Graph Cache And Custom Models Update

Hey there!

My GH setup includes CH and a custom profiles for a couple of profiles. If I’m changing the json associated with one of this profile do I need to regenerate the graph-cache to get updated results?

Yes. You have to regenerate the graph-cache. In fact GraphHopper should not start with a different configuration (error will be “Profiles do not match”).

If you need more flexibility you can use LM and apply the changes per request.

Very interested in the topic.

Have you got any article/documentation on that? With LM you mean Landmarks (A*)? If this is so, does it mean that with Dkjistra you cannot pass the profile to the request?

You can read here and here a bit more about it.

Basically A* is a goal-directed Dijkstra and Landmarks (LM) is an A* with a different goal-directed heuristic. But LM and CH require prepared memory- (or disk-)intense datastructures. But LM allows certain changes per request where CH does not allow this.

Without CH and LM there is no restriction and you can modify the custom model like you want (and GraphHopper supports).