2010/2/3 Lennard <[email protected]>: > Tobias Wendorff wrote: > >> This means: Whenever you zoom out of a map, there's lot of details, >> you don't need at all. The SQL-query grabs all this data of course. > > Mapnik does an explicit && with the bbox of the area to be rendered. > That may still amount to a lot of geometries returned when you are > rendering low zoom tiles, but it by no means grabs *all* the data. >
When I optimized my xml file, I turned off all layers (status=off) and enabled by one and measured time of each separately to find which are slowest and at which zoom. I discovered, that in my case biggest problems were style rules with high MaxScaleDenominator for very rare features mixed with popular ones with small MaxScaleDenominator in one layer. I.e. I had in one Layer/Style almost all POIs, and few of them (airports) had high MaxScaleDenominator. In effect query for low zoom (13) grabbed all possible POI in visible region, but only few were visible, rest needed mapnik filter processing. After that discovery I tried to extract more overview features to separate layer with specific sql query to limit other objects, that are only visible on higher zoom(>15). For forest and water polygons useful is also separate layer with sql query like: "way_area > 2000000" to skip small almost invisible features. Like previously, just make sure, that for rest in second layer for higher zoom, You have style definition with enough small MaxScaleDenominator. Same goes for roads as already mentioned. -- Marcin Rudowski _______________________________________________ Mapnik-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/mapnik-users
