[Qgis-developer] Data for QGIS Hackfest

[Maciej Sieczka]

Andreas Neumann pisze:

> I'd like the mention that the display of very large rasters is
> normally not a problem if you used gdaladdo to add pyramids to your
> data. You can use rasters with the hundreds of gigs of filesize
> without a bottleneck. You'd have to build the pyramids however. If
> you store your data in multiple files instead of just one, you should
> build a tileindex with gdaltindex. See
> http://www.gdal.org/gdal_utilities.html for more infos.
> 
> You should definitively give the display of rasters in qgis a try -
> they should perform well.

If bulding pyramids is not an option, at least convert your rasters to
stripped instead of tiled. From my experience they render ways faster in
QGIS. Also use some performance-oriented compression like LZW instead of
e.g. deflate, if needed at all.

> The display of big vector data is another issues. Did you build
> spatial indices?

Do they help for rendering performance in QGIS? I gues they improve
querying only (correct me please if that's wrong).

You might try to enable "Fix problems with..." in display options. That
enables hardware render accel (though crashes the whole GNOME for me in
case of some vector datasets!). BTW - please devs rename this option if
possible.

> Are you directly loading from the shapefiles or from postgis? In
> realworld scenarios you should use various scale levels and reduce 
> your vector data with SQL WHERE clauses (e.g. only display major
> rivers in a river network). If you display
> tenthousands/hundredthousands of vector features in one display, you
> probably have generalization and map reading problem anyway. There is
> no value in displaying all of your vector features, say on a state
> level when you totally clutter your map display and can't distinguish
> individual features anymore.

Note that this approach is good for display only. For digitizing or
sometimes for querying even it's not a feasible option.

Maciek

-- 
Maciej Sieczka
http://www.sieczka.org