STYLE: Replace std::pow(x, 2) calls with Math::sqr(x)

[N-Dekker]

Math::sqr(x) appears slightly more readable than std::pow(x, 2).

[issakomi]

BTW, sqr sometimes means square root, e.g. in VBA.

Edit:
In e.g. Delphi sqr is square like in VNL. There is some ambiguity.

[seanm]

I had actually started writing a comment wondering why we saved 1 letter and didn't just call it sqrt like C/C++ do, then I realized it's not square root just before clicking "Comment".

Renaming it square might be wise...

[N-Dekker]

Thanks @issakomi and @seanm I don't like the name sqr either. But it's taken directly from the vxl library:

ITK/Modules/Core/Common/include/itkMath.h

Line 822 in 018e0ea

using vnl_math::sqr;

itk::Math::sqr was added to ITK almost nine years ago, with commit 2133446, from 2016.

If it would be implemented in ITK instead, I would prefer it to be named Square, following ITK's "Pascal Case" naming convention.

Anyway, my aim for this PR is to avoid calling std::pow unnecessarily, as it appears somewhat troublesome, as reported by issue #5245

I feel that a new square function for ITK is beyond the scope of this PR. (The new square function might be constexpr, it might be written as a function template, it would need extra unit tests, deprecating the current sqr, considering what to do about Math::cube, etc.) Do you agree that it would be an acceptable improvement to at least replace std::pow(x, 2) calls with the old Math::sqr(x), and then possibly later introduce a new square function?

[seanm]

This one is arguably better left as it was, to be more similar to the next 2 lines... but I don't feel strongly about it.

[N-Dekker]

I see your point, thanks, but on the other hand, the next line (sum3 += std::pow(static_cast<double>(v), 3)) might as well use Math::cube instead 😸

[N-Dekker]

As a follow-up, it might be nice to replace Math::sqr(pointVector.GetNorm() with pointVector.GetSquaredNorm(), but I guess doing so might introduce regression errors, because it would potentially reduce rounding errors!

[N-Dekker]

Note that distanceFromCenter is in fact the dot product of two vectors:

ITK/Modules/Core/Common/include/itkFiniteCylinderSpatialFunction.hxx

Lines 96 to 97 in e9af5c5

	const double distanceFromCenter = dot_product(medialAxisVector.GetVnlVector(), pointVector.GetVnlVector());

So I'm just wondering if the square of distanceFromCenter was originally intended. (But I will leave it "squared" like this, because I don't know the original intention.)

[dzenanz]

This is a step in the right direction. If someone has time, this might be improved further.