Making Maps Navigable - Programming GPS and OpenStreetMap Applications with Java - page 105

files without loss. This file can be used to fine-tune the compiler tools to
the underlying hardware.
The first step, to remove all relations in the file, can be used as a bench-
mark for the compiler and hardware. Osmosis requires more RAM than the
OSMparser, and a choice should be made according to the platform used to
get an idea of the maximum duration of a single conversion step. Note that
the file extensions are hard-coded in the scripts and should follow a general
naming schema. (This way the file names listed in the log files speak for
themselves.) In case a project does require relations, this long process with
little data reduction (1 - 2%) can be skipped. Then, the hard-coded file
names in the batch *.norel.osm should also be replaced with *.clean.osm .
Subfolders are hard-coded and their creation ( mkdir ) can be added.
Network layers. The three layers
highway.motorway waterway.river railway.rail
represent three different networks. The extraction of each network takes
about 25 minutes 2 with the file sizes of 32 MB for highways, 67 MB for
waterways, and 92 MB for railways. These networks were chosen since
they are completely independent of each other and don't overlap. Cars,
ships, and trains don't interfere with each other, while commercial vehicles,
passenger cars, and pedestrians do share parts of one network.
Cartographic layers. The GER project aims to produce a map of Germany
with a pixel size close to a typical computer screen. In order to get some
green and blue color on the map (background), some landuse and natural
features are extracted:
landuse: forest farm residential industrial
natural: wood water
Looking at the image of the Germany map for a computer screen, it is
obvious that the forest polygons smaller than a few pixels could be removed.
Nevertheless, this process is not straightforward with the tools we have
introduced. The process would require a tool with spatial functionality
to calculate the actual size of a polygon and remove it below a certain
threshold. Also, the number of nodes could be significantly reduced without
changing the low-resolution shapes.
The OSMparser could be improved to detect polygons, using methods to
match first and last node. Then, the parser could dispose of all polygons
that have a small number of nodes and every second node of the large
polygons, etc., rerun as needed.
2 Find the exact figures in germany2010.log .