[WIP] updating SMIRKS parsing in ChemicalEnvironments

[bannanc]

This pull request primarily addresses issue #248

I'm still working on testing changes and they aren't quite ready, but I wanted to get this open just to have a record of what I have worked on so far. I'll add some more information tomorrow including updating examples (here and in openforcefield repo) to show how Chemical Environments can be used.

It also adds an isValid() method to Chemical Environments as discussed in issue #96

[jchodera]

We've migrated environment.py out of this repo and into https://github.com/open-forcefield-group/openforcefield as https://github.com/open-forcefield-group/openforcefield/tree/master/openforcefield/typing/chemistry

We should probably update that version instead and migrate the smarty code to use that version, unless both are still rapidly changing at the moment.

[davidlmobley]

Yes, @bannanc - can you make the modifications to chemical environments on the openforcefield repo and instead make a PR here that will use the chemical environments from openforcefield?

And, @jchodera - do we have a conda-installable version yet?

@bannanc - one issue will be that this makes the process of "implementing new code on openforcefield" and then "using new code from openforcefield in smarty" somewhat more involved, in that to make updates to openforcefield conda-installable, we need to do as follows, I believe:

• Cut a new point release
• Do a pull request to update the conda-recipe on github.com/omnia-md/conda-recipes to use the new point release
• Get that PR merged and wait for it to build (usually overnight)
• Then it's conda-installable In the meantime you can of course use/test it locally by installing directly, but Travis-CI won't be able to install it. But, still, having hte chemical environments in openforcefield is the best way to do things longer-term.

[jchodera]

And, @jchodera - do we have a conda-installable version yet?

It's been conda-installable for 11 days: https://anaconda.org/omnia/openforcefield/files

You can also install it via pip if you want to grab the development version:

pip install git+https://github.com/open-forcefield-group/openforcefield.git

[jchodera]

We could have the smarty travis install the latest development version via

pip install git+https://github.com/open-forcefield-group/openforcefield.git

if things are going to be changing rapidly within openforcefield for a while.

[davidlmobley]

Ah, that's a good idea, @jchodera . Note the above, @bannanc .

[bannanc]

@jchodera and @davidlmobley Yes, I know we've moved this, I wrote the issue here initially thinking it might be a problem with smirky, but it seems to be a problem with how the chemical environments were writing out replacement strings. Testing that lead to a few other types of SMIRKS strings that the environment couldn't parse correctly. I started an issue on that repo as well.

I will use the open pull request to make the slight modification I have to smirky and make it so smarty and smirky import from the openforcefield modules.

I'll add the updated environments to my WIP pull request in openforcefield.

[bannanc]

I'm getting closer to solving the parsing issues, I'm working on writing some "extra weird" SMIRKS strings to test it.

Once I'm convinces the parsing works, I'll move everything to openforcefield and make sure the smarty/smirky scripts use openforcefield.

[bannanc]

I thought I was getting closer, but given the way tests are failing, I've clearly broke a few things also...

[bannanc]

Updated environments are now available in the openforcefile repo.

I have changed this PR to be about switch smarty to be dependent on openforcefield, but I think the discussion here is confusing so I am going to close this and open a new PR for removing openforcefield redundancy.