[Qgis-developer] WFS Provider Strategies

[Even Rouault]

Hi,

> 
> I am a bit confused by the WFS option *"Only request features
> overlapping the view extent"*.
> 
> In general, that's what every provider does by default and is a safe
> strategy to use. If this option is unchecked, QGIS will just start
> downloading features randomly until it hits a server limit (or not).

Time to upgrade the servers to support WFS 2.0 paging :-)

> 
> The only reason why one would possibly want to do that is to warm a
> cache on a layer where he know he will be getting all the features (in a
> reasonable time). In this case, the option should probably be renamed
> (to something containing the word "cache").

The provider always cache (meaning keep in a temporary database the features 
retrieved), whatever you check or uncheck this option. For example with the 
default behaviour, zoom in in some region and wait for some or all features to 
be retrieved, and zoom out. The provider will use the already downloaded 
features first and then return the features coming from the background 
downloading of the new request (using gml:id/fid to make sure not to return 
already cached features).

> 
> Question: *is **there anybody using WFS with this option unchecked?* And
> if yes, for what reason?

Yes it is needed for some use cases. There are servers that have buggy 
behaviour with BBOX requests combined with other filters. Or the cost of 
solving a spatial request might be super high, making it more practical to 
issue a GetFeature without a BBOX.

> 
> Apart from this, Raymond noticed that cached features are not dismissed,
> when they are re-downloaded so you end up with duplicated features after
> multiple rendering cycles. Does that ring a bell for someone?

That should be better explained to be sure. It might be an issue with the 
features returned by the server having not a stable gml:id (or even not a id 
at all). Richard raised a ticket about that.

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com