fix: Store all drawn marks in the classification

[eatyourgreens]

After creating a new mark, add the new mark to the previous annotation value and store this as a new annotation. Should fix a bug raised in this comment: #5913 (comment)

You can try this out on How Did We Get Here, in the dev classifier. Draw on both frames of a subject, then check whether the drawn marks persist, in the classification, on the second step of the workflow.
https://local.zooniverse.org:8080/?project=communitiesandcrowds/how-did-we-get-here&env=production

Expected behaviour

When I create a new mark, classification.annotations should be updated to include all the marks that I've drawn for the current task.

Actual behaviour

When I draw a mark, the new drawing task annotation only contains mark IDs from the active frame. Mark IDs from other frames are silently dropped from the classification.

The marks themselves are still stored on their respective drawing tools, but the references to them are removed from the task annotation.

[coveralls]

coverage: 81.214% (+0.008%) from 81.206%
when pulling dbe93b4 on eatyourgreens:fix-previous-drawn-marks
into 1be4462 on zooniverse:master.

[eatyourgreens]

@kieftrav @goplayoutside3 I don't have permission to request reviews or add labels to Zooniverse PRs, so I'm pinging you to take a look at this. There might be an urgent issue, affecting live transcription projects, where work is being lost.

I think it's worth double checking whether classifications are broken for live transcription projects, when you draw on more than one page of a subject. It might be that lines are only being saved for the current, active frame. I'm not 100% certain about that, because the code that serialises a classification (before POSTing it to the API) may actually loop through all drawing tools and marks, and serialise them to the saved classification.

From a quick re-read of the code, the saved classifier snapshot might only include mark IDs that were added to the transcription task annotation, so I'm concerned that volunteer work might have been lost. It depends whether annotation.toSnapshot() here includes all drawn marks, or only the marks that are referenced in annotation.value.

front-end-monorepo/packages/lib-classifier/src/store/Classification/Classification.js

Lines 22 to 30 in 6b421ed

	toSnapshot () {
	let snapshot = getSnapshot(self)
	let annotations = []
	self.annotations.forEach(annotation => {
	annotations = annotations.concat(annotation.toSnapshot())
	})
	snapshot = Object.assign({}, snapshot, { annotations })
	return snapshot
	},

[kieftrav]

@eatyourgreens @goplayoutside3 - this 1 line does fix the issue. I want to be clear that we are not losing any data upon classification submission.

[eatyourgreens]

Excellent. If this is just a rendering bug, then that's less urgent.

[kieftrav]

@eatyourgreens - 100% a rendering bug. I couldn't figure out where annotation.values['markID', 'markID'] was getting modified to only include the current frame's marks. That was the line that did it.

[eatyourgreens]

Drawing task annotations are slightly different from other task annotations. They use types.safeReference to hold references to marks.

front-end-monorepo/packages/lib-classifier/src/plugins/tasks/drawing/models/DrawingAnnotation.js

Line 11 in 1be4462

value: types.array(types.safeReference(GenericMark))

Essentially, a drawing task annotation is an array of mark IDs. A side effect of this is that they're updated differently from other tasks, which hold a direct value rather than a reference.

I think this could be automated by having the drawing annotation model observe self.task.marks and update self.value whenever that changes. That would remove the manual update, which could be prone to breakage.

[eatyourgreens]

self.task is a task key in the annotation model (the name is confusing, a legacy of older code.) It can be resolved to an actual task with something like this line.

front-end-monorepo/packages/lib-classifier/src/plugins/tasks/drawing/models/DrawingAnnotation.js

Line 17 in 6b421ed

const actualTask = resolveIdentifier(DrawingTask, getRoot(self), self.task)

Once you have the actual task, you can observe actualTask.marks for changes, as it's a MobX observable array. Then automate updating the stored annotation value whenever a mark is added or deleted.

[eatyourgreens]

I've opened #5923 with those changes to the drawing annotation models. autorun might be more robust than manually updating annotations whenever a drawing tool changes, but it's a more involved change than the one-line fix here.

[eatyourgreens]

Closing this in favour of #5923.