Cross-Trip Canonical Stop Order for Each Route-Direction

Collapses many observed passages of a route-direction into one stop order. Each stop's canonical position is its median offset from trip start across all passages, tie-broken by stop_ref. This is the cross-trip complement to the per-trip chronological stop_sequence the events converters already produce: when passages disagree on order (e.g. several stops share an arrival second), per-trip chronology is ambiguous but the median offset is not.

Usage

rt2s_obs_stop_order(events)

Arguments

events

Observed stop events (see observed-stop-events).

Value

A data.table with one row per served (route_ref, direction_id, stop_ref): canonical_offset (median seconds from trip start) and stop_sequence (1-based order within the route-direction, by canonical_offset then stop_ref). Only served passages count: skipped/canceled rows and rows without a timed arrival are excluded, so a stop that was never actually served (e.g. skipped on every run) produces no row rather than a placeholder position. A stop visited more than once within a single trip (loop/branch service) triggers a warning, since the median collapses the repeated visits into one position.

Reconstructed versus anchored stop patterns

This function reconstructs a stop order from the observations, which is what you want when no usable published pattern exists. It is the wrong tool when one does exist and an analysis depends on the network being identical across the scenarios being compared: the order here is derived from what was observed, so different observations can yield different stop sets, and a scheduled-versus-observed contrast then confounds "service changed" with "the network changed". For that design, anchor on the published pattern with rt2s_baseline_patterns and vary only service levels.

Worth knowing how easily the two are confused: the median-offset rule below is the same rule a GPS-only pattern builder would use, so a pipeline can slide from anchored to reconstructed without any visible signal.