[QGIS-Developer] Processing: No spatial index exists for input layer, performance will be severely degraded

[Andreas Neumann]

>> @Luigi: if I manually add "create spatial index" often, this very much "pollutes" the model.
>> 
>> I agree with Ujaval, that it would be a better user experience if QGIS would do the indexing automatically (on intermediate layers) for operations where the input doesn't have an index and the algorithm would potentially benefit from an index.
> 
> This is a fair enough request -- we could keep the index generation as
> a "lazy" thing which is only done when a step in the model indicates
> that it's about to benefit from it. I can't see any technical
> blockers/disadvantages to this.

Ok - great. If this requires funding - we will suggest / vote for
several grant proposals from the QGIS-CH user group early next year. I
would like to invest in "data quality" tools/improvements - also, but
not limited, to Processing. 

>> Next to automatic indexing I would think that "auto-repairs" for errors that processing algorithms introduce would be good to have, e.g. remove double vertices, remove empty geometries, empty holes/rings, self-intersections. etc.. If this is potentially dangerous, we can do opt-in with central threshold values, and only if the analysis is 2d only. I understand that for 3D analysis automatic cleaning is not always possible.
> 
> There's some discussion on this here:
> https://github.com/qgis/QGIS/pull/39891#issuecomment-734676128

Ok - great - thanks for being involved here! 

>> Or maybe a switch to GEOS 3.9 with the "fixed precision model" enabled would remove all these errors previously introduced by processing?

I'll open a separate thread for this discussion. 

Andreas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20201127/8edf4dae/attachment.html>