Bug 253486

Summary: Marble changes the inputted streets when trying to get directions
Product: [Applications] marble Reporter: Riccardo Iaconelli <riccardo>
Component: generalAssignee: marble-bugs
Status: RESOLVED UPSTREAM    
Severity: normal CC: nienhueser, wocserve
Priority: NOR    
Version: unspecified   
Target Milestone: 1.0 (KDE 4.6)   
Platform: Fedora RPMs   
OS: Linux   
Latest Commit: Version Fixed In:
Sentry Crash Report:

Description Riccardo Iaconelli 2010-10-07 10:34:17 UTC 
Version:           unspecified (using KDE 4.5.1) 
OS:                Linux

I tried to enter a navigation from "Via Stilicone, Milano" to "Via della Novella, Nerviano". While the second street is correctly set as destination, the first is changed for no explicable reason to "Via della Simonetta, Milano", which is a street very close to it, but not Via Stilicone.

Also the CAP is wrong (20158, instead of 20154) but I think openstreetmaps is to blame there... right?

Reproducible: Always

Steps to Reproduce:
- Enter "Via Stilicone, Milano" as the "A" point.
- Wait for marble to compute the full name (which is: "Via Stilicone, Bovisa, Milano, 20158, Lombardia, Italy")
- Select it

Actual Results:  
The street in the line edit becomes "Via della Simonetta, Bovisa, Milano, 20158, Lombardia, Italy"

Expected Results:  
The street in the line edit should be "Via Stilicone, Bovisa, Milano, 20158, Lombardia, Italy", as selected.

OS: Linux (i686) release 2.6.33.3-85.fc13.i686
Compiler: gcc
Comment 1 Dennis Nienhüser 2010-10-13 20:43:19 UTC 
Via Domenico Cucchiari, Via della Simonetta and Via Stilicone share a common waypoint (the intersection of the three), which is also the closest waypoint to the point you are selecting. My guess what is happening is follows: Nominatim (the search engine we use) searches its addresses and finds Via Stilicone with the source location you are selecting. When you select it, nominatim (now as a reverse geocoder) is queried again, now only given the geolocation. Possibly it only looks at the closest waypoint and selects any of the streets it belongs to instead of taking their directions into account, which would be more correct.

Now there are two possible solutions to improve this:
First, in Marble, I could special-case this and prevent the reverse geocoding call since we have the address information already from the previous search.
Second, in Nominatim, the algorithm could be improved to analyze road directions for nodes shared among more than one road.

The second one is better since it solves the general problem.

Maybe you can forward the problem to the nominatim author [1] and I look into the Marble solution.

[1] http://wiki.openstreetmap.org/wiki/Nominatim#Bugs_.2F_Error_reporting
Comment 2 Simeon 2011-12-18 21:50:58 UTC 
This bug can no longer be reproduced.Report should be closed.