[Qgis-developer] WFS upgrade: how much?

Marco Hugentobler marco.hugentobler at sourcepole.ch
Tue Dec 27 04:19:06 EST 2011


Hi Bill

Thank you very much for the usefull WFS extensions! Your patch is 
applied to master now (that was approach 2. I guess?).

Seems to me approach 2. is good in a zoom-in scenario whereas approach 
3. is also good in a pan scenario (e.g. panning back to areas already 
visited). So my favorite one is 3.
In longer term, it would be cool to have an intelligent caching strategy 
considering the used amount of memory and the already visited areas (and 
that could also be used as cache for other provider types, e.g. 
postgres, shape, ...).

Regards,
Marco

On 26.12.2011 10:26, Bill Clay wrote:
> Hello, all.
>
> I've been working on WFS bug 4604 and feature request 4280 as my first 
> attempt to contribute as a QGIS developer.  I have working code for 
> three different approaches. Which one should be completed and used?
>
> 1. Minimal patch.  Gives you a choice between caching the ENTIRE WFS 
> layer for the life of the layer (slow layer creation for big layers on 
> remote WFS providers) or freshly fetching WFS extents each time they 
> are rendered (slow on each WFS layer rendering).
>
> 2. Medium patch.  If you choose not to cache the entire layer, an 
> expanding extent is cached, starting with the first rendered extent.  
> Each time a new area is rendered, the cached extent expands, combining 
> the new canvas extent with the previous WFS layer extent.  On each 
> expansion, the cache is discarded and the expanded extent fetched 
> fresh.  After a short "training period," local pans and zooms are fast 
> because they work from the cache without requiring a new WFS fetch.
>
> 3. Maximum patch.  The "expanding cache" approach of nr. 2 above, but 
> instead of replacing the cached extent each time the extent expands, 
> we fetch only the NEW areas of the expanded extent (i.e., 1-4 
> rectangles around the perimeter of the cached layer extent) and ADD 
> their features to the cache (discarding duplicate features).  Faster 
> rendering of new areas than nr. 2 when there are many features, but 
> slower for small layers because a separate HTTP GET is required for 
> each of 1-4 new areas.
>
> Is there a "typical" WFS user?  Do they use WFS layers with many 
> features?  If so, patch 3 might be best; otherwise maybe patch 2.
>
> Best wishes for the holidays!
> Bill Clay
> _______________________________________________
> Qgis-developer mailing list
> Qgis-developer at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/qgis-developer


-- 
Dr. Marco Hugentobler
Sourcepole -  Linux&  Open Source Solutions
Churerstrasse 22, CH-8808 Pfäffikon SZ, Switzerland
marco.hugentobler at sourcepole.ch http://www.sourcepole.ch
Technical Advisor QGIS Project Steering Committee



More information about the Qgis-developer mailing list