logo separator

[mkgmap-dev] Admin relations

From Henning Scholland osm at hscholland.de on Thu Jul 12 14:20:50 BST 2018

Hi Gerd,

I felt a little strange, as it is the only tile causing this problem. I 
took a look into the result of splitter, added two nodes with 
test=country and they were set to the matching country code. As the 
issues were matching exactly with the tile boundaries, I was thinking 
maybe it's just swapped for this tile.

You are right, the ways are quite long in that area. But for the east 
level4-border in Mongolia, which is shown, there are plenty of nodes in 
Mongolia. Even after splitting the way at the tile edge and even on the 
west level4-border in Mongolia, there are 4 nodes after splitting the 
way at tile edge. I put relevant data in attachment. Those ways issue 
doesn't match to your explanation.
For the Chinese level4-border it just contains a 2-node way, with or 
without splitting it on the tile edge. So there is no n/2-node for my 
understanding of your explanation. The direction of the way is from 
south-to-north. So the first node should be definitely in China, the 
last one is on the borderline.

Either I got your explanation wrong or there is some issue on that tile. 
Maybe because of it flat and long shape? If you need I can send you the 
result of splitter for that tile by private mail.

In general it could be a idea to calculate an artificial middle node for 
2-node ways to check first, as there is no n/2 one. But as pointed out 
above, I don't think this is the issue in my case.

Henning

(Hope you enjoyed the cycling trip)

On 12.07.2018 13:19, Gerd Petermann wrote:
> Hi Henning,
>
> I think you might hit a problem here because of the way how the LocationHook uses the data in the bounds file.
> When used for a way with n nodes it first tries to get information for node n/2. If that fails it tries the first, then the last, finally all other nodes.
> The lines in your screenshot all look straight, so the corresponding OSM ways might have only 2 nodes. In that case the only tested node
> may be the one on the border between China and Mongolia. I assume that the actual mkgmap:admin_level2 value for those nodes are more or less
> random or that the value is empty.
>
> If the value for mkgmap:admin_level2 is empty mkgmap should probably try another node. This doesn't happen.
> I see no easy way to handle this when the value is set.
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Henning Scholland <osm at hscholland.de>
> Gesendet: Donnerstag, 12. Juli 2018 01:58:16
> An: mkgmap-dev at lists.mkgmap.org.uk
> Betreff: Re: [mkgmap-dev] Admin relations
>
> Hi Andrzej,
> yes they are both in and in general it's working. Just on the borders I
> get these issues. I guess it's because of precision, how mkgmap
> calculates the area of China. The strange thing is, that the missing red
> part on Chinese border (Inner Mongolia/Gansu) Is roughly from the center
> of the last way to the end. For the blue marked ways it's shorter than
> half of it's length. So can't be the reason, a way or so is missing or
> is ignored completly. It's seems more, that there is kind of buffer, as
> both has the same length missing. And the both blue ways seems to end at
> same latitude. Sorry for forgetting the scale on the picture, but it's
> about 50km length missing/too much.
>
> If it's like this I wonder how mkgmap:country can work for address
> search in more dense populated areas.
>
> Henning
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: source.osm
Type: text/xml
Size: 55988 bytes
Desc: not available
URL: <http://www.mkgmap.org.uk/pipermail/mkgmap-dev/attachments/20180712/aa5ebe57/attachment-0001.xml>


More information about the mkgmap-dev mailing list