graphhopper

Coincidentally a few days ago, also for a sorting algorithm, I stumbled over a situation where replacing the branchless cmov with branching instructions actually made the code 30% faster: https://github.com/graphhopper/graphhopper/pull/3380 Or at least the AI explained it this way to me, but I'm unsure if this is correct.

I'm one of the maintainers of the open source routing engine GraphHopper. The project is around 6k stars so not that small and also we have private repositories and pay Github a monthly fee for this. We recently found that one of our longer issues disappeared: https://github.com/graphhopper/graphhopper/issues/446 See this archived version: https://web.archive.org/web/20241112051113/https://github.co... I contacted Github and they wrote: I can confirm that the issue isn't a technical problem on GitHub's side, but is related to the user status of the creator of this issue. For privacy reasons, I cannot share additional details without permission from the user in question, though. I ask them why they don't handle it like reddit does (just removing the comment content and the user name), but they say they can't do this: Unfortunately this is

The backend at this point is just Graphhopper so I think it would make more sense to use it directly and customize it based on your needs: https://github.com/graphhopper/graphhopper

Thanks! I think he meant that we dropped the explicit support of Android and the example app with offline maps (Mapsforge), see this issue: https://github.com/graphhopper/graphhopper/issues/1940 But offline routing should still work on Android. But as you need JDK 1.8 support this will exclude a few older Android versions and devices, I think.

I did a similar visualization but for road network paths: uni-directional A* https://karussell.files.wordpress.com/2012/07/astar.gif (New version: https://www.graphhopper.com/wp-content/uploads/2017/07/astar...) bi-directional Dijkstra https://karussell.files.wordpress.com/2012/06/bidijkstra.gif This uses a primitive Swing UI https://github.com/graphhopper/graphhopper/blob/master/tools... And it can be color encoded too https://www.graphhopper.com/blog/2016/01/19/alternative-road... which is inspired from others: https://www.graphhopper.com/blog/2015/12/14/roads-to-rome-fr... Or here I used Deck.gl with time-dependent data: https://www.graphhopper.com/blog/2018/07/04/high-precision-r...

OpenRouteService uses GraphHopper as library which uses exact routing algorithms: https://github.com/graphhopper/graphhopper (note: I'm one of the co-founders of GraphHopper) And then this A to B routing gives you the distance matrix which is necessary as input for vehicle routing algorithms (usually heuristics), see https://www.graphhopper.com/blog/2017/09/18/route-optimizati... for more details.

OsmAnd has a separate routing algorithm to my knowledge (was one of the first). It is definitely not GraphHopper as I created it later and prefer the routes it creates ;) but you can study our documentation to get a rough understanding how it all works: https://github.com/graphhopper/graphhopper/blob/master/docs/... and reach out in our forum for more details. The algorithm from maps.me / Organicmaps is ported from OSRM to Android and iOS. The CTO of maps.me that did this port left maps.me in 2017 but now seems to work on Organicmaps again: https://github.com/biodranik

We have a simple 4 wheeler config: https://github.com/graphhopper/graphhopper/blob/master/core/...

In a team at a hackathon I implemented a similar thing (safe way to school) with GraphHopper: https://github.com/karussell/nordhopper (Disclaimer: I'm one of the GraphHopper devs) There is no mobile app only a simple web app. But we also considered crash data from the German government which could be important. Additionally with GraphHopper walking routes avoid bigger streets and consider many signals from OpenStreetMap data. My personal route would probably just to avoid some dark parks to be honest as every detour is ugly. And so a nice feature is the customizable routing where you can change your preferences per request because it is not clear if a longer detour along lamps is really safer if it is twice the length and at which point your personal preferences are or how you wish to mix other signals into the cost function of the route.

For OSM-only data you could set up wheelchair routing with two commands (via GraphHopper) for the region of your choice: https://github.com/graphhopper/graphhopper#installation In the config you specify graph.flag_encoders: wheelchair and enable elevation: graph.elevation.provider: srtm It then considers elevation and many restrictions already but it is probably not as fine tuned as it should be. We would be happy to accept improvements.

Area routing is definitely a topic of interest. It looks like it's still a topic of discussion too: https://github.com/graphhopper/graphhopper/issues/82. Maybe it's something I can help with in the future. Good point RE the green index calculation. The very least I could do would be to upstream a reference implementation and documentation back to Openrouteservice. I will take care of that.

GraphHopper is already used for fire trucks and so called "isochrone calculations": https://github.com/graphhopper/graphhopper#for-analysis

Let me know if you have questions or want to try it out on your own. This work will be likely in version 1.0 of our open source routing engine: https://github.com/graphhopper/graphhopper