logo separator

[mkgmap-dev] Commit: r766: Write more of the *.mp-data.

From Robert Vollmert rvollmert-lists at gmx.net on Thu Dec 4 15:06:41 GMT 2008

Hi Alex,

thanks for your help!

On Dec 4, 2008, at 13:44, Alexander Atanasov wrote:
> On 12/4/08, svn commit <svn at mkgmap.org.uk> wrote:
>> what it does. The road class from RouteParams is set in various
>> places, not really sure they should all get the same. (Why is it
> They shouldn't be the same. Some ascii art:
> ----------------TRUNK ROAD------L----------------
>   -- local neighborhood streets -|
>   |---  --- many small streets --- |
>       \
> -------L-  OTHER TRUNK ROAD ----------------
> The small streets should be class 0, TRUNKs class1, etc.
> highways class 6.
>>> From the small streets you have links to L, how to get on the
> next road class /i'm not sure but highway herarchies sounds like this/

Right, the streets come different classes. I'm currently writing this
class in three places:
NOD 2: the first byte is class and speed and two unknown bits (once  
per road)
Table A: one byte is class and speed and oneway and toll (once per arc)
Arc: the destination class (this must be wrong)

For the moment, we're not writing any links. I assume well-placed links
would just optimize routing, but it seems to work without.

Are links only ever between nodes on the same road, or can they go  
where A and B are nodes in a residential area, 1 and 2 are different  
streets, and C is on a trunk road. Can there be a link A->C?

>> "destination" class in an arc?)
> That's for reading. If you are on node class 5  skip nodes with  
> class < 5.
> The graph is searched bidirectional. Start from pos and dest and try
> to go to a bigger road class. At some point they meet each other.
> In other words when you get on the highway stay there, don't route  
> along
> residental streets.

The class of a node is the maximum of the classes of the roads
it lies on? And the destination class of an arc/link is the class
of the destination node?

That makes a lot of sense, but I remember seeing some maps where this
didn't seem to be the case.


More information about the mkgmap-dev mailing list