Introducing a test framework

[guillaumerose]

I am trying this project and it works well for me. It is far more convenient and efficient. I can see some differences between the output of tilemaker and the output of openmaptiles.

Would that make sense to write some tests where we pass a .osm.pbf with few roads, etc. and assert the output is correct ? I see the CI generates a small area but there are no expectations. Should we add some ?

[pnorman]

Are you wanting to test tilemaker, or the Lua scripts that come with it?

[guillaumerose]

I think we should start with an end-to-end test to cover most of the features and ensure no regressions when a PR comes in.

[systemed]

I like the principle. Possibly a pipeline like .pbf -> tilemaker -> .mvt -> .geojson, and then we parse the geojson to make sure the features are correct. Testing the tile filesize would be useful too. Potentially quite a big undertaking though!

[kleunen]

The project would definitly benefit from testing. Even if the conversion fails and generates no output, the CI passes. I have seen the tilemaker executable crashing in CI, but CI says build is complete. So even checking if the output file exists and is not empty, would be an improvement