[Insight-developers] itk vs vnl / Vector & Matrix

[Luis Ibanez]

Robert Tamburo wrote:

>Okie dokie, finding the 1). smallest value in a vector and 2). the average
>of values in a vector would be useful for me.
>

Vector, Point and CovariantVector are intended for representing spatial 
coordinates.
Unfortunately "vector" seems to have too many connotations      :-/

itkVector  is not really intended to be used as the MatLab generalized 
vector...
but rather as a mathematical type for representing the tangent of a 
curve or a speed.
(in which case averaging x,y,z components could not make much sense)

We could do that at the Array level but not adding by a method to the array
it rather be implemented by additional utility classes that access the 
itk::Array.
This is because the elements of the array could be anything, not necesarily
arithmetic types. They could be <std::list<CoreAtoms>>   for example   :-)

A member method relaying on the array elements to provide operator+()
and operator/() will impose too many limitations.

For something like "min" and "average" it could make more sense to use a 
std::vector<>
and take advantage of all the STL algorithms. You can get access to them 
by using the
itkVectorContainer because it derives from a std::vector and wraps 
std::iterators which
is what most STL algorithms expect as input.


What particular use case are you considering for the "min" and "average"   ?


   Luis