Bumb the minimal MDAnalysis version required

[jbarnoud]

We were asking for MDAnalysis >= 0.11, but that version is ancient and
the API changed rather drastically since then. I bumped the minimum
required version to 0.16. The tests run fine with version 0.17 that
should be preferred, especially on python 3.

@HubLot I do not know what should be done for the conda package to be rebuilt. It should now be possible to install PBxplore on python 3 with conda as MDAnalysis was the last obstacle.

[pierrepo]

I totally agree. We need to push toward the use of a recent version or mdanalysis.

@HubLot for conda. Could we create a quick recipe to build the package upon bioconda? This looks like the standard for bioinformatics tools.

[HubLot]

It's doable now since MDAnalysis is already in a conda channel. I think @jbarnoud's modifications are enough to create a conda package in a custom channel. If it succeeds, yes, we can offer a recipe for bioconda.

I can look into it next week.

[pierrepo]

modifications are enough to create a conda package in a custom channel.

@HubLot instead of creating our own custom conda channel, could we join bioconda channel? Their build process seems pretty robust (https://bioconda.github.io/contributing.html)?

[HubLot]

Yes, it is the main objective. I just wanted to try the build first before submitting to bioconda.

[pierrepo]

OK, as you wish. But the build process proposed by Bioconda is pretty safe (everything is build upon Travis)

[jbarnoud]

On 02/12/2018 02:05 PM, Pierre Poulain wrote: OK, as you wish. But the build process proposed by Bioconda is pretty safe (everything is build upon Travis) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#172 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABUWutK9Jiup-JXNpCu0Hp_ZZlMfSutXks5tUDc2gaJpZM4SBTBo>.

For the record, the MDAnalysis conda package is on conda-forge.

[HubLot]

So, I retried to build a conda package from the command line but now it fails... (either on the master or this PR branch) :

see this gist: https://gist.github.com/HubLot/18da1e2063001a1a51c73916ab4adff7

It seems it's an issue on Pytest but I'm not sure. I'll try to dig more into it but I'm quite busy in the next 2 weeks.