gtfs-rt-demo

This is a minimal demo we hacked together during a working session to read vehicle positions from the MBTA's GTFS Realtime feed. GTFS-RT uses a binary format called protobuf, which makes it a bit more complicated to work with than other MBTA products. To extract a GTFS-RT feed, you need three things:

• The URL of the feed itself — e.g. here for the MBTA
• A .proto file describing the structure of GTFS-RT to Protobuf — e.g. the one checked into source as here gtfs-realtime.proto
• Protobuf's protoc command line tool.

You can use protoc to unpack and view data from the feed:

curl -s https://cdn.mbta.com/realtime/VehiclePositions.pb | protoc gtfs-realtime.proto --decode transit_realtime.FeedMessage

Here, transit_realtime is the package name provided in gtfs-realtime.proto, and FeedMessage is a message entity defined in that file. Running this command prints:

entity {
  id: "y1777"
  vehicle {
    trip {
      trip_id: "53317138"
      schedule_relationship: SCHEDULED
      route_id: "9"
      direction_id: 1
    }
    position {
      latitude: 42.3514175
      longitude: -71.0705185
      bearing: 162
    }
    current_stop_sequence: 4
    current_status: IN_TRANSIT_TO
    timestamp: 1662592634
    stop_id: "145"
    vehicle {
      id: "y1777"
      label: "1777"
    }
    occupancy_status: FEW_SEATS_AVAILABLE
    10: 40
  }
}
# and hundreds more...

This repo also contains a minimal Python script to extract GTFS-RT vehicle positions. We need the protobuf PyPI package, but we also need to generate a Python representation of the .proto file, again using protoc:

protoc --python_out=. gtfs-realtime.proto

This file is checked into source as gtfs_realtime_pb2.py. This is a Python module that contains, among other things, a class for each message type in gtfs-realtime.proto:

>>> import gtfs_realtime_pb2
>>> dir(gtfs_realtime_pb2)
[..., 'EntitySelector', 'FeedEntity', 'FeedHeader', 'FeedMessage', 'Position', 'TimeRange', 'TranslatedString', 'TripDescriptor', 'TripUpdate', 'VehicleDescriptor', 'VehiclePosition', ...]

You can inspect example.py to see how to instantiate entities from a feed URL.

Run it yourself

You'll need poetry to manage Python dependencies. Then:

poetry install
poetry run python example.py