[mkgmap-dev] nearby POIs

Wed May 6 07:44:24 BST 2020

Hi all,

thinking again about "Known problems:  The new implementation will not work well for the /all case because a group with different names is not detected as such"
it is probably not a problem of the algo but of the docu.
I can't think of any reason to remove a POI with a given type x named "abc" when it is close to another POI with the same type named "xyz" or when one of them is unnamed, the implementation is probably exactly what is wanted.
So, I think I should simply add the mentioned grouping by type & name to the documentaion?
I think we should define what happens when multiple rules match the same type.
These 3 rules all match for a POI 0x3205 with name "Test" but it is not obvious which one will be used.
0x3205:30:merge-at-mid-point
0x3203-0x3207/named:20:delete-name
0x3200-0x331f:50

With the current code the last one is used because of some internal sorting, probably not what was intended by the user.
I see no obvious way to handle this. What should happen?

BTW: I didn't try but I think default rules (*) with a radius > 5m are quite dangerous. Traffic lights or barriers may disappear. Also when you reander natural=tree
POI you probably don't want to remove most of them. So, if you use them make sure that you specify rules for everything that should not be removed.

Gerd

________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Gerd Petermann <GPetermann_muenchen at hotmail.com>
Gesendet: Mittwoch, 6. Mai 2020 07:04
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] nearby POIs

Hi all,

thanks for testing. I've just noticed that the jar file contained a modified points file, but I guess you all used your own style files.
I thought about the possible effects on routing and found none.

The new algo works like this:
1) Collect all generated POI and the corresponding nodes before adding them to the map
2) after all style rules were processed and WrongAngleFixer has moved points (and POI) to fix visible wrong angles the POI are deduplicated.
This means they are
 - grouped by type and name
 - each group is processed so that
   + the first matching rule is applied using the given radius: Iterate as long as there are multiple POI
     * for each POI, build a group of all nearby POI
     * select the biggest group
     * calculate the center of that group
     * select the POI that is closest to the center
     * apply the action to the other POI of the group, maybe move the selected POI to the middle
     * remove all POI handled in this group

Known problems:
- The new implementation will not work well for the /all case because a group with different names is not detected as such
 - maybe identical POI are not removed with the delete-name case
- Maybe there is an unwanted effect on rendering when WrongAngleFixer moves a point and the corresponding POI is not moved. This already happens sometimes and should be no big problem.

If I find no solution for the first 2 problems today I'll revert trunk version and create a branch for further development.

Gerd

________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Ticker Berkin <rwb-mkgmap at jagit.co.uk>
Gesendet: Dienstag, 5. Mai 2020 12:03
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] nearby POIs

Hi Gerd

That got rid of my multiple benches.

Just looking at the log for the few other POI it also got rid of.

