COMP: Copy BSplineInterpolationWeightFunction::SupportSize within lambda

[N-Dekker]

Copied the constexpr static data member SupportSize to a local
variable, to prevent Clang linker errors, like:

Undefined symbols for architecture x86_64:
"itk::BSplineInterpolationWeightFunction::SupportSize", referenced from:
itk::BSplineInterpolationWeightFunction::'lambda'()::operator()() const in itkBSplineInterpolationWeightFunctionTest.cxx.o

From Mac10.13-AppleClang-dbg-x86_64-static https://open.cdash.org/viewBuildError.php?type=0&buildid=7457677, reported by Sean McBride (@seanm).

[Leengit]

@N-Dekker thank you for your perseverance on this issue! #2721

You know me ... I push to avoid auto and to use {} constructor syntax instead of assignment syntax, so I'd write

const SizeType supportSize{ SupportSize };

... but you can safely ignore me without even responding to this part.

[seanm]

I'm not a big fan of auto either. I find it makes code faster/easier to write, but harder to read, and readability is more important in my books.

Anyway, I'll test if this patch fixes things.

[N-Dekker]

I push to avoid auto and to use {} constructor syntax instead of assignment syntax, so I'd write

const SizeType supportSize{ SupportSize };

Thanks for your comment @Leengit To me, the syntax TypeX variableX{ value } suggests that a conversion takes place from the type of value to TypeX. Or that value is copied into the very first data member of an aggregate of type TypeX. I would typically use this syntax when that's my intention.

On the other hand, the syntax auto variableX = value communicates most clearly to me that both variableX and value have the same type, and that variableX will be an identical copy of value.

Hope that clarifies my motivation 😃

[seanm]

I confirm it fixes the link error.

[Leengit]

Thanks for your comment @Leengit To me, the syntax TypeX variableX{ value } suggests that a conversion takes place from the type of value to TypeX.

You responded, so that's on you. :-) C++ is more restrictive with types conversions with {} syntax than with = syntax.

Or that value is copied into the very first data member of an aggregate of type TypeX. I would typically use this syntax when that's my intention.

Yes, it looks like that. But I am thinking that to actually achieve that you'd have to use { {SupportSize} } or maybe ={SupportSize}.

On the other hand, the syntax auto variableX = value communicates most clearly to me that both variableX and value have the same type, and that variableX will be an identical copy of value.

Yes, I agree that auto communicates "same type" very clearly.

[Leengit]

Looks good -- thanks!

[N-Dekker]

@hjmjohnson @dzenanz @Leengit Thanks for your approvals!