A service's day as flat tables: the headline, every scheduled journey run or not, and every stop of every journey tracked. Pick the services and the days and it is built on demand — the same code the command line runs, so a correction to how a journey is matched reaches this file the day it is made.
journey_key, not trip_id.
The trip id is BODS's, and it is a content hash of the journey: 22.8% of one
Monday's journeys were renamed by a republication a week later, so a report
joined across two weeks on that column would quietly drop a quarter of them.
journey_key is built from what a re-issue leaves alone — date,
operator, line, direction, origin stop and aimed departure — and resolved
99.9% of a day against the timetable published a week earlier.route_id and route so the rows say which.
Where a number is one service, and almost all of them are, it stays a bare
number.jny_canc and
jny_part_canc are written empty rather than 0, because 0 would say
nothing was cancelled and this data cannot say that.observations 0, where theirs holds observed stops only. Filter to 1 to
match their row count exactly.not recorded
rather than blank, because blank has to mean we looked and saw none.