Hi Gerd, while playing a little bit with splitter-r269 and bounding-polygons I found something strange with my scandinavia-polygon. If you compare scandinavia.poly and scandinavia.kml, the tiles are around iceland a little bit larger than bounding-polygon. If you open both files in josm and also bing sat-image, you'll see, that in the overlapping parts typical shouldn't be any data. Is it a bug in your algorithm or is it in the limit? If second, it wont be a problem at all. files: http://www.aighes.de/data/scandinavia.zip Henning Am 21.12.2012 00:02, schrieb GerdP:
Hi all,
I've committed r269 in the problem-list branch. It contains the changes proposed by WanMil and a few optimizations regarding throughput and memory needs. It is now possible to split europe on a 4GB machine with 64 bit JVM with --keep-complete and --max-areas=2048, no matter what output format you use. I've changed the default "batch_limit" in the pbf writer from 4000 to 1000. This increases the output file size by 2-3%, but it reduces the heap requirements with output=pbf drastically if you have many areas. The limit of 4000 was ok when splitter wasn't able to write more than 100 files at a time.