Port topology module

[utkbansal]

Fixes #

Changes made in this Pull Request:

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

[utkbansal]

This should hopefully get coverage back to normal.

[utkbansal]

@richardjgowers @kain88-de Minor increase in coverage! A review would be helpful.

[kain88-de]

@jbarnoud & @mnmelo can you have a look I don't have time currently.

[jbarnoud]

Style is much better than in the PR on utils. Just one file where the imports are reordered.

I did not notice any tests that are missing, but I figured out why I had test that appeared discovered in pytest but not in nose: while we generally avoided docstring in tests to have a readable log in nose, we still have some. Nose does display the docstring instead of the test name in the verbose view; pytest doesn't. It means that it is difficult to accurately compare the two logs, but also that we will be able to have docstrings in tests. Yeah!

[jbarnoud]

Please, fix the import order.

[jbarnoud]

This all file will be much better with fixtures. Not for this PR, but this test suite will really profits from the move to pytest.

[utkbansal]

@jbarnoud @richardjgowers This should be ready to merge now.

[utkbansal]

Apparently, the hostname "dropbox.com" cannot be resolved during the full build.

[kain88-de]

Apparently, the hostname "dropbox.com" cannot be resolved during the full build.

We need more information. This message alone doesn't give me much when the build is restarted. When did it try to access dropbox would be valuable information for example.

[jbarnoud]

On 06/23/2017 04:48 PM, Max Linke wrote: Apparently, the hostname "dropbox.com" cannot be resolved during the full build. We need more information. This message alone doesn't give me much when the build is restarted. When did it try to access dropbox would be valuable information for example. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1420 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABUWuiSOGsbEOOPvKDxoUFXyDX9YTqemks5sG9BCgaJpZM4OBZLH>.

I did restart the job. Hole is downloaded from dropbox apparently. This time, dropbox.com could not be resolved. It seams to be just a transient error, the install phase is done now which means that dropox was found.

[utkbansal]

if [[ $NAME == 'full' ]]; then \
      bash ./maintainer/install_hole.sh $TRAVIS_OS_NAME "${HOME}/${MDA_OPTPACKAGES}"; \
      HOLE_BINDIR="${HOME}/${MDA_OPTPACKAGES}/hole2/exe"; \
      export PATH=${PATH}:${HOLE_BINDIR}; \
  fi
  
Downloading hole2-NotForProfit-2.2.004-Linux-x86_64.tar.gz from https://www.dropbox.com/s/jukpwlohhi20r17/hole2-NotForProfit-2.2.004-Linux-x86_64.tar.gz?dl=1...
curl: (6) Couldn't resolve host 'www.dropbox.com'
[install_hole.sh] ERROR: Failed to download hole2-NotForProfit-2.2.004-Linux-x86_64.tar.gz from https://www.dropbox.com/s/jukpwlohhi20r17/hole2-NotForProfit-2.2.004-Linux-x86_64.tar.gz?dl=1 [1]

It tries to download some hole dependency from dropbox and that is a part of the ci-helpers we are using.

[jbarnoud]

@kain88-de @richardjgowers Can I merge, or should I wait for 0.16.2 to be released?

[richardjgowers]

@jbarnoud yeah tests aren't user facing, so we can merge this

[richardjgowers]

@jbarnoud I take that back, maybe we're not doing pytest in 16.2... #1401

[orbeckst]

#1420 (comment)

It tries to download some hole dependency from dropbox and that is a part of the ci-helpers we are using.

Just some background: hole is not part of the ci-helpers. It is an external software that we use for the MDAnalysis.analysis.hole module. In order to test the module, we need to install the software. The license of hole is very unfriendly so the only legal approach (as far as we could figure out) is to always download it from the official site, which happens to be a dropbox location...

Or at least that was the story until 2 minutes ago when I figured out that hole is now open source: https://github.com/osmart/hole2 ... so I think we will soon have conda packages. ;-) #1429

[richardjgowers]

@utkbansal So this PR didn't actually add topology to the pytest list in travis.yml 😆

[utkbansal]

Oh man! Rebasing and stuff can be very confusing if you have 4 PRs in parallel. 😆

[richardjgowers]

Put it in the xdist pr

[jbarnoud]

On 06/25/2017 01:09 PM, Richard Gowers wrote: Put it in the xdist pr — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1420 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABUWukmyOaDP_1CI1hqmEbk3_rapJM3Hks5sHj_jgaJpZM4OBZLH>.

Just to reiterate a previous suggestion: having a file that list the modules for nose and an other file that lists the modules for pytest would reduce the need for rebasing. One problem we currently have is that every PR from Utkarsh change the same two lines in .travis.yml. Having the lists in separate files would hopefully reduce the conflicts.

[utkbansal]

@jbarnoud Makes more sense now!