daubechies_scaling coverage.

[jzmaddock]

This one really needs @NAThompson 's help.

Any ideas on obtaining meaningful tests for:

math/include/boost/math/special_functions/daubechies_scaling.hpp

Line 366 in 9f03f29

constexpr std::array<int, 20> r{ -1, -1, 18, 19, 16, 11, 8, 7, 7, 7, 5, 5, 4, 4, 4, 4, 3, 3, 3, 3 };

and

math/include/boost/math/special_functions/daubechies_scaling.hpp

Line 473 in 9f03f29

return m_interpolator->double_prime(x);

In daubechies_scaling.hpp?

And then similarly:

math/include/boost/math/special_functions/daubechies_wavelet.hpp

Line 123 in 9f03f29

std::array<int, 20> r{ -1, -1, 18, 19, 16, 11, 8, 7, 7, 7, 5, 5, 4, 4, 4, 4, 3, 3, 3, 3 };

In daubechies_scaling.hpp.

[NAThompson]

@jzmaddock : Yes, unit tests for the derivative can be obtained just by doing finite differencing with the optimal spacing: https://github.com/boostorg/math/blob/develop/include/boost/math/differentiation/finite_difference.hpp

That being said, the accuracy really can degrade for both methods and picking a tolerance will be quite arbitrary. We can certainly demonstrate accuracy "in the eyeball norm", but I doubt in any other.

As to the arrays, I'm not convinced those can really be covered. They were selected rather arbitrarily by the conditions that 1) they didn't explode the memory use, and 2) they achieved a good accuracy under the constraint of 1).

My apologies-my brain is DDOS's until at least June!