[Tagging] Carriageway divider
gregory at gregorywilliams.me.uk
Tue Aug 21 09:37:06 BST 2012
> -----Original Message-----
> From: Markus Lindholm [mailto:markus.lindholm at gmail.com]
> Sent: 20 August 2012 11:51
> To: Tag discussion, strategy and related tools
> Subject: Re: [Tagging] Carriageway divider
> On 20 August 2012 10:55, Gregory Williams
> <gregory at gregorywilliams.me.uk> wrote:
> >> -----Original Message-----
> >> From: Markus Lindholm [mailto:markus.lindholm at gmail.com]
> >> Sent: 19 August 2012 19:26
> >> To: Tag discussion, strategy and related tools
> >> Subject: Re: [Tagging] Carriageway divider
> >> On 19 August 2012 18:23, Tobias Knerr <osm at tobias-knerr.de> wrote:
> >> > On 19.08.2012 15:09, Markus Lindholm wrote:
> >> >> On 19 August 2012 14:49, Fabrizio Carrai
> >> >> <fabrizio.carrai at gmail.com>
> >> wrote:
> >> >>> This could be a solution but it is against the reality: this kind
> >> >>> of road are indeed a single entity. The "legal" division, i.e.
> >> >>> the "solid_line" is just an attribute.
> >> >>
> >> >> There's a multitude of cases where a single entity is represented
> >> >> by multiple objects in the database, e.g. when the road changes
> >> >> speed limit it has to be split into two highway objects. The same
> >> >> with bus routes, to accommodate then the road was to be split into
> many parts.
> >> >
> >> > A major difference is that it is comparatively easy to re-assemble
> >> > a way that has been split (because they have common nodes).
> >> >
> >> > It's not so easy with two parallel ways that somehow "belong
> >> > - the connection could only be established by rather complex
> >> > heuristics based on proximity among other things. In practice, it
> >> > would simply result in gaps or overlaps appearing randomly
> >> > depending how
> >> "parallel"
> >> > the mapper has actually drawn the ways, and on the width assumed
> >> > (or
> >> > tagged) for the ways.
> >> For which purpose would the two highways be "reassembled"?
> > Split highways may be reassembled when you're not interested in the
> > attributes that do change between them. For example when you want to
> > reassemble the portions of the same road with the same class and name
> > together but aren't interested in the fact that the speed limit
> > changes partway down, the lighting changes, the surface changes, or
> > that a small portion of it has a cycle or bus route which crosses it
> > for a few tens of metres. If building a routing graph from the data
> > you'd want to keep the graph as simple as possible by ignoring the
> > tags not relevant to your routing and reassembling the adjacent
> identical segments.
> Yes, I understand why one would reassemble highway segments on a route
> that only differ on the maxspeed tag or other such minor issue. But why
> would one want to reassemble two highways going in opposite direction and
> from which there is no direct legal route to the other?
Most of this still applies to two parallel opposite highways. When building
a routing graph you may want to combine the opposing ways together into one
way that represents both directions in order to simplify the routing graph.
More information about the Tagging