Travis CI failing about scona HOT 9 CLOSED

This appears to be a networkx 1 to 2 migration issue. The migration guide seems to recommend using dict(G.degree()) to get the degrees as a dictionary. This should be backwards compatible with networkx 1

from scona.

Do you want to require networkx version 1.9 to fix the Travis tests first? Or are you planning on migrating to version 2.0 to fix things?

from scona.

There were a few people at Mozfest who had the AttributeError: 'DegreeView' object has no attribute 'values' error that interrupts builds of python 3.3 and later. I'm using 3.6 but I don't get it.
Verbatim the error is:

Traceback (most recent call last):
  File "/home/travis/build/WhitakerLab/BrainNetworksInPython/tests/regression_tests.py", line 17, in setUpClass
    write_fixtures('/tmp')
  File "/home/travis/build/WhitakerLab/BrainNetworksInPython/tests/write_fixtures.py", line 80, in write_fixtures
    recreate_network_analysis_fixture(folder, corrmat_path)
  File "/home/travis/build/WhitakerLab/BrainNetworksInPython/tests/write_fixtures.py", line 49, in recreate_network_analysis_fixture
    names_308_style=True)
  File "/home/travis/build/WhitakerLab/BrainNetworksInPython/wrappers/network_analysis_from_corrmat.py", line 229, in network_analysis_from_corrmat
    names_308_style=names_308_style)
  File "/home/travis/build/WhitakerLab/BrainNetworksInPython/scripts/make_graphs.py", line 383, in calculate_nodal_measures
    deg = G.degree().values()
AttributeError: 'DegreeView' object has no attribute 'values'

At this point in the code, G.degree() should be a dict object. G.degree is a DegreeView object.

from scona.

Great find @Islast (and our fellow 🐛 finders at MozFest!)

from scona.

Good question @schae234 ! I think we will definitely want to migrate to networkx version 2.0 sometime. But for now I we should probably change Travis to use version 1.9. If our Travis builds pass after that we can close this issue and create a new "networkx verison 2.0 compatability issue"

from scona.

Hi @Islast , that sounds like a good plan! I have a fork thats passing with python version 3.6, but the regression tests fail for other versions of python. I'm still looking into it! 👍

Do you know if there is a way to get the tests to be more verbose? I'm not familiar with the unittest package

from scona.

Great job @schae23 ! I believe you use a -v to get "verbose" detail, as in:

 python3 -m unittest discover -v -p '*_tests.py

in the .travis.yml

from scona.

Some thoughts about why the non-3.6 tests might be failing.

• The regression tests compare to files that were generated (using a fixed random seed) in python 3.6
• I notice that the corrmat tests don't fail - this is the only file that is created without using networkx
• The global and rich club measures both use the random seed, and these tests both fail, but the local measures do not use the random seed and they fail too

@schae234 if you feel like keeping going on this please do! If not, the work you've already done is extremely helpful 🌺
Feel free to PR whenever

from scona.

Travis builds are now passing as of pull request #49

from scona.