[mkgmap-dev] actual splitter branch

Thu Nov 4 10:07:25 GMT 2010

Hi Steve,

>There was no license indication on the osmpbf files last time I looked, 
>but mainly I didn't check anything in because I just bodged it to get 
>it to build and it needs sorting out properly.

Maybe the easy way would be to ask the Osmosis developers, because they 
are bundling some of the jars.

>> By the way, I figured out that by carefully adjusting the tile 
>> borders, it is not necessary to filter the input (move the coastline 
>> endpoints). This approach also creates less "imaginary sea" around 
>> the map tiles. Polygon-shaped tiles would solve this problem in an 
>> even more elegant way.
>
>That sounds very interesting, is it something you do (or could do) 
>automatically or are you adjusting by hand?

I did it by hand, but I guess that it could be assisted with a GUI, 
e.g., extracting a natural=coastline extract of the map extract with 
Osmosis and then drawing tile rectangles (using appropriate 
snap-to-grid) on top of that data. Maybe there should be a JOSM plugin 
for editing areas.list? (If the alignment of the coordinates were not 
that important, a converter from a subset of OSM XML to areas.list could 
do the trick.)

If you look at the map of Finland, you can see that the south coast is 
sloping a little in the NE direction. Thus, if the southern (or 
southeastern) tile has its NE corner near where the coastline extract 
stops, you will have a nice sea in that tile. The tile to the north from 
that must extend further to the east, since the land border is sloping 
to the NE direction.

In the north, it was a little tricky. I cut the Northern Finland tile so 
that it takes a little bit of sea in the south. In this way, its 
southern neighbour tile had a continuous coastline extending all way to 
the tile borders. I tried to split two tiny pieces from the northern 
tile to silence a warning and fix the flooding of an island, but in the 
end I enabled generate-sea: multipolygon,extend-sea-sectors for the 
northern tile. That apparently extended the abruptly ending west end of 
the coastline straight to south, to the tile border.

I will update the files on http://www.polkupyoraily.net/osm/ hopefully 
tomorrow, if the data is OK. Someone had broken some oneway streets and 
a multipolygon in today's extract.

	Marko