Fixed __eq__ issue

[Hadimius]

Changed eq so it accounts for differences in frequency and all primary properties. Refers to and solves issue #914

[jhillairet]

Great contribution, thank you!

[Ttl]

Comparing all PRIMARY_PROPERTIES is unnecessary. Network object stores only S-parameters and other parameters are converted from them as necessary. It's enough to compare f, s, s_def and z0. s_def comparison is only necessary if any of z0 has non-zero imaginary component. Different network with the same real z0 and s are equal even if s_def is different. Also noise parameters should be compared if they exist.

numpy will raise error in future when comparing two different length arrays by equality. It would be better to first compare the number of frequency points. Tests would be nice to have to check all these different cases.

[jhillairet]

I agree that comparing all properties is maybe overkill. Is the more the better?

Or should we really restrict the comparisons to:

• f
• s
• z0
• s_def (if any of z0 has non-zero imaginary component)

for noise parameter, eventually too.

[FranzForstmayr]

The abs is at the wrong place.
Maybe use something like return npy.allclose(self.z0.real, other.z0.real, atol=ZERO)

[FranzForstmayr]

Maybe one more test with f2 = f1 * 10, as this test returns when checking the length of the vectors.

[Hadimius]

It seams to be failing because of the following test in test_media:

    def test_impedance_mismatch(self):
        """
        Test the naming of the network. When circuit is used to connect a
        topology of networks, they should have unique names.
        """
        name = 'impedanceMismatch,50to25'
        qucs_ntwk = Network(os.path.join(self.files_dir, name + '.s2p'))
        self.dummy_media.frequency = qucs_ntwk.frequency
        skrf_ntwk = self.dummy_media.thru(z0=50, name = name)**\
            self.dummy_media.thru(z0=25)

        self.assertEqual(qucs_ntwk, skrf_ntwk)
        self.assertEqual(qucs_ntwk.name, skrf_ntwk.name)

I'm not 100% sure about this, but the Networks in this test should not be equal, since they differ in z0, or?
So should the test be adapted to compare only f, s and name?

[FranzForstmayr]

I think the problem is, that the file stored in sNp has always the same port impedance on all ports, in this case 50.
You cannot store 50/25 Ohm in sNp, so this test has to fail by definition.

You could manually set qucs_ntwk.z0 to [50, 25] to avoid this issue. If you do, please write a small note :)

[FranzForstmayr]

@Hadimius don't worry about CI, this issue is from the latest numpy release from yesterday. We need to fix this ASAP :)

[FranzForstmayr]

Can you merge in master please?

[FranzForstmayr]

Looks good!
Thank you @Hadimius for your effort!