Add Motis APIs for network topology and other non-routing tasks

[twatzl]

Hey,

I just learned about motis in the context of the Transitous project. I think it is really cool what is happening there and the task of compiling and handling all that data is really huge. That is why it think it could be very profitable to extend motis to support some non routing tasks.

Namely I have in mind/looking for the following use cases:

• I want to get the whole network of bus/tram/subway lines (routes) in a city or long-distance train routes in a country
• I want to be able to get all trips for one route
• I want to get all trips for a set of routes between hours x and y
• I want to get all routes for operator x

The idea behind this is that i would like to do some analysis on transport networks and also visualize them, maybe using https://github.com/ad-freiburg/loom

Or using the data to automatically generate speed maps for certain types of transport like shown in https://marcochitti.substack.com/p/getting-bus-priority-right-lessons

Why do I think it should be part of Motis?
Because Motis seems to be already a place where multiple GTFS streams are combined and so it seems logical to start these queries here as otherwise one would have to build a complete stack again to answer the above questions. It might not be part of the core question that motis is trying to answer, but i still think it would be the right place to add this.

Furthermore, especially in urban networks it is often hard to get useful transit diagrams from official sources. Sometimes they do not exist at all, or sometimes when they exist they might leave out regional busses or trains, because the map is just provided by one operator. When this kind of automation would work, then this would also be a benefit to everyday users.

[felixguendling]

We're interested to have something like Loom directly inside MOTIS (e.g. as additional tile layer).

MOTIS merging of GTFS feeds is quite simple at the moment:

• prefix all trip and stop IDs with a dataset tag
• add footpath transfers and equivalence relations of stops from different datasets based on proximity

For your goals, probably only the first one is needed.

I'm not opposed to add a few more endpoints to MOTIS (including functionalities you mention). However, I don't think that MOTIS is the only tool that can accomplish this. Adding a prefixing logic to a tool that loads GTFS into a Postgres database could probably do the same. And there, each of your endpoints would be a simple SQL query instead of hundrets of lines of C++ code.

[twatzl]

Ah I see. Yes that makes sense. Maybe I'll write a small server at some point.

So you'd argue for the structural queries that do not require routing it would be more beneficial to just do static queries to a database. I guess the only downside is then that you'd have to run the ingestion pipeline twice. (in case you want to use the structural data in an application and not just for generating nice graphics.)

[felixguendling]

The question is: on which scale do you want to do this? If you just want to connect a few GTFS datasets locally, anything will do the job. You can add endpoints to MOTIS and query them or you can use a SQL database. It will all be reasonably fast and RAM usage will be low.

If, however, you want to do this on planet scale, querying MOTIS a million times and then doing a lot of work outside MOTIS will be slow and you do a lot of unnecessary serialization+deserialization. In this case, I would propose to do everything (e.g. creating a tile layer for transit network plans) in MOTIS and then provide an endpoint for the end result. Otherwise, if you add a few endpoints but try to do everything outside of MOTIS, it will still be slow.

I think what you can already do now is to query map/stops endpoint and the stoptimes endpoint. Together you can retrieve the whole timetable.

[laem]

Checkout https://codeberg.org/cartes/serveur

We're producing transport maps and stop hours APIs that can be seen here : https://cartes.app/?style=transports#7.43/48.255/-3.119

We're using node-GTFS : a local DB filled with GTFS and easy to query (despite not as quick as Motis, far from it).

[laem]