W> ok, I understand. W> For me it's no problem if points that fall outside the bounding box W> are included in the tile. They can be easily filtered. W> I have the following requirements: W> 1st and most important: All nodes that fall inside the bounding box W> should reliably be contained in the tile. In my previous mail I W> mentioned two nodes where this wasn't the case.
Ah, I hadn't realised this from your original description, apologies for missing that. Indeed, this should not be happening. Would it be possible for you to make the OSM file you're using available, or send me a link to it? It's not obvious to me why nodes are being excluded so having the same OSM file as you will make debugging much easier.
Aah, I don't have any server where I can put the files. I have retried the splitting. http://download.geofabrik.de/osm/europe/austria.osm.bz2 from today (Jan 13th) and that gave the same results. This is my splitter call (v103):
java -Xmx950m -jar splitter.jar data/*.osm.bz2
These are the generated areas: # List of areas # Generated Wed Jan 13 19:44:27 CET 2010 # 63240001: 2174976,442368 to 2228224,581632 # : 46.669922,9.492188 to 47.812500,12.480469 63240002: 2164736,581632 to 2213888,659456 # : 46.450195,12.480469 to 47.504883,14.150391 63240003: 2213888,581632 to 2275328,659456 # : 47.504883,12.480469 to 48.823242,14.150391 63240004: 2160640,659456 to 2226176,798720 # : 46.362305,14.150391 to 47.768555,17.138672 63240005: 2226176,659456 to 2287616,724992 # : 47.768555,14.150391 to 49.086914,15.556641 63240006: 2226176,724992 to 2283520,802816 # : 47.768555,15.556641 to 48.999023,17.226563 The attached way 39612875 is still missing two nodes 474857138 and 474857142. I hope you can reproduce this. Otherwise I have to look for a server where I can upload the files (Does anyone have a temporary FTP-account for ~200MB? ;-)
W> it W> is also ok if all inner nodes and for all parts of the way the first W> node out of the bounding box is included. This enables me (and W> others) W> to calculate the intersection point with the bounding box.
Sure, as I understand it that's why the overlap handling exists in the first place. As it currently stands it's not robust though, there are various corner cases where it gets things wrong (eg if the first node of a way that falls outside a tile also falls outside the overlap area then the way will appear to stop before the tile edge).
I can think of a few ways to improve the splitter to help with this, will try to take a look once I find a bit of spare time.
I think it would improve the splitter files if the overlap is removed and the ways are bounding box complete plus first nodes outside the bounding box. Anyhow I was impressed how fast the splitter worked on my lame old computer. That's good work! WanMil