From Operator to NumPy operator

[pmli]

See #234

[coveralls]

Coverage decreased (-0.07%) to 33.542% when pulling 84ade54 on pmli:to_numpy_operator into 8500dc2 on pymor:master.

[coveralls]

Coverage decreased (-0.09%) to 33.528% when pulling dd44796 on pmli:to_numpy_operator into 8500dc2 on pymor:master.

[sdrave]

Thanks for the nice work! I only have two remarks:

1. The suite for LincombOperator will not work when the operator is parametric, in which case you would need to call op.evaluate_coefficients(mu) to obtain the linear coefficients. I would say one should at least check that all coefficients are numbers and not ParameterFunctionals. Even nicer would be to add an optional mu parameter to _to_matrix and pass it along.
2. I would prefer to leave out the to_numpy_operator function and rename _to_matrix to to_matrix. This would make it clearer that by using this function you leave the realm of pyMOR's abstractions and directly work on the data. - If you really need the matrix again as an Operator, it should not be too much of a pain to explicitly instantiate the NumpyMatrixOperator. Hoever, others (@ftalbrecht, @andreasbuhr) might disagree? Do you use to_numpy_operator yourself very often?

[pmli]

I actually noticed that I kept using ._matrix after to_numpy_operator, so I agree about renaming it.

[andreasbuhr]

Hi Petar,

the code looks good! :-)

I agree: The name to_matrix is more clear.

[ftalbrecht]

Thx for your work! I agree with Stephan that working with data structures directly (as opposed to operators and vector arrays) means leaving the (current) realm of pyMOR abstractions, and the payment for return shall be the extra instantiation of a NumpyMatrixOperator :). I would thus also vote for the to_matrix variant (at least for the time being).

[coveralls]

Coverage increased (+47.3%) to 80.947% when pulling 070aa56 on pmli:to_numpy_operator into 8500dc2 on pymor:master.

[sdrave]

Do you really mean sparray? Never heard of that before. Also, |foo| is a subtitution in reST (http://www.sphinx-doc.org/en/stable/rest.html#substitutions). We define all our substitions in docs/source/substitutions.py.

[pmli]

Whoops! Thanks for noticing!

[sdrave]

Apart from the two minor remarks above, I think this is ready to get merged?

[coveralls]

Coverage increased (+43.2%) to 76.776% when pulling 43cc916 on pmli:to_numpy_operator into 8500dc2 on pymor:master.

[coveralls]

Coverage decreased (-0.03%) to 80.983% when pulling 1b1542d on pmli:to_numpy_operator into 0ae424a on pymor:master.

[sdrave]

FYI, I have updated AUTHORS.md and added

algorithms.to_matrix

to your contributions list. Thanks again for your work!