Check on expected behaviour of dodgr_contract_graph and junction points

[al-obrien]

I'm a new user of this package and it has been an excellent experience so far.

I wanted to confirm if an observed behaviour of dodgr_contract_graph is expected. I understand that it will try to reduce the volume of vertices while preserving junctions. However, when plotting how a route navigates through vertices in the full vs contracted network, there are some jagged features and missing vertices that otherwise appear to be a junction. This seems to be a more obvious problem in dense neighborhood regions. Is this expected?

I have provided a side-by-side image below that shows the vertices and route shape on the full and contracted networks. I have also circled, in green, some of the "artifacts": areas that have no vertices when contracted but are junctions. I imagine this is simply how the algorithm behaves under these curvy roads? The road networks were sourced from the {osmextract} package.

[mpadge]

Thanks @al-obrien for reporting. To investigate further, I'll need reproducible code. Could you please use the reprex package to paste a reproducible script here? You can start with a sample of your data covering the area shown, and upload that to this issue by dragging-and-dropping into a comment here. You might need to do that first, then after it's been uploaded, you can get a URL for the data, and use that to download it and create a fully reproducible script. Once I have that, I'll be able to investigate further.

One note though: osmextract delivers data in sf format only, which is then strictly geometrical, and so loses much of the data necessary for accurate routing. This may well end up being the cause of your problems. Accurate routing with dodgr requires using the osmdata package (or dodgr_streetnet() here, which does the same), and ideally extracted as osmdata_sc()-format.