+1 for dropping the filter -- trust the user ! --strk;
On Tue, Jul 10, 2012 at 05:10:23PM +0200, Vincent Picavet wrote: > Hi, > The postgres provider apparently does a check on the input query and only > tolerates queries beginning with "SELECT". > > The more general SQL query form allows for queries not beginning with select, > but still being valid, as it is the case for CTEs. The only way to have qgis > load them is to wrap CTEs inside a select, which is not really nice. > > Example. > This won't load in qgis : > > with points as (select gid, st_makepoint(random() * 10, random() * 10) as > the_geom from generate_series(1, 100) as gid) select * from points > > But this will : > > select * from (with points as (select gid, st_makepoint(random() * 10, > random() * 10) as the_geom from generate_series(1, 100) as gid) select * from > points) as foo > > This is also the case for other kind of queries like UPDATE... RETURNING, > which should be valid but don't conform to the filter. > > Is the filter useful anyway, since there are so many ways to write non-valid > queries ? > > So the question is : do we get rid of the SELECT filter in postgres provider > ? > > Those in favor of keeping it raise your hands and arguments now, or I think > we'd better drop it... _______________________________________________ Qgis-developer mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/qgis-developer
