For spatial buffer query, a simple C code was used to make a shared library with functions that can be loaded by Postgres backend by commands like:
CREATE FUNCTION pt_in_path_buffer(path,point,float8) RETURNS boolean AS 'geo_addfuncs.so','path_buffer_contain_pt' LANGUAGE 'c';
(syntax is for 7.1.3 version - different in 7.2)
Now we can query polygons (and sites) in different ways:
a) center of polygon boundary box is within buffer (picture)
b) overlap (any point within buffer)
c) contained (all points within buffer),
etc.
To draw the selected (falling in 1000 m buffer) polygons on screen:
d.vect.pg -f -s map=kuruma_id sql=test_buffer1000.sql
color=grey
test_buffer<i>.sql:
select rec_id from info_kuruma_bnd, riv_names_arc where (
(pt_in_path_buffer(riv_names_arc.segment, center(boundary::box),<i>)
= 't')
and (riv_names_arc.riv_id = 205001 or riv_names_arc.riv_id = 104004
or
riv_names_arc.riv_id = 204007))
Make buffers in psql with command:
select path_to_file(return_buffer(path_reduce(segment,2),100))
from riv_names_arc where riv_id = 204007;
Each time it writes a single result buffer line (in picture in violet
- 100 m) in the
file '/tmp/boundary.pol'. This name can be changed to whatever you
like. The line is further imported to grass by v.in.arc module.
Functions 'path_to_file', 'return_buffer' and 'path_reduce' are from the same library, and are created by Postgres superuser using CREATE FUNCTION. Check the code for correct parameter definitions before you create these SQL functions in psql.
Note: 'path_reduce' second parameter is number of times the input
line is generalized before buffering. This parameter is 1 if radius of
buffering is less or equal to segments lengths. However, when radius is
much greater than distances between vertices, the true UNION of circles
is intrapolated by 'thinning" the number of vertices. Thus we can produce,
e.g.,
1000 m buffer (in red in picture) with minimal (or none by adding more
to this parameter) 'swallow tails' that are resulting from algorithm based
on only 2 neighboring segments analysis. This algorithm is correct to the
limit of radius mentioned above. For large buffers, increase the parameter
in 'path_reduce' until the needed smoothness of buffer. For example, the
river 205001 with 1000 buffer (top of picture) needed 24-times thinning
to produce the shown line.
d.vect.pg -s map=rivers sql=show_rivs.sql color=yellow
show_rivs.sql:
select riv_id from riv_names_arc where (
riv_names_arc.riv_id = 205001 or riv_names_arc.riv_id = 104004 or
riv_names_arc.riv_id = 204007)