Re: Advice on geolocation

I've done quite a bit of this as I used to work on a GIS
product.  Given to lat,lon pairs it is not hard to find
the great circle distance between them.  This assumes a
spherical Earth but good enough for your purposes as the
error will be under a few hundred meters

What you do is assume a plane defined by three points: your two
geolocations and the Earth's center. Next assume a circle
the same radius as the Earth (look this up) with center
at the Earth's center.  Your two geolocations are on the
edge of the circle.  Find the angle between the points then
divide that angle over 360 degrees and multiply by the
circumference of the circle (Earth).

So we have converted a spherical geometry problem (hard)
into a plain geometry problem (easier high school math)

The trick to computing the angle between the two locations
is to convert lat,long, at one Earth radius to (X,Y,Z)
then you have two vectors and can use length,length,length
to define a plain triangle.

This is an easy one because you can convert it basic plain
trig and you don't have to hunt down a specialist in
spherical trigonometry


--- Bruno Wolff III <bruno@wolff.to> wrote:
> I use a static postgresql database to serve web pages containing
> information
> about board game players and their ratings in various games. There
> are
> about 7000 players listed and about 300 games. I don't expect this to
> grow
> quickly in the short run. Most of the players are located in the US.
>
> One of the planned improvements is to publish contact information and
> to let people search for other people who live nearby and have common
> interests in games. I have street addresses for most of these people,
> though they currenty aren't in the database. I am also in the process
> of using Daniel Egnor's geocoder project (for the Google contest)
> working so that I can get Latitude and Longitude for most of these
> addresses. I expect to be able to do this for something like 5000
> people.
>
> I am looking for hints on how to best use this information for radius
> queries. I have tried looking at Gist and PostGIS documentation.
> It looks like for Gist I would need to create a new datatype (since
> I don't want to use a flat earth model making point unsuitable) and
> figure out an efficient indexing scheme. I am pretty sure this is
> beyond my
> capability to do (in the short run at least). It wasn't clear to me
> that
> PostGIS would be fast for the small amount of data I have as their
> data
> models looked pretty general and probably had a lot of overhead. Can
> people
> recommend using either of the above approaches or any other open
> source
> solution?
>
> I also am not sure if using 2d lat+long or 3d rectangular is a better
> way
> to go. 3d is simpler as there is no wrap around. If I use a spherical
> model for the earth, then for any great circle distance I can compute
> a secant distance and using that as a bounding box distance for
> searches.
> This seems a lot easier than doing things with lat+long. If I do this
> there may be tricks for setting up the coordinates so that indexes
> work
> better. Since initially I will only have US data (and even later I
> expect
> to have mostly US data) setting up coordinates so that they are used
> from
> most discrimating to least discrinimating seems to be a good idea. I
> didn't
> see much on this issue at the main parts of the PostGIS site, but I
> haven't
> tried to read any of the referenced papers so far.
>
> ---------------------------(end of
> broadcast)---------------------------
> TIP 5: Have you checked our extensive FAQ?
>
> http://www.postgresql.org/users-lounge/docs/faq.html


=====
Chris Albertson
  Home:   310-376-1029  chrisalbertson90278@yahoo.com
  Cell:   310-990-7550
  Office: 310-336-5189  Christopher.J.Albertson@aero.org

__________________________________________________
Do You Yahoo!?
Yahoo! Health - Feel better, live better
http://health.yahoo.com