Breaking up the LCV

[rogerhutchings]

Package

lib-classifier

Feature or Issue Description

I'm not happy that the LCV is a big ball of code with no tests, so here are some thoughts:

• We break up the LCV into a plain viewer (for Talk), viewer with non-interactive annotations (for feedback) and marking tools
• We add tests
• We add an event bus for the classifier to handle button interactions; there's a discussion to be had as to whether viewer state is kept in the viewer or the store, but a common interface to trigger these would be a good first step
• We break off the LCV into a separate package, and dynamically import it when required. It comes with dependencies like D3 that won't be required

cc #366

[eatyourgreens]

#648 makes the big ball of untested code even bigger. Could feedback use a decorator: withFeedback(SubjectViewer) to add feedback marks to the viewer?

@rogerhutchings have you got time to comment over there about making feedback generic, and less coupled to the light curve viewer?

[srallen]

@eatyourgreens Mark is working on that refactor but I asked him to push up what he had done now so that we can have the UI in place for beta testing with volunteers.

[srallen]

We break up the LCV into a plain viewer (for Talk), viewer with non-interactive annotations (for feedback) and marking tools

The LCV shouldn't even be coupled with marking tools. We have an incoming project with CSSI that will want to visualize the light curves of variable stars, but the task is to identify them using the survey task, so no marking will be happening at all.

[srallen]

The LCV should be able to take props to set:

• Axis labels
• Zoom type: x-axis only or x and y (brush zoom)
• Axis inversion (For LCs this is typically the y)
• Initial zoom level

[eatyourgreens]

In my mind, any feedback between task and subject should happen via a common parent or store. Or the subject viewer should be a child of the task component, so that the task can pass props to the subject.