Scaffolding for pluggable components

[drewoldag]

A placeholder issue for implementing system for pluggable components. This likely needs to broken up into a handful of smaller, more focused issues.

[drewoldag]

One place where we can add simple scaffolding is in database.py around lines 1347-1390. @jeremykubica suggested a simple registration system with a parent class and then assorted subclasses this makes a lot of sense in this case.

@AmandaWasserman pointed out that down the line, we'll probably want to let people bring their own search strategy (or other classes) - we'll keep that in mind and make sure the registration system can handle that.

[drewoldag]

Also checkout database.py lines:

• 268-365 load_features_from_file These are the hardcoded feature extractor names.
• 949-971 classify These are the different classifiers
• 1240-1289 make_query_budget Similar to the second comment in this issue - basically everything that is going on in query_budget_strategies.py.

[drewoldag]

After chatting with Amanda, the three classes (or will likely become classes) that should be made pluggable are:

• classifiers.py
• metrics/query strategy in query_strategies.py
• feature extractor resspect/feature_extractors/[bazin|bump|light_curve|malanchev].py

[drewoldag]

Pluggable classifiers has been mostly taken care of with PR #45. There are a couple of follow up issues that should be addressed with #48 and #49.

After those, we should move over to making the feature extractors pluggable.