Add router tests

[nilsnolde]

We’re resting already, which is great. But we only test the self contained example instances. It’d be worthwhile to test the integration with all routers with Andorra or so, so we get notified ahead of time when smth breaks. Best in a scheduled GitHub Action, e.g. every 2 weeks as there’s hardly any activity here unless we’re releasing.

[jcoupey]

there’s hardly any activity here unless we’re releasing

For that exact reason, I would not bother to test frequently. I mean if we only run tests on release PR, we'd still catch any problem before releasing.

[nilsnolde]

this would just test integration with the current versions of routers, not any code of this repo or vroom. we offer a docker-compose here which I kind of take responsibility for, since it's inside the repo. any of the routers could change their latest image at any point (that's kinda the same as depending on master branch instead of a release), breaking stuff in this repo's docker-compose.yml. so it'd be regular integration tests basically. few alternatives exist of course:

1. remove the docker-compose.yml, which I'm pretty open to tbh. it does make startup so much easier and faster for new users, but it also attracts lots of people who have zero clue about software or docker (see recent issue..). people who can handle docker will only spend extra 20 mins setting up a docker-compose.yml. in this case I wouldn't feel like I'd need to test integrations
2. pin the images we reference in the docker-compose.yml to a specific version of the routers (instead of latest) and let the community keep it up-to-date. in this case I would feel like we should test integation. it would be nice to have one place that tests valhalla & ors no? vroom tests osrm extensively, but not the others.

option 1 sounds tempting to me personally though :) what's your thoughts @jcoupey ?

[jcoupey]

I agree we should go for option 1 here since:

• the scope for this repo is to provide a Docker setup for VROOM, not a full Docker infra working out of the box with any router;
• the scope of this organization is to develop an optimization engine, so of course routing is related but users should be able to handle their own routing in any case (be it from source, using Docker or whatever).

Taking responsibility for maintaining working integration for all routers is just asking for trouble for a small benefit, especially if docker-composing is straightforward as you point out. And it's always possible to deal with the occasional questions in the issues.

We should probably adjust the current readme if removing docker-compose.yml?

[nilsnolde]

yeah, it's a thin line for these kind of wrapper projects..

We should probably adjust the current readme if removing docker-compose.yml

jep, I'll draft smth soon.