[OSM-talk] Splitting and merging ways (was Re: road on western side of Loch Ness)

[Richard Fairhurst]

Quoting matthew-osm at newtoncomputing.co.uk:

> Just out of interest, what happens if you try to split a way that
>
>  * contains cycles (loops); or
>  * is disjoint?
>
> As you know, because a way is purely an ordered list of segments, with no
> constraints on connectivity or otherwise, these are entirely   
> possible. (If a way was an ordered list of nodes, of course, it would
> make this easier.)
>
> I would guess that an error is given?

Ok. Essentially the idea behind Potlatch is to create a simple  
newbie-friendly editor (which we don't have yet), as opposed to a  
power-user editor (which we already have). The main thrust of this is  
that segments are never exposed to the user: it's ways and nodes alone.

Consequently, you can't use Potlatch to draw a new line with loops  
(except in the special case of first node=last node) or disjoints -  
you'll need to use JOSM for that. However, given that there are lots  
of loops and disjoints in existing OSM data and it mustn't break  
these, Potlatch does attempt to support these within the limitations  
of the UI.

If we have a way

        A--B--C--D  E--F--G

(i.e. disjoint between D and E) and split it at C, the result is:

way 1: A--B--C
way 2:       C--D  E--F--G

(The key/values of the original way will be copied to both new ways.)

Loops are fairly similar. If we have:

              F--E
              |  |
        A--B--C--D
              |
              G

and split at E, the result is one A-B-C-D-E way, another E-F-C-G.

If we split at C, it will assume splitting at the first instance; the  
result will be one A-B-C and another C-D-E-F-C-G.

You didn't ask, but I might as well explain merging ways, because the  
limitation is more obvious here. If you have two ways:

way 1:  A--B--C
way 2:           D--E--F

then you can merge them by drawing a connecting line from C to D. Easy enough.

But if you have two disjoint ways:

way 1:  A--B--C  D--E--F
way 2:                     G--H--I

you can't join C to G. It will only work for the first or last node in  
a way, not for any intermediate nodes. This does therefore require  
that the source segments are well-ordered until I implement a "reorder  
segments" option (which won't be in the first release).

Unfortunately it doesn't yet have a "render way as ASCII art for  
mailing list message" option... I think that's more a JOSM power-user  
thing. :)

cheers
Richard