Data & Filtering Architecture

[itlackey]

After thinking about it, I think we should classify filters into "global filters" and "chart filters" and then separately we can have "regions".

Global filters

• Currently time, cohort, and cases.
• They remove patients from the patient overlays (dot-density, 2d-density, contours) and most charts
• Some charts shouldn't be affected by the filters. For example, some time series charts should probably show the full time scale and have a vertical line to indicate the current date (or date range).

Chart filters

• Any filter found in the charts (age, race, symptoms, etc).
• They remove patients from the 2d-density and contour overlays, but will change the appearance of the dots in the dot-density (no fill, just a border). This way we'll highlight the patients that pass the chart filters, but still leave some way to compare against the patients that were filtered out.
• Chart filters might also change the appearance of the charts rather than completely remove them.. not sure yet.

Regions

• Gives in a black border to the dots in the dot-density overlay. Has no effect on 2d-density or contours.
• Each region will keep track of which patients are in the region.
• Some charts will use the patients after global filters, but some (like some time series charts) will use the full time scale.

So I'm thinking from a data structure standpoint, App should keep track of:

• Patients
• filtered_patients (after applying global filters)
• chart_filtered_patients (after applying global+chart filters)

Each region should keep track of:

• patient_ids within the region
• filtered_patient_ids within the region (after applying global filters)
• chart_filtered_patient_ids within the region (after applying global+chart filters)