Hi, I ran into a problem with flooded areas around the river Rhine. A europe.osm.pbf from geofabrik.de dated 2011-07-31 was used. I extracted a contiguous area containing roughly Germany and France using osmosis. This extract was spit up into 369 parts using tile splitter r179. After they have been converted using mkgmap I recognized at least 3 flooded areas at the river Rhine on my Oregon 550. The riverbank areas extended wide into the land. The riverbank in that europe.osm was defined by a really large multipolygon. (in the meantime someone else split up the gigantic multipolygon into a number of smaller multipolygons). This large multipolygon consists of 320 ways (counted inner and outer) and extends over over at least 4 of 5 tiles that were created by tile splitter. I had a look at one of the tile-files an noticed, that tile splitter removed all the ways that were completely outside the bounding box of the respective tile. However the multipolygon contained all 320 ways as members, but most of them (and the corresponding nodes) were not included in the tile-file. Im not sure wether tile splitter did some processing in order to close the multipolygon along the bounding box, but probably this ist not the case. This incomplete information in the tile file probably makes it impossible for mkgmap to correctly clip the multipolygon along the bounding box and possibly leads to the flooded areas. As I already mentioned, this large multipolygon is no more present in the actual osm database. But I probably will be able to provide some test data if required as I still have the original data and can reproduce this behavior. Regards Wolfgang
On Sun, Aug 14, 2011 at 10:13:36PM +0200, Wolfgang Hammel wrote:
Im not sure wether tile splitter did some processing in order to close the multipolygon along the bounding box, but probably this ist not the case.
It does not. This is a known problem in the tile splitter. mkgmap does try to work it around to the extent possible. For what it is worth, the splitter in Osmosis implements an option not to break relations (such as multipolygons). It costs much more processing time. I do not know how much additional effort it would take to clip multi-polygons (and polygons, for that matter) along tile borders in splitter.
As I already mentioned, this large multipolygon is no more present in the actual osm database.
There are some large multipolygons elsewhere on the planet. One example is Lake Päijänne in Finland. The largest lake, Saimaa, is in several multipolygons, and every now and then someone adds an island to the wrong multipolygon relation. Best regards, Marko
i'm running a few tests with the new locator and a very simple command line. java -jar mkgmap.jar --index --tdbfile --remove-short-arcs --route --add-pois-to-areas -c template.args this works as expected. but: if i leave --route out, the address index is not generated.
Sounds strange. Is the mdr file created when leaving out --route? I will test that during the weekend. WanMil
i'm running a few tests with the new locator and a very simple command line.
java -jar mkgmap.jar --index --tdbfile --remove-short-arcs --route --add-pois-to-areas -c template.args
this works as expected.
but: if i leave --route out, the address index is not generated.
the mdr file is created, but quite a bit smaller (roughly 30%). map installs fine in mapsource. when i send the map to the gps, the "creasting index" progress bar part is only a fraction of a second long (so i guess no index is created at all). search for adresses/crossings is then not active on the gps (oregon 450). in addition to that: i did some more testing on that, using a test map of mine that adds road_speed and road_class to the street's names, so for instance i'd get "street a (rc0-rs0)" for one section of a street and for another "street a (rc1-rs1)". both these streets sections are findable in the index. if i leave out road_speed and road_class, "street a" appears once in the index. i guess that the index creation pocess of mapsource somehow overwrites streets with the same name/city combination, so only the first or last(?) processed section of a street is actually part of the index. this could be the reason for not finding crossings even if they're clearly there, because the crossing is in an overwritten section of the street. micha Am 16.08.2011 22:34, schrieb WanMil:
Sounds strange. Is the mdr file created when leaving out --route? I will test that during the weekend.
WanMil
i'm running a few tests with the new locator and a very simple command line.
java -jar mkgmap.jar --index --tdbfile --remove-short-arcs --route --add-pois-to-areas -c template.args
this works as expected.
but: if i leave --route out, the address index is not generated.
mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
michael lohr <micha.lohr@web.de> wrote:
i'm running a few tests with the new locator and a very simple command line.
java -jar mkgmap.jar --index --tdbfile --remove-short-arcs --route --add-pois-to-areas -c template.args
this works as expected.
but: if i leave --route out, the address index is not generated. _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
Hi The route option is required to add road information to the map tiles which is then read to generate the addresses in the index. Without route then the index will only contain poi and city information (assuming no bug prevents that from happening) ..Steve
i gathered as much in the meantime. is this a mkgmap or a mapsource mechanism? Am 17.08.2011 13:45, schrieb Steve Ratcliffe:
michael lohr<micha.lohr@web.de> wrote:
i'm running a few tests with the new locator and a very simple command line.
java -jar mkgmap.jar --index --tdbfile --remove-short-arcs --route --add-pois-to-areas -c template.args
this works as expected.
but: if i leave --route out, the address index is not generated. _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev Hi
The route option is required to add road information to the map tiles which is then read to generate the addresses in the index.
Without route then the index will only contain poi and city information (assuming no bug prevents that from happening)
..Steve _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
participants (5)
-
Marko Mäkelä -
michael lohr -
Steve Ratcliffe -
WanMil -
Wolfgang Hammel