> It doesn't let you avoid a road or favour another
This might be a computational cost thing. Modern routing algorithms preprocess the graph to find the optimal route orders of magnitude faster than you normally would without preprocessing. If you let each person customize the graph then such queries would be much more costly.
Then again, I'm sure there are lots of other complications already and that most users wouldn't want alternate routes, so maybe the marginal cost here is irrelevant?
Hire some game devs that have worked on path finding in AAA games. They will find a way to partially pre process the map data in such a way that on-device computation can then let the user pick and choose what roads to favor and what roads to exclude.
And as an added bonus they’ll be unrelentingly optimizing your map application to the point where the UI will run at 144 Hz ;)
contraction hierarchies https://en.wikipedia.org/wiki/Contraction_hierarchies made cross-continent route planning practical and practically instant, eg in GraphHopper. A lot of the options listed there work offline.
Game routing is not remotely similar to real world routing. Neither in scale, nor in other aspects like road traffic or incorporation of human preferences.
In any case, I'm not suggesting it's impossible, just costlier than people imagine when they ask for the feature.
I wouldn't be so sure about that. I have seen many different users intentionally deviate from the suggested route because they know more than Google about what is best for them (or at least they think they do).
I wasn't saying this never happens. I myself deviated from Google Maps just yesterday. But, like, this was probably the only one out of my last 10 queries. And I was trying to say this is a minority of the queries.
Those overrides could be stored locally and the maps service could still return the same optimized graph. The client device can then swap out and override it and do the calculation on device.
This might be a computational cost thing. Modern routing algorithms preprocess the graph to find the optimal route orders of magnitude faster than you normally would without preprocessing. If you let each person customize the graph then such queries would be much more costly.
Then again, I'm sure there are lots of other complications already and that most users wouldn't want alternate routes, so maybe the marginal cost here is irrelevant?