4 ms·

Instead of indexing the geofences using R-tree or the complicated S2, we chose a simpler route based on the observation that Uber’s business model is city-centr

Instead of indexing the geofences using R-tree or the complicated S2, we chose a simpler route based on the observation that Uber’s business model is city-centric; the business rules and the geofences used to define them are typically associated with a city. This allows us to organize the geofences into a two-level hierarchy where the first level is the city geofences (geofences defining city boundaries), and the second level is the geofences within each city. What about Uber's business model impacted this choice of algorithm? Edit: left only the unanswered question

chris11 9y ago
I'm not quite sure what you are referring to by strict background checks. But here's I think the service is used. A user initiates a call to the service requesting a ride. The service takes in the user's location and first finds the city (first tier), then the service searches and returns the georeferences in that city for ones that contain that user location. These georeferences might attached to drivers, and those drivers might be analyzed to find good drivers to send that ride request to, or they could be something like neighborhoods, where Uber would search that neighborhood for drivers. I'd guess that city specific requirements could be associated with georeferences in either tier.