When label does not fit in the center of area, render label off-center

[matkoniecz]

related to #957 but I am not sure is it doable

http://www.openstreetmap.org/?mlat=50.0917&mlon=19.9920#map=14/50.0917/19.9920

on zooming in label disappears

http://www.openstreetmap.org/way/54965742#map=16/50.0916/19.9924

[daganzdaanda]

In #478 (comment), Vincent mentioned an interesting talk:

Maxim Rylov, Cartographically Plausible Label Placement.
A Multi-criteria Model for Good Point-label Placement on OSM Maps
http://sotm-eu.org/en/slots/16

From the Slides:

The model can be reproduced in your own solution or in
Mapnik, GeoServer, Mapserver, etc.

A map with a sophisticated label placement for the whole globe
http://openmapsurfer.uni-hd.de (OSM Roads layer)

The proposed model is implemented within the MapSurfer.NET framework
http://openmapsurfer.uni-hd.de/Framework/info.html

The example map looks great (when you turn off the experimental boundaries). If you look at the Czech Republic, the "Cesko" moves south when "Praha" appears.
Let's hope Mapnik can adapt these algorithms, if it hasn't yet.

[dieterdreist]

Am 22.05.2015 um 22:33 schrieb daganzdaanda notifications@github.com:

A map with a sophisticated label placement for the whole globe
http://openmapsurfer.uni-hd.de (OSM Roads layer)

this is a improvement but still fails on Rome for certain zoom levels (disappears and reappears on zooming in)

[kocio-pl]

related to #957 but I am not sure is it doable

That would be possible with a placement type grid, if it was merged into the Mapnik and available in the Carto CSS - see #2962.

[kocio-pl]

Maxim Rylov, Cartographically Plausible Label Placement.

Interesting to compare the label placement density in OSM Carto/GMaps then (2014) and now (2017)...

Then - 5 vs 54:

Now - 9 vs 27:

[kocio-pl]

More about Google change in the number of labels:

https://www.justinobeirne.com/what-happened-to-google-maps

[jeisenbe]

Requires change to v5.x since we would need to require a new Mapnik version, correct?

[kocio-pl]

I think you're right.

[jeisenbe]

OK, that should happen fairly soon, so that the unpaved rendering PR can be deployed.