[Tilesathome] Simple collision avoidance for or/p

[Frederik Ramm]

Hi,

   I have built a very primitive "collision avoidance" into or/p. It's
not yet in SVN because I wanted to await some feedback, but I'll make
it available in a few days.

You can see the results (compared with the standard t at h slippymap) at
http://openstreetmap.gryph.de/slippymap/. A typical result is
something like this:

http://openstreetmap.gryph.de/slippymap/index.html?mt0=orp&mt1=tah&lon=8.4451469&lat=49.0033785&z=15

What it does is the following: Whenever it writes a text along a path
(usually that's road names or refs), it stores that in a hash table:
"wrote string <x> along a path that went from node <a> to node <b>"
(where <a> and <b> are the end nodes if the way and are swapped so
that <a> is on the left and <b> on the right).

Before it writes a text along a path, it checks whether there is an
entry in the hash table for the same text and a path whose end nodes
are within 100 metres of the end nodes of the current path. If yes, it
silently ignores the text instruction.

This method has many shortcomings, most of all it simply writes the
first text it encounters and ignores the later ones, and because it
only ignores text that's basically on the same path, it is still
possible to have overlapping labels. There's also some potential for
wrongfully weeding out labels that could have been placed without
problem, if you imagine an U shaped way and another way that connects
the two legs of the U, and both had the same labels - the algorithm
would say "start and end node within 100m of each other, so discard
one of the labels". 

However, for the very common case of a dual carriageway or similar
construct inside a city, where both ways in OSM are roughly the same
just a bit displaced, this method nicely avoids the double-printing of
names/refs.

I am a bit unsure how I should build this into or/p. I could make it
so that you enable/disable the collision avoidance by placing an extra
attribute on the text instruction. That would make it very
controllable but bloat the rules file. Or I could make it a global
config option in or/p so that when you switch it on, it works for all
labels (that's what I did in my test script that generated the demo
slippy map).

One could also make the "100 metre" setting configurable. And lastly,
the algorithm could of course be widened to not look at the text, i.e. ignore
*anything* that is written along a way when *anything else* has been
written along a similar way. It could also be changed to better
compare the ways (perhaps take first and last and middle point or so),
or tons of other things.

Also, I haven't looked too closely at the existing "proximity" code in
Osmarender/XSLT and I don't know if that basically does the same.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"