Vivek, This approach sounds really interesting.
Are you able to share a small example? Link to working code would be a bonus! Also, geocouch is hardly experimental these days, it's at least 2 years old and has been well used. A+ Dave On 12 December 2013 04:02, Vivek Pathak <[email protected]> wrote: > Hi > > I have been using this approach successfully via a view. The view emits > the geokey based on latitude and longitude of the document (document has a > location). > > The idea is to traverse the view in ascending and descending orders - from > a start key corresponding to center of bounding box - and to get sufficient > number of "candidates". Then filter out what is not needed (eg: must be > within a given shape etc.) > > I believe using both ascending and descending should solve the 00 and 20 > problem you state below. > > My best wishes for the geo couch enhancement. No reason why it couldnt be > k dimensional as well... > > I just mentioned my approach since it is a common misconception that > multidimensional can not be efficiently mapped to single dimension. I found > the solution to work very well, and it is conceptually simple also. > > Thanks > Vivek > > > > On 12/11/2013 02:22 PM, Alexander Shorin wrote: > >> On Wed, Dec 11, 2013 at 11:16 PM, Florian Westreicher Bakk.techn. >> <[email protected]> wrote: >> >>> That's an interesting idea but it will not work in all situations. You >>> could not query for areas that span multiple quadrants. If you would like >>> to get all locations around 0,0 in a 20km square this won't work. >>> >> This task looks suitable for GeoCouch extension: >> https://github.com/couchbase/geocouch >> >> -- >> ,,,^..^,,, >> >> >
