Scalable Recipe Tests - Githubissues

phurwicz / hover

:speedboat: Label data at scale. Fun and precision included.

https://phurwicz.github.io/hover

MIT License

323 stars 19 forks source link

Scalable Recipe Tests #26

Open phurwicz opened 2 years ago

phurwicz commented 2 years ago

Sketch / Food for Thought

Current testing procedure SELECTING POINTS TO LABEL

make selection in
[movable] enter in annotator_input
action_view_selection() -> check selection table
make sub-selection in selection table
action_evict_selection() -> check old/new selection
to be continued. Apply twice then commit? Commit twice then deduplicate? Repeated deduplication should have no effect? Neither does repeated push? Check selection sync across multiple explorers at random points? Generate filter conditions and expected outcome?

The state of a plotted recipe

dataset/explorer consistency
- rows for each subset
- recovered by PUSH
- affected by COMMIT, DEDUP
- unchanged by VIEW, EVICT, PATCH
- data values (including SUBSET)
- recovered by PUSH
- unchanged by
dataset/selection table consistency: mutable data values
explorer/selection table consistency
- selected indices
- recovered(or reset) by VIEW, PUSH
- unchanged by PATCH
- affected by COMMIT, DEDUP, EVICT
- data values (including SUBSET)
- recovered(or reset) by VIEW, PUSH
- unchanged by EVICT
- affected by COMMIT, DEDUP, PATCH
widget states
- checkbox toggles
- input/slider values