[Tile-serving] Hstore benchmarking results

[Sven Geggus]

Paul Norman <penorman at mac.com> wrote:

> If you're using the database to render a map with the common OSM features,
> --match-only is never going to do much.

This is simply not true. As I already said it is pointless to use
"match-only" with the standard osm2pgsql style.

Instead we import _only_ those objects which have one of the "important"
tags.

"Importance" in this case is derived here from the left side of the SQL
select functions but not the where clause inside the standard mapnik style.

Fortunately those are very few! 22 to be exact:

aerialway	addr:housenumber	aeroway
amenity		barrier			boundary
building	highway			historic
landuse		leisure			lock
man_made	military		place
power		railway			route
shop		tourism			waterway
natural

Thus e.g. an osm object featuring only "name=foo" but nothing else will not
get imported into our database at all and will thus save disk-space.

What we try to do with "match-only" is importing all the data which might be
needed for rendering now and in future, but discard all the rest.

This works because it is unlikely, that new important tags which should be
rendered in future will get introduced.

Strictly speaking we would not even need a real database column for our
"important" tags. We could as well opt for a hstore-only database.

Howerver, there is currently no support in osm2pgsl other than "match-only"
to specify required tags for objects to be imported and discarded otherwise.

Without ever doing a comparison of database sizes it is certainly the case,
that the match-only option is the one which produces the smallest possible
databases while keeping all tags available which might be of interest for
renderring in future.

Regards

Sven

-- 
"Ich fürchte mich nicht vor der Rückkehr der Faschisten in der Maske der
Faschisten, sondern vor der Rückkehr der Faschisten in der Maske der
Demokraten" (Theodor W. Adorno)
/me is giggls at ircnet, http://sven.gegg.us/ on the Web