189343
jabber: balle.ient@...

http://www.ient.rwth-aachen.de/