Use a geohash index when clustering tables in the optimization phase

[tbonfort]

c.f. http://workshops.opengeo.org/postgis-intro/clusterindex.html

will submit a PR if it has a chance of being integrated

[olt]

Cool finding! It says that "an R-Tree built incrementally on spatial data might not have high spatial coherence of the leaves". Does this also apply when the index was created on a full dataset?

@pramsey can you answer that?

If not, then we don't need this with your change #14. Otherwise a patch would be welcomed (maybe with ST_Centroid before ST_Transform).

[pramsey]

Yes, the effect can be the same either way. Data with highly variable density will show this effect more than homogeneous data. I wouldn't necessarily expect any huge performance gains from clustering though, these are all improvements on the margins. P.

On Thu, Sep 27, 2012 at 2:10 AM, Oliver Tonnhofer [email protected]:

Cool finding! It says that "an R-Tree built incrementally on spatial data might not have high spatial coherence of the leaves". Does this also apply when the index was created on a full dataset?

@pramsey https://github.com/pramsey can you answer that?

If not, then we don't need this with your change #14https://github.com/omniscale/imposm/issues/14. Otherwise a patch would be welcomed (maybe with ST_Centroid before ST_Transform).

— Reply to this email directly or view it on GitHubhttps://github.com/omniscale/imposm/issues/15#issuecomment-8930319.