Re: Architecture question

Dear Douglas,

Thank you for your interest in MTL and your suggestions.

On Mar 6, 2006, at 9:19 PM, Douglas Aberdeen wrote:

> Dear MTL developers,
>
> I've recently come across MTL and I'm starting to use it. It seems to
> be the only c++ matrix package that handles sparse and dense
> matricies in a fairly uniform way, with basic linear algebra
> supported seemlessly between matrix formats. So thank you!
>
> There seem to be a couple of obvious things missing and I'm wondering
> if the architecture of MTL makes them bad, or if they are just things
> that the developers haven't gotten to?
>
> In short, I find it weird to call routines like
> 	set_value(A, val);
> 	print_vector(x);
> 	print_all_matrix(A);
> 	copy(A, B);
> It seems more intuitive (and notationally neater when using pointers
> to objects) to use
> 	A.set_value(val)

Member functions are less flexible than free functions in their  
interplay.  Other than that for given data types it is favorable to  
add a free function instead of writing a wrapper to introduce more  
member functions (and later a wrapper of the wrapper ... ).

> 	cout<<x;
> 	cout<<A;

Agreed.

> 	B =  A.copy()
>
What should be the type of 'A.copy()' ?  And where should it be  
stored?  Anyway, there will be B = A which deals with conversion and  
other issues.  Operator notation has often the problem of requiring  
temporary variables to store intermediate results.  We will deal with  
this in terms of expression templates (which in turn might induce  
compile time overhead).
> I understand the dangers of operator overloading in a matrix context,
> but things like stream operators should be okay?
>
> I'm also wondering whether development is still active since the last
> sign of activity was a gcc 4.0 compliant version more than 6 months
> ago?	

Yes, it is.  But as we completely redesigned MTL, it takes some time.

Best Regards

Peter
>
> --
> Dr Douglas Aberdeen
> Senior Researcher
> Statistical Machine Learning
>
>
> National ICT Australia Limited
> Locked Bag 8001
> Canberra ACT 2601
> Tel. +61 2 6125 8647
> Fax. +61 2 6125 8645
> Email. [email protected]
> Web. www.nicta.com.au
>
> The imagination driving Australia's ICT future.
>
> To receive the latest NICTA information register at http://
> nicta.com.au/registration.cfm
>
>
> _______________________________________________
> This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/

------------
Peter Gottschling
Research Associate
Open Systems Laboratory
Indiana University
301i Lindley Hall
Bloomington, IN 47405
Tel.: +1 812 855-8898   Fax: +1 812 856 0853
http://www.osl.iu.edu/~pgottsch


_______________________________________________
This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/