[OSM-talk] the coast and colour of the sea

[Jon Burgess]

On Wed, 2007-10-31 at 19:38 +0000, Andy Allan wrote:
> On 10/31/07, Jon Burgess <jburgess777 at googlemail.com> wrote:
> > On Wed, 2007-10-31 at 19:32 +0100, Martijn van Oosterhout wrote:
> > > On 10/31/07, David Groom <reviews at pacific-rim.net> wrote:
> > > > I belive Artem had some tools which took the OSM natural=coastline data and
> > > > converted this to the files needed for the sea layer in Mapnik.  This is why
> > > > if you look at an area around
> > >
> > > I could really use such a tool right now, any ideas where I might find
> > > it? Otherwise I'll probably start writing one...
> > >
> > > Have a nice day,
> >
> > We don't have such a tool.
> 
> So how did Artem do it originally for the english coast? I'd really
> like to have a go at extending the shapefiles up the west coast of
> Scotland, so that my old hometown looks better.

I'm not exactly certain. I believe Artem started by import either the
PGS coastline data or OSM coastline ways for the UK. He then edited this
with Jump (see below) and made sure that the coastline formed a closed
way (many of the OSM and PGS lines have small gaps, especially around
the coast of Scotland!). 

You probably want to start by importing the shapefile which is currently
used for the coastline polygons:
http://tile.openstreetmap.org/shoreline_a.tgz (3MB)

The complete set of shapefiles we use is much larger ~600MB
http://artem.dev.openstreetmap.org/files/world_boundaries.tar.bz2

The majority of this is a single shapefile called shoreline_l which I
believe is a PGS import for the whole world. This is the source of the
thin blue line shown at high zooms. You can use this as a reference to
add more coastline to the shoreline_a file.

The shapefile can be edited in a tool like Jump. I used the one from
Vivid solutions http://www.vividsolutions.com/jump/ 
I see there is a newer version here: http://openjump.org/ but I have not
tried this.

I'd be happy to take an improved shapefile. There are a couple of things
which discourage anyone from spending much time updating them manually:-

- If lots of people want to make update then it will be difficult to
merge the changes together. 

- At some point we will probably replace the shapefiles with something
auto-generated from the natural=coastline ways from the main DB. I think
it would be better if we spent the effort on writing such an
auto-conversion tool.

- There is also an outstanding question as to whether the shapefiles are
in the correct projection. I know the projection string we use currently
("+proj=merc +datum=WGS84  +k=1.0 +units=m +over +no_defs") is wrong and
at some point we probably need to regenerate the shapefiles in the
correct projection (I've mentioned this on the dev list previously). The
error cancels itself out over the complete rendering pipeline but makes
it tricky to use the Postgres DB or shapefiles for anything accurate.
The fix requires updates to practically the entire set of mapnik tools
(osm2pgsql, these shapefiles, re-import the planet data, update the
render_from_list.py and probably a couple more scripts I can't
remember). 

  Jon