Michael Zilske
Michael Zilske
One benefit of edge attributes separate from topology would be that it would allow different attributes for different types of edges without wasting memory, like a `union`. Very useful for...
> reading from a large int array was slightly, but predictable slower Hmm. Now I played myself, and I can't reproduce that. This is what I get (measuring throughput in...
Repeating, going 10 seconds per trial (coming closer to covering the entire data set while reading), and putting `MMapDataAccess` into the mix: ``` REMEMBER: The numbers below are just data....
Is that a problem if this is done for all cases and I'm interested in relative performance?
> Would it help you to be able to store a byte array per edge? OMG yes. Except that I won't be the only one. Meaning, very soon, n people...
I think this is because when a virtual _destination_ node cannot be reached via its _closest_ tower node, but only via a _specific_ one (enforced by the curbside-aware algorithm, but...
From my point of view, we should even narrow our current scope. Anything that could be solved by pre-processing the OSM data before it hits GraphHopper perhaps *should* be solved...
To bluntly manage expectations, I don’t think it’s likely we will merge this unless we would use it ourselves. (Either of it.)
Looking forward to this, as this has applications to everything from map-matching to tour planning.
*How interesting*, thank you! To me, this looks like he should try map-matching rather than via-routing. I think the problem is solved there.