When we call graphHopper.load with a wrong path that doesn't exist, the process informs us with an exception about that, but also creates unnecessarily the path.
That's inconvenient as we have to remove the unneeded created empty folder afterwards.
Can we skip the GHDirectory.mkdirs in case the graph loading has failed?
Sure, that is ugly.
Should be fixed with #700
Found a way to fix this and keep backward compatibility: https://github.com/graphhopper/graphhopper/commit/0de10bfd2ba7bd33c7e9980a73ad00b66b7a6898
More complex than #700 :(. See #724 for discussion why this is still necessary.
Nice! :+1:
Also the needed removal of the mkdirs in GHDirectory constructor.
Most helpful comment
https://github.com/graphhopper/graphhopper/pull/700