[mkgmap-dev] Re : --generate-sea for Australia?

Sat Feb 6 12:24:44 GMT 2010

> Hello Ben,
> I sent the same remark for Ireland last night (mkgmap-dev, Digest vol
> 19, issue 11). The latest version I found which is able to generate sea
> mutlipolygon is 1512. Your data should be sufficient.
> I hope this will help you agaisnt global warming.
>
> Regards,
> David

Hello Ben, hello David,

I don't know the generate-sea code exactly but I am the developer of the 
multipolygon code, which has to deal with the data generated by the 
generate-sea-code.

A lot of work has been done on the multipolygon and generate-sea part of 
mkgmap in the last weeks. But mkgmap is still very dependent on the 
quality of the input data. We have analysed the quality of the Geofabrik 
dumps. They have a big disadvantage: The boundary shapes of the 
Geofabrik dumps sometimes are too short, so that some ways tagged with 
natural=coastline are not contained in the dump and mkgmap does not have 
the extact boundary shape to substitute the holes. This problem could be 
resolved in the osmosis software used by Geofabrik but at the moment the 
developers do not have time for it (see 
http://lists.openstreetmap.org/pipermail/osmosis-dev/2010-January/000452.html). 
If you want to speed up a solution for this please answer to the given 
thread.

If the quality of the input data does not fit to a minimum standard 
mkgmap must guess a lot of things which randomizes the results. We will 
continue to improve mkgmaps guess capabilites but we cannot propose the 
lottery numbers...

If we should analyze in detail what fails for you data we need the 
following information:
* Which dump do you use (Geofabrik, Planet dump etc.) and the day of the 
dump
* Your splitter settings (the areas.list file, splitter version and 
splitter parameters)
* Your mkgmap style file
* Your mkgmap parameters

Sorry that I cannot give you a simple solution but as long as things are 
complicated their solutions also tend to be complicated. Maybe Mark can 
give you more information as he has done some work on the generate-sea-code.

WanMil