Dear postgres users,
I would like to set up a table that contains a 3D position, a time and a measured value, e.g. columns: x, y, z, t, val.
Does
bog-standard postgres (i.e. no PostGIS extension, if possible) provide
a way to index this table to allow the following 2 types of query to
be performed efficiently?
1. Select all the measurements in a given spatio-temporal box.
2. Select the N points nearest (in the euclidean/pythagorean sense) to a specified point.
If so, how do I define the correct index type, preferably in a manour which is supported by both postgres versions 7.4
and8.*? The 7.4 manual suggests postgres can do an R-tree index, but 8.4's suggests that this kind of index has been
subsumedwithin the GiST framework. Either way, I don't really know where to start.
I've played with PostGIS a couple of years ago, but, for the sake of installation simplicity,
I'd really prefer to avoid it--I don't require any of the fancier
geometry types, coordinate conversion or GIS functionality. And (back
then, at least) PostGIS didn't provide any inbuilt 'find nearest'
capability (probably because it would be very tricky to implement
efficiently if the geometries use ellipsoidal coordinates), although queries like (2), above, are easy to do quickly
whenyou have an R-tree index of points in cartesian coordinates.
Nathaniel
