Determine local environments based on bonding analysis

[JaGeo]

Summary

This is work in progress to add an analysis of the local environments based on bonding analysis. The classes will allow to both use localenv and chemenv in combination with Lobster output. It will also provide additional methods to analyse the ICOHPs/COHPs for sites with neighbor atoms.

@mkhorton , let me know if you already have comments.

I hope it is okay to put it in pymatgen.analysis.lobster_env

[shyuep]

I would like to have a better organization of all the local environment analysis packages, including local_env and chemenv. Can we create a parent package called pymatgen.analysis.local_environment to house all of them. I would also prefer some common API that unifies all the different kinds of analysis and the various options. Right now, the packages are so big, fragmented and difficult to find and use.

[JaGeo]

I am happy to move my code to another place.

I am not sure regarding the common interface and moving the localenv and chemenv code (it will probably break a lot of existing code and documentation in publications and so on) but there are other people that might be able to comment better on this.

[shyuep]

Well, all this is open for discussion. There is no need to create breaking changes. We can always retain existing code where they are with deprecation messages. @mkhorton what do you think?

[mkhorton]

The main problem is that chemenv and local_env are developed by different people/teams and have different philosophies.

The code @JaGeo is proposing is related to neither, but based on the Lobster code.

There has been some attempt at having a common interface, namely the NearNeighbors class (in local_env) is an interface to define neighbors and StructureGraph in pymatgen.analysis.graphs is a container to store the resulting graph. I think both of these are good.

@JaGeo here is using NearNeighbors here, so is conformant with that interface, so I think this is a good PR. I think it should be in its own file too, due to its length, but perhaps io.lobster is the better location since this is so closely coupled to the output of the Lobster code.

Chemenv (unrelated to this PR) pre-dates the NearNeighbors class, but it would be nice to get a NearNeighbors interface to Chemenv as well. It is too large a module to merge with local_env however, but a better page on all our local environment algorithms would be good in the docs to explain that we do provide multiple algorithms, and to give the relevant publications for each.

[JaGeo]

I am totally fine with moving the code to io.lobster.
Building some interface to ChemEnv via the NearNeighbors class also seems to be feasible (but I would expect that it would only be able to use a limited number of the capabilities of ChemEnv).

[mkhorton]

Building some interface to ChemEnv via the NearNeighbors class also seems to be feasible (but I would expect that it would only be able to use a limited number of the capabilities of ChemEnv).

Yes, perhaps it depends on what kwargs are available for ChemEnvNeighbors(NearNeighbors), I hope it may be possible to include the advanced functionality too. But ChemEnv is outside the scope of this PR so let's not worry about it here.

[coveralls]

Coverage decreased (-0.4%) to 79.521% when pulling c13852b on JaGeo:master into 754c259 on materialsproject:master.

[JaGeo]

I am now quite happy with the current state of the code. I would be happy if this could be merged for now. (I will probably extend the code within the next months)
I am still having problems with some of the tests though and I am not completely sure what is going on. At least all tests on ubuntu work.

[shyuep]

I know what the issue with the mac/win tests is. I have made changes. I have merged this PR. Thanks.

[JaGeo]

Thank you.