Add deregister_toolkit method

[mattwthompson]

#468

• Maybe add round-trip tests?
• Maybe add tests using GLOBAL_TOOLKIT_REGISTRY (codecov counts the relevant lines of code as tested, but I think they're only ever called by things that use this: grep -R "GLOBAL_TOOLKIT_REGISTRY" openforcefield/tests/ comes up empty; related GLOBAL_TOOLKIT_REGISTRY does not behave as anticipated #493)
• Tag issue being addressed
• Add tests
• Update docstrings/documentation, if applicable
• Update changelog

[lgtm-com]

This pull request introduces 2 alerts and fixes 1 when merging 3f3be8a into 97d31e6 - view on LGTM.com

new alerts:

• 1 for Syntax error
• 1 for 'import *' may pollute namespace

fixed alerts:

• 1 for Unused import

[codecov-commenter]

Codecov Report

Merging #614 into master will increase coverage by 0.12%.
The diff coverage is 100.00%.

[lgtm-com]

This pull request introduces 2 alerts and fixes 1 when merging 5365cf9 into 131d86c - view on LGTM.com

new alerts:

• 1 for Syntax error
• 1 for 'import *' may pollute namespace

fixed alerts:

• 1 for Module is imported more than once

[mattwthompson]

I bumped this up in priority since it's been desired multiple times in recent memory. I've merged in the other 0.7.0 changes, updated tests and docstrings, and also catch the case of attempting to deregister an un-registered toolkit. I believe this is ready for review now.

I've completely punted on retooling GLOBAL_TOOLKIT_REGISTRY since I want to get this feature out sooner and the two changes should be mostly de-coupled from each other.

Tangentially, the toolkit wrappers could probably do with a __str__ or __repr__ override to make things like this a little prettier:

In [4]: registry = ToolkitRegistry(toolkit_precedence=[AmberToolsToolkitWrapper])

In [5]: registry.deregister_toolkit(RDKitToolkitWrapper)
---------------------------------------------------------------------------
ToolkitUnavailableException               Traceback (most recent call last)
<ipython-input-5-0e9acceb2423> in <module>
----> 1 registry.deregister_toolkit(RDKitToolkitWrapper)

~/software/openforcefield/openforcefield/utils/toolkits.py in deregister_toolkit(self, toolkit_wrapper)
   4348                     f"Currently registered toolkits are {self._toolkits}"
   4349                 )
-> 4350                 raise ToolkitUnavailableException(msg)
   4351
   4352         for toolkit_to_remove in toolkits_to_remove:

ToolkitUnavailableException: Did not find <openforcefield.utils.toolkits.RDKitToolkitWrapper object at 0x129036750> in registry. Currently registered toolkits are [<openforcefield.utils.toolkits.AmberToolsToolkitWrapper object at 0x111610990>]

In [6]: registry.registered_toolkits
Out[6]: [<openforcefield.utils.toolkits.AmberToolsToolkitWrapper at 0x111610990>]

I could probably sneak that in here - but when I say that, it typically works to not be true after trying it.

[lgtm-com]

This pull request introduces 2 alerts and fixes 1 when merging 45ab3da into 131d86c - view on LGTM.com

new alerts:

• 1 for Syntax error
• 1 for 'import *' may pollute namespace

fixed alerts:

• 1 for Module is imported more than once

[lgtm-com]

This pull request introduces 2 alerts and fixes 1 when merging 3fa8d0f into 131d86c - view on LGTM.com

new alerts:

• 1 for Syntax error
• 1 for 'import *' may pollute namespace

fixed alerts:

• 1 for Module is imported more than once

[lgtm-com]

This pull request introduces 2 alerts and fixes 1 when merging d78e111 into e1b6e74 - view on LGTM.com

new alerts:

• 1 for Syntax error
• 1 for 'import *' may pollute namespace

fixed alerts:

• 1 for Module is imported more than once

[j-wags]

Nice tests -- Very clever and thorough. Made a few direct suggestions, and there's one docstring that seems to have wandered off, but after that I think this is good to merge.

Thanks, Matt!

[j-wags]

Yup. This is a pretty weird one. Hopefully this can be made more sensical when we move to a singleton design pattern or context manager approach.

[j-wags]

Hooray for specific exceptions!