Closed scmcca closed 3 years ago
I think the intention of locations.geojson was to only represent area-based polygons (whether that's a route deviation buffer zone, a contiguous demand-responsive service boundary, a single polygon with multiple isolated areas, etc.). From the spec: "Individual stops should be defined in stops.txt
."
I've been assuming an on-demand point to point service (previously unordered stops in v1) would be represented by a location_group consisting of regular stops that has a single start_pickup_drop_off_window
/end_pickup_drop_off_window
with pickup_type
/drop_off_type≠0
. Could/should this kind of service be modeled a different way?
Representing an on-demand point to point service in this way would make arrival/departure times forbidden ("Forbidden when stop_times.stop_id
references a location_groups.locationg_group_id
"). Does that solve the issue?
That makes sense! I hadn't thought through the data enough for the use case. Thanks!
Currently stop_times.arrival_time and .departure_time are forbidden if .stop_id refers to geojson areas. This is an ad-hoc way to say that arrival and departure times should not be defined for on-demand service, but does not capture point to point on demand service (i.e., point type polygons).
Therefore, a different approach should be considered for "hard coding" on-demand data in GTFS to avoid future conflicts.