Is there any way to know if a road has railway (tram) tracks embedded in it?

In the Netherlands it is common to map embedded tram tracks in a single way as the highway, and it is rendered without problems on Mapnik, see http://www.openstreetmap.org/browse/way/7372002 So I dont understand why this method isn't recommended in the wiki.

AFAIK, no official way to represent this configuration on the map.

Drawing two ways, one for the road, the other for the tram track, reusing the nodes where they overlap will trigger the "overlapping ways" validation warning in JOSM. Other mappers will be tempted to "fix" it. Also, they are two different, separate ways from the point of view of routing software.

Drawing two ways as above, but not reusing the nodes is even worse, because there will not be even the hint of the overlapping segments telling that there are two ways there, one for the trams and other for the vehicles. Software will try to "get" the nearest way from your point of view, and will select one of the two, since it is very hard to manually adjust two ways to exactly the same coordinates. Probably will trigger lots of "crossing ways" validation warnings.

Drawing one way with both the railway and highway tags set is not recommended as per the wiki page, but I could not find exactly why for the specific case of street that have the tram line embedded. I believe that the main reason is because it would be difficult for the rendering software to decide what to render (if the highway, the tram line would appear broken, else the street would). And I think, why not, since the street and the embedded tram line effectively use the same space on the ground? Also, for bus lines, there is already the "lanes:psv" (and other tags) to represent multiple types of vehicles sharing the same way.

Drawing parallel, but not overlapping ways would be right only when the tram line and the street didn't share the same physical space, i.e. no cars/pedestrians allowed over the rails and (obviously) no trams running over the street...

In short, if I had to choose, I would pick the third option, because, even if it makes things difficult for the rendering software, is the best form to allow for the fact that both share the same physical space on the ground.

From a data processing point of view, there are several ways to attack the problem.

A relatively fool-proof way would probably be this:

• Import your data into a spatial database (e.g. use osm2pgsql to import into PostGIS)
• To find out whether a road is "dangerous", convert the road's linear geometry into an area (called "buffering", usually with ST_BUFFER or similar), and ask the database whether there is anything called railway=tram inside that buffer.

This would capture all the different ways of tram mapping, but it might require some refining (what if the tram is on a bridge over the street or what if it just crosses the street).

Another option is to first process all ways tagged railway=tram, split them into two-node bits ("there is a tram directly connecting node A with node B", "there is a tram directly connecting node B with node C", etc.); store that in a suitable memory structure, and when processing the roads later, simply check if any pair of two consecutive road nodes is found in your tram segment table.