[mkgmap-dev] splitter that generates problem list

Sun Nov 4 13:51:46 GMT 2012

Hello Henning,

I did not make sure, but I guess your split-file contains overlapping tiles. I did never test such a case and probably my algorithm runs into an endless loop. It is not easy to convert the log back to the format used in the split-file. 
Can you please send me your split-file?

reg. your suggestion:
To be precise, splitter reads the nodes, ways and rels (in this order). As I described before, if you want to calculate the problem list in one pass you have to store almost all information contained in the input file (except for the tags). This would require extremely large heap
(I guess more than 32GB) and will not work on a 32 bit system 

You have a 64bit system with 10000Mb heap, so I suggest to use a higher value for --max-areas, e.g. 1024. If possible, specify higher 
values.
With this value splitter might be able to collect all information in a few passes.

Ciao,
Gerd

Date: Sun, 4 Nov 2012 14:30:50 +0100
From: osm at aighes.de
To: mkgmap-dev at lists.mkgmap.org.uk
Subject: Re: [mkgmap-dev] splitter that generates problem list

    Am 04.11.2012 14:05, schrieb Gerd
      Petermann:

       Please try again using e.g. only the tiles
        starting with 14.

    Will do so.

      @ all: I need an algorithm that uses a list like
        below and fills "the rest of planet" with a small number of
        rectangles. There is no need to find the smallest number, but
        the result should be near.

    Maybe another solution, but don't know if this is useful.

    First, splitter reads relation, ways and nodes. So I think after
    these three steps it should be possible to have all problematic
    objects with all ways and nodes.

    Then check for all nodes, in which tile they would be and save this
    information to the problematic ways/relation, they belong to.

    Afterwards start normal splitting and if splitter check a
    problematic object, then stop splitter writing it in normal mode and
    instead write the hole object to the tile. If the order of the
    splitted tile isn't important, you can also look for each object, if
    it is in problematic part. If so, skip this object (node, way,
    relation) and write these at the end of the (node, way and relation)
    block.

    Don't know if this would be faster.

    Henning

_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mkgmap.org.uk/pipermail/mkgmap-dev/attachments/20121104/dd631027/attachment.html 