-
Notifications
You must be signed in to change notification settings - Fork 43
Roadmap February through April 2016
In the November - January roadmap, I set out to (1) add basic shape visualizations, (2) restructure our database to enable a simple and correct ETL process for CSV and shape datasets, and (3) remove the unique ID restraint we had placed on CSV datasets. Thank you to all of the contributors of Plenario for your help making these happen. (In particular, @lw334 churned out (1).) Also, we were glad to work with @chicago to support OpenGrid's launch in early January.
In roughly this order:
- API: Add spatial joins and filtering.
- Frontend: Add new spatial visualizations and fix longstanding usability issues.
- ETL & Dataset Submission: Improve ETL reliability. Improve submission process.
While Plenario has supported queries that filter on user-provided geometries from the start, one of the most asked-for features is supporting spatial queries based on geometries stored in Plenario. Within a week or two, ESRI Shapefile support will be robust enough for us to allow public shape submission. The next step is adding endpoints that will let users join and filter datasets based on spatial relationships. As we build out a general interface, we want to make sure we're addressing these two use cases (and ship solutions for them first):
-
Filtering points in polygons (OpenGrid use case): Right now, opengrid.io lets users filter queries based on Chicago wards and zip codes by including GeoJSON in the OpenGrid source code. They would like to be able to do this through Plenario on any polygon dataset.
-
Aggregating points to polygons (health sciences use case): The addresses of patients participating in clinical trials are often anonymized up to political boundaries like census tracts (or so I'm told). Health researchers would like to aggregate point data available in Plenario up to those polygons to get a richer understanding of how patients' environments affect clinical outcomes.
We'll want to expose Plenario's new geospatial capabilities in the Explorer. That will include better ways to explore individual shape datasets (Definitely: seeing details on hover. Maybe: finding ways to visualize datasets too large to naively dump in clients' browsers). More exciting will be letting users generate chloropleth views by joining datasets in the UI.
Outside of spatial features, I'd like to do a more general cleanup. Here are some examples of usability fixes I have in mind:
- Improving the response time of timeseries queries by splitting them up and running them asynchronously.
- Automatically selecting a time frame that contains data when inspecting an old dataset. (The default of looking only 90 days into the past is confusing, because it looks like old datasets have no data at all.)
We've been living with ETL issues like #184 for too long. I need to set aside time to tackle reliability issues before adding the next set of features.