Fix test sg03ad

[bnavigator]

This test fails on some architectures while building for openSUSE. Replacing range() which returns an int generator with arange() fixes this.

Plus some whitespace (pep8) fixes.

[bnavigator]

Hmm, the build service now reports failure again even with the patch applied. Seems there are more issues with powerpc left.

[  429s] ======================================================================
[  429s] FAIL: test_sg03ad_a (slycot.tests.test_sg03ad.test_sg03ad)
[  429s] ----------------------------------------------------------------------
[  429s] Traceback (most recent call last):
[  429s]   File "/home/abuild/rpmbuild/BUILDROOT/python-slycot-0.3.5.0-18.1.ppc64/usr/lib64/python3.7/site-packages/slycot/tests/test_sg03ad.py", line 32, in test_sg03ad_a
[  429s]     assert_almost_equal(Xref, X)
[  429s]   File "/usr/lib64/python3.7/site-packages/numpy/testing/_private/utils.py", line 584, in assert_almost_equal
[  429s]     return assert_array_almost_equal(actual, desired, decimal, err_msg)
[  429s]   File "/usr/lib64/python3.7/site-packages/numpy/testing/_private/utils.py", line 1029, in assert_array_almost_equal
[  429s]     precision=decimal)
[  429s]   File "/usr/lib64/python3.7/site-packages/numpy/testing/_private/utils.py", line 841, in assert_array_compare
[  429s]     raise AssertionError(msg)
[  429s] AssertionError: 
[  429s] Arrays are not almost equal to 7 decimals
[  429s] 
[  429s] Mismatch: 58.6%
[  429s] Max absolute difference: 0.65333008
[  429s] Max relative difference: 1.31888793
[  429s]  x: array([[1., 1., 1., ..., 1., 1., 1.],
[  429s]        [1., 1., 1., ..., 1., 1., 1.],
[  429s]        [1., 1., 1., ..., 1., 1., 1.],...
[  429s]  y: array([[1.       , 1.       , 1.       , ..., 1.6032463, 1.5698183,
[  429s]         1.5585347],
[  429s]        [1.       , 1.       , 1.       , ..., 1.5900677, 1.557722 ,...
[  429s] 

[bnavigator]

Back in business! The calculation of matrix A did not fit the definition of the paper. I added the discrete time case from the paper as well.

[roryyorke]

Just for my info: how important is the range to np.arange change? It looks like np.diag(range(n)) works OK.

I'm going to try to push a rebase onto lastest master.

[bnavigator]

Just for my info: how important is the range to np.arange change? It looks like np.diag(range(n)) works OK.

I think I did that to avoid having an array of integers. Some Python version back then introduced precision problems because of the implicit conversion. Not sure anymore if it is really needed. Numpy on Python 3.8 however frequently produce DeprecationWarnings when an array needs to be casted to float implicitly. So better keep the np.arange?

[roryyorke]

On range/arange: I was just curious.

I can't push to this branch - are you willing (and do you have the know-how---I don't think I do!) to make it editable by other Slycot devs? I want to push the rebase (see https://github.com/roryyorke/Slycot/tree/fix_test_sg03ad_static)

[roryyorke]

Merged via 9d2bfe3

[bnavigator]

This PR was missing the

• Allow edits from maintainers.

checkbox. All the other remaining PR's do have it.

[roryyorke]

FWIW I just tried pushing a rebased branch for #83, it also failed. When I try to push it asks for a username (assume it tries my normal SSH credentials first, and fails).

[bnavigator]

Did you add the remote as https:// or git@github...?

[roryyorke]

Ah-ha, thank you! Had it as https. Now pushed.

I'll be offline for an hour or two now.