[GRASS-user] how does one merge two street maps?

[Moritz Lennert]

On 26/08/13 03:31, James Powell wrote:
> Hi list!  I'm a long-time listener, first-time poster.
>
> I have struggled long with a simple problem, and I beg you to help me.
>
> I have two separate shapefiles,
>
> 1) streets: a street map with attributes list like address range
> for a segment, zip code, length, and (most importantly for me) road
> class.
>
> 2) segment: a second street map covering the same area, but with a
> completely different (and complementary) set of attributes.
>
> I want to merge these two into a new shapefile whose attribute table
> has all of the attributes of both streets and segment, with a single
> row for each feature.
>
> I've tried all kinds of variations of v.patch and v.clean.  I've
> successfully given segment and streets both separately the identical
> attribute table structure (identical schema in SQL) because as man
> v.patch says
>
>    If the table structures are identical, attributes are transferred to
>    the new table.
>
> I've used v.category to make sure the category numbers of the two
> don't overlap.  By the way, there is absolutely no key allowing to
> find a streets record from a segment record or vice-versa except for
> the geometry.
>
> My process looks like this:
>
>    :  v.patch --overwrite input=segment at PERMANENT,streets at PERMANENT output=segstr -e
>    :  v.clean --overwrite input=segstr at PERMANENT output=segstr_clean tool=break,snap,rmdupl
>
> or similarly, but different,
>
>    :  g.remove vect=segstr
>    :  g.copy vect=streets,segstr
> 	:  v.category option=sum cat=200000 input=segstr output=segstr_cats --overwrite
>    :  v.patch -a --overwrite input=segment at PERMANENT output=segstr_cats -e
>    :  v.clean --overwrite input=segstr_cats at PERMANENT output=segstr_clean tool=break,snap,rmdupl
>
> Note I'm hardcoding 200000.  I am pretty sure that's safely high
> enough to separate the two.  I recall from
> grass-user/2012-March/063958.html,
>
>     If you're on a *nix command line, you can do this:
>
> 	 v.category in=gr_help1 out=gr_help1_cat option=sum value=`echo 'SELECT
> 	 MAX(cat) FROM gr_help1' | db.select -c`
>
> and that's useful, but let's assume for now that my 200000 is
> separating the categories so that there is no overlap.
>
> I do not know what v.patch does with the overlapping cats in the
> non-append (no -a) case, thus I am not sure how the two approaches
> above differ.
>
> In any case, either way, after, the resulting attribute table is a
> concatenation of the input tables, i.e. there are two rows for each
> line segment but as far as I can tell only one piece of line segment
> geometry remaining (as you'd expect from rmdupl in v.clean) which
> itself points to exactly one of the two rows by category.
>
> The other row is orphaned it seems like.  I try v.what with
>
> v.what --q -a map=segstr_clean at PERMANENT east_north=7627976.73,709484.80 distance=0.37619
>
> and I get just one of the records (the one from segment).  I do not see
> any way to progress.
>
> What I would like to have happen is either there are two rows per
> feature where each row one cross references the other (via perhaps a
> "cat_b" attribute in analogy with a_ and b_ in v.overlay)

And v.overlay doesn't work for you ?

Another option would be v.distance. Assuming that there are more 
segments than streets:

- use v.distance to load the category id of the respective street 
corresponding to each segment into the attribute table of the segments 
map. If the geometries overlap perfectly you can use dmax=0, otherwise 
leave a tolerance.

- Use sql or (v.)db.join to join the attribute table of the streets map 
to the attribute table of the segments map.


Moritz