[OSM-dev] OSM History Splitter

[Peter Körner]

Hi devs,

some months ago I announced my History Splitter and the availability of 
extracts. In the last months I've completed my work and implemented a 
second cutting-algorithm. While the --hardcut algorithm will stay the 
default as a, the --softcut algorithm features the following attributes:

- version-completeness (if one version is inside the bbox, all versions 
are included in the extract)
- way-completeness (ways are not cropped but included as they are in the 
database)
- way-reference-completeness (all nodes of ways touching the bbox are 
included)
- relation-completeness (relations are not cropped, they may however 
have children which are not included inside the extract)
- cascading-relations (all super-relations of relations touching the 
bbox are included in the extract)

those features are implemented using a two-pass process. It runs 
single-threaded, so to use multiple cores you are required to run 
multiple independent instances of the program. A python script that is 
able to control this parallelization is included, however multiple 
simultaneous splits increases the number of disk-seeks, which possibly 
outruns the speedup gained through distributing the CPU-Load to multiple 
cores.

The tool is, without changes, suitable for normal planet-extractions, 
too. By choosing the sequence and parallelization carefully, it's 
possible to accomplish the specific task of extracting a bbox or a 
polygon in 1/3 of the time required by osmosis.

The source and building instructions are available at github:
https://github.com/MaZderMind/osm-history-splitter

If you have bug reports or problems compiling or running the splitter, 
don't hesitate to mail me directly at osm at mazdermind.de.

Peter