[postgis-devel] temporal indexing

Chris Hodgson chodgson at refractions.net
Tue Aug 7 10:19:27 PDT 2007


You didn't mention if you would be doing spatio-temporal queries, ie. 
querying for objects which existed within a certain region of time and 
space. For these types of queries, the cube type and index which is 
provided in the postgres-contrib package is quite effective. You can 
store your x,y coordinates in the first two dimensions of the cube,and 
the date/time in the third dimension. The gist-based cube indexes are 
are very fast, but do consume a lot of disk space.

If you're only querying based on date then a standard b-tree index on 
the date itself should provide reasonable performance.

Chris

Elena Camossi wrote:
> Hi list,
> 
> I'm designing a spatio-temporal schema with support for coverage and 
> temporal data in postgresql by using postgis spatial extension. The 
> temporal support so far is quite limited,  just timestamps which state 
> the time at which data have been collected. Anyway, the temporal 
> information is extensively used to access data, so the corresponding 
> fields should require an index.
> 
> The scientific literature states that R-Tree perform quite well with 
> this kind of data, but there exist other indexes that should perform better.
> Does anyone in the list has some previous experience on using GiST 
> postgres implementation for indexing postgresql temporal columns? Does 
> anyone extend the GiST operator classes with specific support for 
> temporal data?
> 
> Thank you for helping.
> 
> Best regards,
> -Elena
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-devel




More information about the postgis-devel mailing list