Cleanups in convolution operators

[stigrj]

This will give VAMPyR the ability to build ConvolutionOperator from arbitrary GaussExp<1>

• Remove unnecessary intermediate class GreensKernel
• ConvolutionOperator now accepts generic GaussExp<1> kernel constructor argument
• OperatorTreeVector = std::vector<OperatorTree *> -> std::vector<std::unique_ptr<OperatorTree>>
• MWOperator.band_max = Eigen::VectorXi -> std::vector<int>
• Remove kern_mra and oper_mra from MWOperator class, now computed on-the-fly
• Add MRA<D> member to MWOperator, where D is the dimension of the functions we are operating on.
• Moved getKernelMRA() and getOperatorMRA() from MRA class to MWOperator
• Moved calcMinDistance() and calcMaxDistance() from operator classes to MRA

[codecov[bot]]

Codecov Report

Merging #196 (c0c8582) into master (0e079a9) will decrease coverage by 0.68%. The diff coverage is 83.75%.

@@            Coverage Diff             @@
##           master     #196      +/-   ##
==========================================
- Coverage   63.86%   63.17%   -0.69%     
==========================================
  Files         172      174       +2     
  Lines       13056    13077      +21     
==========================================
- Hits         8338     8262      -76     
- Misses       4718     4815      +97     

Impacted Files	Coverage Δ
src/operators/ABGVOperator.h	100.00% <ø> (ø)
src/operators/BSOperator.h	100.00% <ø> (ø)
src/operators/DerivativeConvolution.cpp	0.00% <0.00%> (ø)
src/operators/DerivativeKernel.h	0.00% <0.00%> (ø)
src/operators/HelmholtzOperator.h	100.00% <ø> (ø)
src/operators/IdentityConvolution.h	100.00% <ø> (ø)
src/operators/PHOperator.h	100.00% <ø> (ø)
src/operators/PoissonOperator.cpp	43.47% <41.17%> (-56.53%)	:arrow_down:
src/operators/HelmholtzOperator.cpp	56.52% <57.89%> (-43.48%)	:arrow_down:
src/operators/MWOperator.h	80.00% <77.77%> (-20.00%)	:arrow_down:
... and 29 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 0e079a9...c0c8582. Read the comment docs.

[robertodr]

For OperatorTreeVector, does the 2nd CTOR of unique_ptr suffice? I recall wrapping smart pointers into STL containers as having some pitfalls...

[stigrj]

Which is the 2nd CTOR of unique_ptr?

[robertodr]

https://en.cppreference.com/w/cpp/memory/unique_ptr

template <
    class T,
    class Deleter
> class unique_ptr<T[], Deleter>;

[stigrj]

Sorry, I do not understand the problem

[stigrj]

You mean it's unsafe to do the following?

std::vector<std::unique_ptr<OperatorTree>> oper_vec;
auto o_tree = std::make_unique<OperatorTree>(mra, prec);
oper_vec.push_back(std::move(o_tree));

[robertodr]

You mean it's unsafe to do the following?

std::vector<std::unique_ptr<OperatorTree>> oper_vec;
auto o_tree = std::make_unique<OperatorTree>(mra, prec);
oper_vec.push_back(std::move(o_tree));

yes, or rather, there used to be a warning against using smart pointers like that. But some Googling shows it's pre-C++11 thing.

[stigrj]

I will update the doc string before merging