[OSM-talk] JOSM Modeless???
tom at compton.nu
Wed Oct 24 16:13:01 BST 2007
In message <E1IkhXB-0004yV-9p at streetmap.vm.bytemark.co.uk>
Rick Collins <gnuarm.2006 at arius.com> wrote:
> For example, if I want to make two ways share a common node, I have
> to add a node to one way, extend the way to a point on the other way,
> continue that back to the first way. Then I have to disconnect the
> existing section of the first way that is being bypassed and delete
> it. Finally, I have to combine the new way sections with the
> existing way and delete the new nodes. Instead, it would be nice to
> have a feature to just allow nodes on two ways to be merged.
I have to say I'm having a hard time following your description
of what you're doing, but it doesn't sound like anything I've ever
had to do when creating a new way that shares nodes with an
> One feature I would love to see could facilitate the use of track
> data for generating ways. Rather than use JOSM as electronic tracing
> paper to draw ways over top of the tracks, I would like to have a
> means to select nodes from the track to be transferred to a new
> way. In essence, let the track define the points and I am just
> culling the ones that are superfluous to the way. I think this could
> be facilitated by letting JOSM tell you which points can be
> eliminated and keeping the way to within a user specified distance of
> the track. In fact, I am working on a tool to test how well this can
> work. When I am done, I will be happy to share the algorithm, or
> perhaps I will learn to code in Java and I can share the code.
Historically this sort of automatic or semi-automatic generation of
ways from traces has tended to be discouraged as people have felt
that better results can be achieved manually.
There is a well known algorithm (Douglas-Peucker) for doing it though
so you don't need to invent your own.
One problem is that although you can automatically generate way that
follows a trace it is hard to know where to split it, and any single
trace is not likely to be accurate so it would be better to be able
to average several traces.
> Lastly, it would be very nice to be able to see the name of the
> action to be undone or redone. It is more than once that I can't
> tell if I moved an object when I selected it. If I undo and it was
> not moved, I then have undone something else and have to figure out
> what that was. The redo button is not always available for a given action.
If you open the command stack window then I think you will see
what you want - or have I misunderstood?
Tom Hughes (tom at compton.nu)
More information about the talk