[mkgmap-dev] problematic_polygons file

Tue Oct 23 09:32:07 BST 2012

Am 23.10.2012 09:59, schrieb Gerd Petermann:
> Hi Carlos,
>
>
> > I've seen problematic_polygons file in growing quite fast in the wiki.
> > May it affect splitter performance? If so, would it make sense to split
> > the file by continents or countries?
>
> regarding performance:
> a) if you specify the --problem-file parm, splitter has to do a lot 
> more work, so that affects performance, no matter how many ids you put 
> ino your list.
> b) I don't expect a measurable performance impact for any id that does 
> NOT occur in your input OSM file(s) as long as this list doesn't contain
> millions of ids. The information is stored in HashMaps, so the only 
> negative impact is a higher possibility of hash collisions and a 
> slightly higher
> memory usage for these HashMaps. Most of the additional time that is 
> required to handle the list is caused by the fact that splitter has to 
> read the input
> file more often (3 times). With o5m and pbf format, only parts of the 
> file(s) are read, with XML input the complete file is processed.
>
> Of course, those ids that occor in your input data will require more 
> heap and probably produce more output data, so that will affect 
> performance.
>
> OK?
>
> The bigger problem that I see is this:
> The list now contains some relations like
> rel:52822 # Border Sweden
> rel:1059668 # Border Norway
> If you use the complete list to split e.g. finland.osm.pbf, the input 
> file will only contain parts of the needed data for these polygons.
> I wonder what splitter should do in these cases?
> Currently it might print a few messages regarding missing nodes or 
> ways, but it will use the incomplete data and it might
> write the incomplete relation to more tiles than r202, thus mgkmap 
> might produce even more error messages.
>
> I think it would be better to change this so that splitter returns to 
> the default handling for every relation or way that is not
> complete, with a corresponding message.
>
> Would that be better?
I think splitter should write everything he could find in input-data to 
each tile. The intention is a "complete" boundary. Of course if you only 
create a map of Finland, border of Sweden would only partly in the map, 
but these lines shouldn't be broken.
But is this a new problem? I think the relation of eg. Sweden would be 
written to a tile containing a part of Swedish boundary and there are 
also missing ways and nodes.

@Carlos: Maybe it would be better to separate the wiki-list in several 
parts like ferry, boundary, lakes, forrest. So it is easier to find out, 
if the ID is already in the list.

Henning
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mkgmap.org.uk/pipermail/mkgmap-dev/attachments/20121023/79d456bb/attachment.html 