Affine transforms - matrix algebra

[Kirill Prazdnikov]

Hi Martin,

I perfectly understand the value of indexed access.
And I agree with you.

Thanks
   -Kirill

On 12-Jul-12 16:20, Martin Desruisseaux wrote:
> Le 12/07/12 14:08, Kirill Prazdnikov a écrit :
>> Transformation matrix consist of four vectors. In case of affine, it 
>> is coordinates of local orts + translation.
>> Accessing this vectors is the most get/set use-case.
>> I can not imagine where accessing individual matrix values could be 
>> used (useful) at all.
>>
>> Does anybody uses it ?
>
> Well, then intend is to access the vectors by looping over 
> (row,column) indices. In the current API, we can only access 
> individual matrix values through named properties. This means that we 
> can not use loops neither parametrized methods. For example in order 
> to compute the magnitude of a vector (my previous email about "square 
> root of the sum of ..." was a complicated way to said exactly that), 
> we have to:
>
>  * explicitly sum every terms as in sqrt(mxx*mxx + mxy*mxy + mxz*mxz);
>  * duplicate the above code for every rows.
>
>
> While I admit that a loop doesn't save much in this example, some 
> other use case are a bit more complicated. And the problem of 
> duplicated code for every vector (row) remain. If a 'get(row, column)' 
> method is provided, then a user can compute a vector magnitude in a 
> single method where the vector number (row index) is given as a method 
> argument.
>
>     Martin
>