Removing Dead, Deprecated, and Unreachable Code

[goldmanm]

I am curious if we want to deprecate any functionality in RMG. I am thinking that not-often-used and poorly understood functionalities could be removed to allow easier bug checking, code support, and development. We may also be able to just cut some redundant methods. One example I came across would be to remove the method rmgpy/molecule/molecule/Bond/isSpecificCaseOf and integrate its code into rmgpy/molecule/molecule/Bond/equivalent.

If people think it may be worthwhile to have a list of methods to cut, we may want to have a warning placed in docStrings, similar to how Cantera placed some in their code to warn users.

[goldmanm]

#1021 mentions some to get rid of

[goldmanm]

I couldn't find a place where the attribute Reaction.label is clearly assigned and utilized. Possibly we could deprecate it.

The Reaction.__str__ doesn't even use the label attribute.

[goldmanm]

I just thought we might want to deprecate the restart file functionality. It didn't seem like it was working over 2 years ago, and as of now, it doesn't seem to be usable either (due to long saving times).

[alongd]

Does anyone use RMG-Java?
Perhaps we could depreciate the duplicate code for dealing with this older format?
I think that leaving just the import scrips is enough.

[mliu49]

I agree with deprecating RMG-Java methods. See #1021 for additional discussion related to this.

[mliu49]

As far as I'm aware, we don't have a existing deprecation process. It might be good to start one though. I think for typical software projects, deprecation means that a feature is no longer supported or recommended for use. Complete removal of the feature is done later on a fixed timeline.

This might be difficult for us though, since we don't have a fixed release schedule. A possible plan for us might be to add the deprecation warning in one release, then remove the method in the next release. Unless we start making small releases more often, in which case we could have a couple releases between deprecation and removal.

[alongd]

I was too fast to vote for removing on #1021. I agree that we should develop a deprecation process.

[goldmanm]

We may also want to deprecate the non-web versions of scripts like diffModels, and only support one type of output for them.

[alongd]

Perhaps allowSingletO2 option in the input file can now be depreciated.
If O=O is generated, an unidirectional template (ReactionMechanismGenerator/RMG-database#239) converting it to [O][O] is applied.

[alongd]

We could also depreciate the saveRestartPeriod now that automatic seed generation works
Edit: I see @goldmanm already mentioned that above. It might be time to remove it.

[goldmanm]

I'd add group additivity methods in 'family.py' to the list of deprecatable methods.

[alongd]

We could deprecate the FAME (used in RMG-Java) to CanTherm/Arkane file type converter.
(Has anyone ever seen a FAME input file lately?)