Weighted incidence

[PuneethaPai]

Fixes Issue: #229

I have handled following cases w.r.t permutations of params:

• Allow empty str as weighted attr name
• Raise ValueError when multiple and weighted can not co-exist.
• weighted=True graph becomes weighted with edge_attr=weight
• weighted="some_str" graph is un-weighted with edge_attr=some_str
• directed=True, mode="out", weighted=True graph is weighted, direction out is respcted
• directed=True, mode="in", weighted=True graph is weighted, direction in is respcted
• directed=True, mode="all", weighted=True graph is weighted, direction all is respcted

[PuneethaPai]

Hi @ntamas ,
In test code I have used:

• all(list([bool])) instead of bool and bool and ...
• I have also used self.assertListEquals(list1, list2) instead of self.assertTrue(list1 == list2)

I did not modify other parts of code.

Now to maintain homogeneity:

• If you liked this I can apply the same to other parts of code.
• If you do not like this I can change it back to how you all are following.

[PuneethaPai]

I think i haven't handled negative weight scenario. I will refer to behavior w.r.t R implementation and accordingly handle in python.
Let me know if you have any thoughts how to tackle negative weight.
Thanks

[PuneethaPai]

@ntamas
Just checked with R-igraph. It allows negative values to the weights.
So I think we are good. I was in assumption that we would not allow negative weights.
You can merge this. Let me know if there are any cases I have missed or any improvements in code.

Thanks

[ntamas]

Thanks a lot for your contribution! I have requested a few minor changes, but otherwise it seems good to go. I'll merge this when the requests are addressed.

[PuneethaPai]

@ntamas why is wheel build is not happening in win32 x86 system? How to debug this issue?

[ntamas]

This is probably an issue with the C core and is unrelated to your PR. I'll merge this and then fix the failing build in a separate commit. Thanks for your contribution!