[GraphHopper] Disconnected areas

Ludwig ludwigbrinckmann at gmail.com
Mon Sep 23 18:25:11 UTC 2013


Hi Peter,

fiddling with that parameter seems to do the trick. I managed to get
routing to work for Naxos with a value of 500, but had to drop this further
to 200 to get routing on some even smaller islands. I have not had time to
really find the minimum value for this particular constellation.

For me at this point the question is what the motivation is to exlude
certain areas at all. Even if this was a private area (behind a barrier)
why exclude it from roiuting?

Regards

Ludwig


On 23 September 2013 16:44, Peter K <peathal at yahoo.de> wrote:

>  Hi Ludwig,
>
> probably we are a bit too aggressive with removing subareas! Ie. it could
> be that the island is 'too small' for graphhopper and will be removed (as
> graphhopper thinks this is a subnetwork hidden after e.g. barriers etc).
> Please change PrepareRoutingSubnetworks.minNetworkSize to something smaller
> than 3000 like 500 and let me know if this helps.
>
> Kind Regards,
> Peter.
>
>     Thanks for your quick reply, I only had time to look into this again
> now.
>
>  When I read your reply, I assumed that maybe the points I chose for
> illustration of the problem were too close together or maybe on the same
> road, but with my larger area I cannot get any routing to work on Naxos and
> I do get the same errors on the graphhopper website, even if I choose
> distant points clearly on different roads.
>
>  However, if I just choose a smaller area to create the routing data, then
> the routing starts working. There is no other change to anything.
> Boundaries of smaller area as given in original mail.
>
>  Example: 37.120359,25.410004 to 37.1,25.38 does not work on the extended
> area (and does not work on the website), but for the smaller routing area
> it works. See screenshot.
>
>  I have not made any effort to understand how the routing works, but my
> hunch was that it is disconnected areas (in this case multiple islands)
> that cause the problem. Is that even remotely possible?
>
>
>
>
>
>
>
> On 15 September 2013 03:17, Peter K <peathal at yahoo.de> wrote:
>
>>  Hi Ludwig,
>>
>> it looks like GraphHopper Maps has a similar problem (I tried your
>> coordinates<http://graphhopper.com/maps/?point=37.169792%2C25.449737&point=37.101957%2C25.375918>).
>> Currently the routing goes only from junction to junction and so the second
>> point resolves to the same junction. Will be fixed (in the very near
>> future) via #27
>>
>> Not sure why there is a difference for the "two" areas. It would be
>> interesting for me to know where the second point resolves to for your
>> unexpanded area and how the route looks like (screenshot?).
>>
>> Regards,
>> Peter.
>>
>>
>>     Following problem: I create a graphhopper app for one of the Greek
>> Islands and it works. Now I expand the area to include other islands and
>> the same app (no code changes) does not find the route (start and end point
>> are the same as previously).
>>
>>  To me this suggests that once I include a wider area that is not
>> connected to a subarea, it somehow confuses the routing. The exception is
>> an IllegalArgumentException that says it cannot find point1 and cannot find
>> point2 (then it also says the two are the same, I guess they are internally
>> nulled after being unable to find them.).
>>
>>  The bounds for my OSM extraction are as follows:
>>
>> #Naxos only: works
>> export LEFT=25.3
>> export RIGHT=25.7
>> export TOP=37.3
>> export BOTTOM=36.9
>>
>>  #Wider Aegais: does not work
>> export LEFT=24
>> export RIGHT=28.5
>> export TOP=38
>> export BOTTOM=35.2
>>
>>  # Slightly more verbose error message from the error stack:
>>  Error:[java.lang.IllegalArgumentException: Cannot find point 1:
>> 37.16979166666667, 25.449736666666663, java.lang.IllegalArgumentException:
>> Cannot find point 2: 37.101957, 25.375918,
>> java.lang.IllegalArgumentException: Point 1 is equal to point 2]
>>
>>  Any ideas why this is so?
>>
>>  Ludwig
>>
>>
>
> _______________________________________________
> GraphHopper mailing list
> GraphHopper at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/graphhopper
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/graphhopper/attachments/20130924/14495b71/attachment.html>


More information about the GraphHopper mailing list