Hi Matthias,
I think I can follow that, and it makes sense that QGIS does the
calculation independently of the provider for consistency.
I will indeed look into the details of the view and see if it is one
particular join that is so expensive or that it is just the total number
of joins that causes it. At the same time I'll keep my eyes open for
potential projects that might be able to contribute, but that's a bit
more complicated.
Thanks for helping me understand the background of the process,
Janneke
On 30/11/2015 14:46, Matthias Kuhn wrote:
Hi Janneke,
The reason for individual statements are mainly that expressions
(field calculator) and SQL (table manager) are not completely equal.
Some functions are only supported by QGIS, some only by
postgres/postgis. The expressions should be supported by all data
providers (different database systems, non-database systems) and
therefore QGIS plays safe and just does the calculation always on its own.
Some background:
We have this summer started to be able to generate postgres SQL from
expressions (where possible) but this has only been applied to some
isolated areas (like filtering for rendering). It would theoretically
be possible to use it for provider side updates as well but that's an
ambitious task that will need some work by an experienced developer.
I'd propose looking into what makes updating your view so expensive
(probably easier) or consider to contribute towards provider side updates
Matthias
On 11/29/2015 10:03 PM, Janneke van Dijk wrote:
Hi Jürgen,
After your remarks I thought I would try the individual update
statements (7000 of them ) directly on postgres. Indeed, that gives
an out of memory error after 10 minutes. Not sure if that is entirely
up to the expensive view or that there is something more going on. I
also tried writing the update statement as an update on the view with
the where clause being 'where id in (list of ids)', this again goes
quickly (6 seconds).
To sum it up: unless there is a way in QGIS to update selections
using one update statement with a where clause (as opposed to
individual update statements per record), it is not possible to work
with larger selections on 'expensive' views in postgres. Would that
be difficult to implement, or are there other reasons for issuing
individual statements?
Thanks again all for your thoughts,
Janneke
On 29/11/2015 22:36, Jürgen E. Fischer wrote:
Hi Janneke,
On Sun, 29. Nov 2015 at 20:44:54 +0300, Janneke van Dijk wrote:
Using the field calculator either from the attribute table or from main
window with the attribute table closed both takes too long to be useful. Not
the actual calculation, but the saving of the edits. To be clear: I
originally used the update expression bar to update the selection, but all
three methods take too long to save.
That's because - as Matthias already said - the expression are evaluated for
each record and the changed attribute value is temporary stored in QGIS. When
you commit the changed records are iterated and a separate UPDATE statements is
issued for each record. And that's what takes long - although an expensive
view contributes to the time each UPDATE needs.
The comment about the attribute table was just because for each change also the
attribute table(s) UI is updated. So execution might be quicker if there are
no open attribute tables. But the actual updates should perform the same from
wherever you invoke the feature calculator.
Jürgen
_______________________________________________
Qgis-user mailing list
[email protected]
List info:http://lists.osgeo.org/mailman/listinfo/qgis-user
Unsubscribe:http://lists.osgeo.org/mailman/listinfo/qgis-user
_______________________________________________
Qgis-user mailing list
[email protected]
List info:http://lists.osgeo.org/mailman/listinfo/qgis-user
Unsubscribe:http://lists.osgeo.org/mailman/listinfo/qgis-user
--
Matthias Kuhn
OPENGIS.ch -https://www.opengis.ch
Spatial • (Q)GIS • PostGIS • Open Source
_______________________________________________
Qgis-user mailing list
[email protected]
List info: http://lists.osgeo.org/mailman/listinfo/qgis-user
Unsubscribe: http://lists.osgeo.org/mailman/listinfo/qgis-user
_______________________________________________
Qgis-user mailing list
[email protected]
List info: http://lists.osgeo.org/mailman/listinfo/qgis-user
Unsubscribe: http://lists.osgeo.org/mailman/listinfo/qgis-user