It would be useful for the log message to include the distance between
the POIs

Ticker

On Tue, 2020-05-05 at 09:19 +0000, Gerd Petermann wrote:
> Hi all,
>
> attached is a patch that seems to work. I've kept the logging as
> memory is not such a big problem.
> I've not yet tested special cases like duplicated barrier nodes or a
> group of nodes where only some are within the tile boundary.
>
> Please check if your problems are solved with this version, a binary
> is here:
> http://files.mkgmap.org.uk/download/469/mkgmap.jar
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag
> von Joris Bo <jorisbo at hotmail.com>
> Gesendet: Dienstag, 5. Mai 2020 08:32
> An: Development list for mkgmap
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi,
>
>
>
> I have been testing the nearby poi rules as well with 4485
>
> A while I struggled thinking misconfiguration because I see nothing
> happening but then after extending the ranges, an occasional poi
> disappears. (toilet 0x2a0d, busstop 0x2f17, custom lift-gate 0x6905)
>
>
>
> On the cmd line
>
> --nearby-poi-rules=*/unnamed:100:delete-poi,*/named:100:delete-poi
>
>
>
> On my densed big city area (Antwerpen) In that case I expect maybe a
> couple of thousand benches, trees and bicycle parkings disappear
> which do not.
>
>
>
> Can I help testing better by increasing some logging? How could I do
> that?
>
>
>
> Kind regards, joris
>
>
>
> Silo’s (0x2a19) don’t disappear / N51° 11.992' E4° 24.916'
>
> An occasional lift gate (0x6905) does disappear
>
> [cid:image001.jpg at 01D622B7.BA7DE260]
>
>
>
> Bicycle stands (0x2f18): don’t disappear
>
> N51° 13.685' E4° 24.281'
>
> [cid:image002.jpg at 01D622B7.BA7DE260]
>
>
>
>
>
> -----Oorspronkelijk bericht-----
> Van: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> Namens
> Ticker Berkin
> Verzonden: maandag 4 mei 2020 16:13
> Aan: mkgmap development <mkgmap-dev at lists.mkgmap.org.uk>
> Onderwerp: Re: [mkgmap-dev] nearby POIs
>
>
>
> Hi Mike
>
>
>
> No, I just have --nearby-poi-rules=0x6605:30
>
>
>
> Gerd has explained why I see a message from NearbyPoiHandler about a
> POI for that wasn't 0x6605 - the logging of the removal of an exact
> duplicate has moved from StyledConverter
>
>
>
> Ticker
>
>
>
> On Mon, 2020-05-04 at 14:35 +0100, Mike Baggaley wrote:
>
> > Hi Ticker, have you got a --nearby-poi-rules-config option as well?
>
> >
>
> > Cheers,
>
> > Mike
>
> >
>
> > -----Original Message-----
>
> > From: Ticker Berkin [mailto:rwb-mkgmap at jagit.co.uk]
>
> > Sent: 04 May 2020 11:33
>
> > To: Development list for mkgmap <mkgmap-dev at lists.mkgmap.org.uk<mai
> > lto:mkgmap-dev at lists.mkgmap.org.uk>>
>
> > Subject: Re: [mkgmap-dev] nearby POIs
>
> >
>
> > OK, looking up the # as a way instead gives a polygon that matches
> > the
>
> > description and would generate a typ code as per the message
>
> >
>
> > I was more worried about it logging messages about POI with typ
> > codes
>
> > that I hadn't mentioned in rules and not saying anything about the
> > one
>
> > I had.
>
> >
>
> > I get the same behaviour with just --nearby-poi-rules=0x6605:30
>
> >
>
> > Ticker
>
> >
>
> > On Mon, 2020-05-04 at 10:14 +0000, Gerd Petermann wrote:
>
> > > Hi Ticker,
>
> > >
>
> > > the new option doesn't work well, Mike and I are working on it.
> > > See
>
> > > http://gis.19327.n8.nabble.com/nearby-POIs-tp5962630p5964759.html
>
> > > The log messages are for POIs generated from areas, so the
>
> > > "generated from" ids are either ways or multipolygon relations.
>
> > >
>
> > > Gerd
>
> > >
>
> > >
>
> > > ________________________________________
>
> > > Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk<mailto:
> > > mkgmap-dev-bounces at lists.mkgmap.org.uk>> im Auftrag
>
> > > von Ticker Berkin <rwb-mkgmap at jagit.co.uk<mailto:
> > > rwb-mkgmap at jagit.co.uk>>
>
> > > Gesendet: Montag, 4. Mai 2020 12:04
>
> > > An: mkgmap development
>
> > > Betreff: [mkgmap-dev] nearby POIs
>
> > >
>
> > > Hi Mike & Gerd
>
> > >
>
> > > I'm trying --nearby-poi-rules=0x6605:30:merge-at-mid-point
>
> > >
>
> > > with, in points:
>
> > >
>
> > > amenity=bench [0x6605 resolution 24 continue]
>
> > >
>
> > > and the resultant map is showing benches close to each other (<
> > > 10
>
> > > m)
>
> > >
>
> > > and, in the log file I'm getting messages like:
>
> > >
>
> > > INFO: uk.me.parabola.mkgmap.osmstyle.NearbyPoiHandler
>
> > >  ../mapHants/74210001.osm.pbf: Removed duplicate POI with type
>
> > > 0x2d02
>
> > > The Dungeon for element generated from 159255400
>
> > > INFO: uk.me.parabola.mkgmap.osmstyle.NearbyPoiHandler
>
> > >  ../mapHants/74210001.osm.pbf: Removed duplicate POI with type
>
> > > 0x2a07
>
> > > Wah Ying for element generated from 468906008
>
> > >
>
> > > The "generated from" number doesn't seem to correspond to a
> > > relevant
>
> > > OSM node and the message doesn't have the trailing "and moved..."
>
> > > that
>
> > > I would have expected.
>
> > >
>
> > > Ticker
>
> > > _______________________________________________
>
> > > mkgmap-dev mailing list
>
> > > mkgmap-dev at lists.mkgmap.org.uk<mailto:
> > > mkgmap-dev at lists.mkgmap.org.uk>
>
> > > http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>
> > > _______________________________________________
>
> > > mkgmap-dev mailing list
>
> > > mkgmap-dev at lists.mkgmap.org.uk<mailto:
> > > mkgmap-dev at lists.mkgmap.org.uk>
>
> > > http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>
> >
>
> >
>
> _______________________________________________
>
> mkgmap-dev mailing list
>
> mkgmap-dev at lists.mkgmap.org.uk<mailto:mkgmap-dev at lists.mkgmap.org.uk>
>
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev