motis: Enable duplicate matching

[jbruechert]

No description provided.

[felixguendling]

Loading performance with this option for timetables with large overlap is not really well tested. It works well for the MOTIS demo instance (where only a few long distance trains overlap).

This vecvec<K, V> data structure is not really made for many push_back operations. If you want to use this for large overlaps it might make sense to try a different data structure (like mutable_fws_multimap).

https://github.com/motis-project/nigiri/blob/d806245b6b313f143e1705a01df4dc7d9abbbe19/src/loader/build_footpaths.cc#L152

https://github.com/motis-project/nigiri/blob/d806245b6b313f143e1705a01df4dc7d9abbbe19/include/nigiri/timetable.h#L356

[vkrause]

Loading times still seem ok here locally (but then we don't really have large overlaps yet). However it's not filtering the DE <-> NL Regionalexpress duplicates that triggered this originally either it seems.

Your optimization suggestion however seems to be the solution for a problem we have with GTFS-RT updates :)

[felixguendling]

Maybe it makes sense to create a minimal example with those trips? Usually what I do is:

• click on the trains on the web UI, get the trip_id from the URL
• grep for the corresponding trip_id in the GTFS data and create a stop_times.txt with only times from this trip
• reuse all other files (stops.txt, routes.txt, etc. from the GTFS feed(s) and call gtfs-tidy
• voilà - you have a minimal testing example where you can print some output on where the code bails out and decides that they can't be duplicates

Maybe it would even make sense to write a very simple command line tool to do this extraction (like gtfs-extract feed.zip trip_id1 trip_id2 ...) to be able to generate minimal examples and debug them quickly.