duplicate_key: duplicate primary key in a GTFS file Error
Two rows of the same GTFS file share one primary key. What duplicate_key means, where duplicates usually come from and how to remove them from the feed.
What this rule means
Every GTFS file has a primary key that must be unique: stop_id in stops.txt, route_id in routes.txt, trip_id in trips.txt, service_id in calendar.txt. Some files have a key made of several fields: trip_id and stop_sequence in stop_times.txt, service_id and date in calendar_dates.txt, shape_id and shape_pt_sequence in shapes.txt. The rule fires when two rows share the same key.
Severity: error.
Why it matters
A reference to a duplicated key is ambiguous: a trip with route_id L3 does not say which of the two L3 routes it follows. Consumers usually keep the first or the last row and drop the other without telling anyone. In stop_times.txt, two stop times with the same stop_sequence leave the order of stops undefined.
Common causes
- Two exports appended together (two periods, two operators) without renumbering.
- A row copied to create a variant, with its identifier left unchanged.
- A
stop_sequenceretyped by hand after a stop was inserted. - An exception in
calendar_dates.txtentered twice for the same date, sometimes with two contradictoryexception_typevalues.
How to fix it
Compare the duplicated rows. If they are identical, delete one. If they describe two different things, give one of them a new identifier and update every reference that should point to it (trips.txt, stop_times.txt...). In stop_times.txt, renumber stop_sequence in the real order of calls. In calendar_dates.txt, keep the exception that matches the service intended that day.
In GTFS Express, the validation report lists each occurrence with its file and row. "Fix one by one" opens the rows concerned in the editor; each change becomes a retouche in the draft, and "Re-validate" confirms the duplicate is gone.
Example
Two different stops share one stop_id:
stop_id,stop_name,stop_lat,stop_lon
S100,Station,49.8920,2.3070
S100,Town Hall,49.8945,2.2990After the fix, the second gets an unused identifier and the stop times that serve it are updated:
stop_id,stop_name,stop_lat,stop_lon
S100,Station,49.8920,2.3070
S101,Town Hall,49.8945,2.2990See also the list of validation rules.