logo separator

[mkgmap-dev] [PATCH v4] sea polygons

From Robert Joop 8500547528183gmap at rainbow.in-berlin.de on Tue Aug 18 22:57:05 BST 2009

On 09-08-17 17:38:46 CEST, Clinton Gladstone wrote:
> On Sun, Aug 16, 2009 at 12:49 AM, Christian
> Gawron<christian.gawron at gmx.de> wrote:
> > With this version of the patch, the intersection of the bounding box and the
> > landmass does not have to be simply connected as in version 3 of the patch.
> Hm... I tried out this patch with a tiled map of Germany. The results
> were kind of amusing: pretty much all of Germany was flooded, with the
> exception of some areas around Lübeck. I suppose this is because I am
> using an extract (from Geofabrik) which cuts off coastline at the
> country border.

same here: different input, similar result.
using osmosis, i've cut a bbox from planet.osm, from northern germany
to croatia.
some observations:
- one danish island at the northern end of the bounding box has the bbox
  edge cutting through it.
  but what is displayed of the island does not reach to the top of the
  bbox, instead there's a strip of sea, some 600 m wide (which is fake,
  in reality there's (is)land cover).
- its sister island, equally cut through by the bbox edge, is flooded.
- main-land reaching into the bbox from the left exhibits the same as
  the first island, i.e. has a fake strip of sea, some 900 m wide.
- a number of islands seem to come out perfectly, land surrounded by sea.
- even with most of germany flooded, there's a strip at the coast which is
  dry--interestingly the dividing line through the land is not exactly
  east-west but goes from about N54.03 E11.6 to N54.16 E13.3.
  following this line further to the west, land and sea are sometimes
  inverse, i.e. the sea is shown as land, the land as sea.
  it toggles where this line and the coastline cross.
- one tile (northern germany/poland at the coast) is fine, but the tiles
  further inland are flooded...

there were nine warnings of the form

SEVERE (Osm5XmlHandler): Non-closed coastline segment does not hit bounding box - expect strange results: 25010841 (54.17039/13.35783) 21906107 (53.99328/11.00206) http://www.openstreetmap.org/?lat=54.17039&lon=13.35783&zoom=17

i'm wondering how this can be fixed, by
- different use of osmosis (unlikely as some warnings come from
  inside the bbox?)
- changing the splitter (some problems seem to be tile related?)
- changing the sea patch
- fixing osm data--some problems seem to be at neither bbox nor tile edge,
  for example lastovo and its sister islands are flooded

thank you,

More information about the mkgmap-dev mailing list