[mkgmap-dev] Dead End Check

Thu Aug 30 15:48:47 BST 2012

On Thu, Aug 30, 2012 at 03:32:52PM +0200, Felix Hartmann wrote:
>While wondering how to best clean up all those micromapping 
>footway/service/track/path stuff, and thinking about the 
>mkgmap::length filter, I got the idea, that mkgmap got an dead end 
>check.
>Is there anywhere a description of how it works?

As far as I know, it only works on oneways (oneway=*). It will complain 
about oneways going to or coming from nowhere, or same-direction oneways 
joining at a node that connects to no other way.

I have added some style rules to suppress the dead-end warnings in 
certain cases. Also, in the core, I suppress the dead-end checks if the 
start or end node of the way carries a FIXME=* or fixme=* tag.

I routinely fix the dead-end warnings (or suppress them by setting 
fixme=continue) in Finland. Currently there is some polygon breakage, 
since the license bot took its toll. I am gradually working on fixing 
the data (in Finland only).

>At least the three 4m ones, I would like to get rid of. Why is this so 
>bad? Well if you make a map without buildings, you'll see the disaster, 
>short ways going nowhere:

I think that a more severe problem are routing islands, because they 
really break routing. Some mappers seem to think that a 
highway=pedestrian area is enough and may even delete an overlapping 
highway=* way as garbage.

As far as I know, currently mkgmap does not check for routing islands.  
It might be hard to avoid false positives near tile (or map extract) 
borders. If we ignore that, is it computationally expensive to detect 
routing islands? (Run Tarjan's algorithm and complain if there are 
several strongly connected components in the routing graph, perhaps?)

Best regards,

	Marko