[mkgmap-dev] default style lines enhancements

Fri Jul 31 17:50:34 BST 2020

Hi

I agree that mkgmap:set_un/semi_connected_type can't remove circular
ways (eg car park circumferences) that cross a tile boundary and that
this might result in a routing island.

Routing inlands, unrelated to the car park issue, exist in the OSM data
for other reasons. --check-routing-island-len=+ve logic can't remove
these if they cross a tile boundary either, for the same reasons.

It would be good if mkgmap could remove these islands/semi/unconnected
but, as Gerd outlines, it would a really major piece of work. Also I
can see the possibility of this other tag, like mkgmap:set_unconnected
but it ignores the tile edge NODs, ie chuck edge loops unless they have
1 or more real connections. I don't know what happens if 1 tile has
these but the adjacent one doesn't.

A way of finding problem car parks might be nice, but joining all the
ways in suitable positions would probably need to be manual and take  a
few mins per car-park. 

Routing islands can cause route-calculation-error if at the start or
end of the planned journey; in either case the problem can be
circumvented by moving outside the unconnected network and/or choosing
an end-point outside the unconnected network.

Unconnected car-parks in the middle of a hoped-for route are not so
easily resolved; it might not even be noticed that parts of the
calculated route are ridiculous.

I can see no advantage in not having the circumference way logic in the
default style. However, as it is causing this contention, then it could
be there but commented out, along with an explanation of the issues. 

I see the default style as the ideal place to document these sorts of
issues, with sample rules etc, For instance: typ-codes with meanings
introduced on newer devices, how to show old railways, etc.

Ticker