[GraphHopper] Distances are 2x wrong?

[Peter K]

Hi Jaak,

good you found the issue!

Regarding the maven repo: that is a good idea!
Pushing to maven central is another (complicated) option or to bintray.com:
https://github.com/JFrogDev/bintray-examples/blob/master/maven-example/pom.xml

Deploying it automatically (server side git push hook) via our
travis-ci.org <https://travis-ci.org/graphhopper/graphhopper> would be
even cooler as I don't like to maintain anything ;)

But currently I'm heavily focused to push 0.1 out ...

Regards,
Peter.

> Hi,
>  regarding 1) . Maybe I used somehow older version of lib. Now I tried
> with Maven dependencies and it seems to get correct results, like
> in https://dl.dropbox.com/u/3573333/public_web/graphhopper_sample1.png
>
>  ps. I uploaded the graphhopper-1.0-SNAPSHOT to my Maven repo
> (https://repository-nutiteq.forge.cloudbees.com/release/), is there
> better place to fetch it from?
>
> Jaak
>
> On 04.04.2013, at 19:06, Peter K wrote:
>
>> Hi Jaak,
>>
>> cool, this is really great news! As I already wanted to look into
>> this, but hadn't the time so far.
>>
>> Regarding 1) The returned value is as expected (normal) meters.
>> Hopefully I can digg into the code next week.
>>
>> Regarding 2) ghz is a zipped folder and can be opened via the
>> Helper.unzip
>>
>> > mapsforge library
>>
>> Oh, so the mapsforge data can be opened? Nice!
>>
>> Regarding 3) This will be solved in issue-17. It is the problem of
>> mapping from "real world gps coordinates" to graph node ids.
>> You can partially work around this by increasing the precision of
>> Location2IDQuadtree index
>> or checkout the master (not really recommended yet, as incompatible
>> file layout and probably other minor issues)
>>
>> Regards,
>> Peter.
>>
>>
>>> Hello,
>>>
>>>  I just pushed a sample with Graphhopper routing to Nutiteq 3D SDK
>>> AdvancedMap3D samples: https://github.com/nutiteq/hellomap3d/,
>>> see GraphhopperRouteActivity.java . Nutiteq SDK is quite similar to
>>> Mapsforge, just with 2.5D perspective and rotation capabilities. And
>>> with a lot of different data source Layer implementations, from
>>> geotiff and shapefiles to mapsforge library.
>>>
>>>  Issues with grasshopper:
>>>
>>> 1) major problem - all the routes in my area (Estonia) are a bit
>>> more than 2x too short. Instead of 100 km I have 45 or so. Could it
>>> be that it is so much north and calculations of distances become
>>> wrong? If the "meter" is really "mercator meter" then the value
>>> could be even right, but calculation of mercator meters to real
>>> meters, when the route is from south to north, is not trivial (or
>>> not possible at all). I created test dataset for Baltics 3
>>> countries: https://dl.dropbox.com/u/3573333/baltics.ghz , where it
>>> is observable. 
>>>
>>> 2) some minor issues. I did not figure out how to open properly .ghz
>>> file with the API. So the sample asks now for .map file inside
>>> <map>_gh folder, so the files must be uncompressed in SDCARD. But it
>>> would be nicer to enable just to open ghz file, and GH would
>>> uncompress it automatically.
>>>
>>> 3) start and endpoints are sometimes not optimal, quite far from
>>> clicked location, and there seem to be much closer nodes to given
>>> points. I added lines from clicked location to first and from last
>>> location to make it nicer. Maybe some tolerance in graph generation,
>>> or in routing?
>>>
>>>
>>> Your pull requests with fixes to the demo app are very welcome,
>>> quite possibly I just did something wrong
>>>
>>> Jaak
>>> Nutiteq.com <http://Nutiteq.com/>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/graphhopper/attachments/20130405/b643ca3e/attachment-0001.html>