The Geospatial Matching Problem

When a passenger taps "Request Ride", the dispatch server must:

  1. Find all active, unoccupied drivers within a 3-kilometer radius.
  2. Sort drivers by estimated arrival time (accounting for one-way street topology).
  3. Sequentially dispatch the ride offer with a 15-second acceptance countdown timer.

Sub-Millisecond Radius Lookups with Redis GEO

Querying a relational SQL database with WHERE ST_DWithin(...) on every driver's 3-second GPS ping will saturate disk I/O. Instead, we stream high-frequency coordinate updates into Redis in-memory spatial sorted sets: 

import redis

r = redis.Redis(host='localhost', port=6379, db=0)

def update_driver_location(driver_id: str, longitude: float, latitude: float):
    '''Records driver position in Redis GEO sorted set.'''
    # GEOADD key longitude latitude member
    r.geoadd("active_drivers_geo", (longitude, latitude, driver_id))
    # Keep 60-second expiration heartbeat to clear offline drivers
    r.setex(f"driver_heartbeat:{driver_id}", 60, "online")

def find_nearby_drivers(passenger_lon: float, passenger_lat: float, radius_km: float = 3.0) -> list:
    '''Finds nearest active drivers within radius sorted by ascending distance.'''
    # GEORADIUS key longitude latitude radius unit WITHDIST WITHCOORD ASC
    results = r.geosearch(
        "active_drivers_geo",
        longitude=passenger_lon,
        latitude=passenger_lat,
        radius=radius_km,
        unit="km",
        withdist=True,
        withcoord=True,
        sort="ASC"
    )
    
    nearby = []
    for member, distance, coords in results:
        driver_id = member.decode('utf-8') if isinstance(member, bytes) else member
        # Verify driver is not in a ghost state
        if r.exists(f"driver_heartbeat:{driver_id}"):
            nearby.append({
                "driver_id": driver_id,
                "distance_km": round(distance, 2),
                "coordinates": (coords[0], coords[1])
            })
    return nearby