Re: [CHANGESET]: First attempt at a single precision type.

[David Bateman]

John W. Eaton wrote:
> On 30-Apr-2008, David Bateman wrote:
> 
> | John W. Eaton wrote:
> | > On 30-Apr-2008, John W. Eaton wrote:
> | > 
> | > | Did you mean to write this as
> | > | 
> | > |   ComplexMatrix ComplexMatrix::stack (const Matrix&) const
> | > |   {
> | > |     return MArray2<ComplexMatrix>::stack (a);
> | > |   }
> | > | 
> | > | in the ComplexMatrix class?
> | > 
> | > Oops, I meant "MArray2<complex>::stack (a)".
> | 
> | Or even "MArray2<Complex>::stack(a)", sorry the fault was only in the
> | e-mail not in what I implemented.. Thinking about the problem is
> | probably that the template needs to be instantiated somewhere, though I
> | thought the call to the templated method should have done that itself..
> | I think I see in any case why these functions haven't been templated in
> | the past :-)
> 
> OK, it will work with the following function in MArray2.h:
> 
>   template <class U>
>   MArray2<T> stack (const MArray2<U>&) const
>   {
>     ...
>   }
> 
> and the following in CMatrix.cc:
> 
>   ComplexMatrix
>   ComplexMatrix::stack (const Matrix& a) const
>   {
>     return MArray2<Complex>::stack (a);
>   }

THis is what I had, but in CMatrix.h

> Of course it would be better to put this in CMatrix.h so it could be
> inlined.  To do that, you'll need to include Matrix.h in CMatrix.h.
> Without that, it fails because Matrix is not a complete type at the
> time CMatrix.h is parsed (I think).  The error message from g++ didn't
> really help me zero in on the problem immediately...

The error message didn't even reference Matrix!!! Ok, it seems to work.

> Something else to consider is that these are binary operations, so
> maybe they should't be member functions?  These are some of the older
> function in liboctave which were written when I didn't really know
> what I was doing (as if I do now).

Or maybe they shouldn't exist at all... Where are stack, extract,
extract_n actually used in Octave? If it is used could the call be
replaced will something else? As far as I can see extract_n and stack
aren't used and extract is only used in the pseudo_inverse method. Can
we just drop these methods or is there code external to Octave that
relies on them? This stuff seems like cruft to me..

D.