[postgis-tickets] [PostGIS] #2777: [raster] Huge literal in enforce_max_extent_rast check constraint

[PostGIS]

#2777: [raster] Huge literal in enforce_max_extent_rast check constraint
--------------------+-------------------------------------------------------
 Reporter:  strk    |       Owner:  dustymugs
     Type:  defect  |      Status:  new      
 Priority:  medium  |   Milestone:           
Component:  raster  |     Version:  trunk    
 Keywords:          |  
--------------------+-------------------------------------------------------

Comment(by robe):

 I don't think it's worth changing.  In raster it can't have any number of
 vertices, and I'm more concerned about breaking people's assumptions and
 forcing them to rebuild all their indexes if we do this.  And even though
 the end seems like the same, to the planner it is not so it would never
 use an existing index unless we require everyone to drop thier indexes and
 reindex for 2.2.

 I think originally a long time ago we use ST_Envelope, but the fundamental
 difference bewteen the ST_Envelop and ST_ConvexHull is that ST_ConvexHull
 handles rotated rasters and will give the right answer.  So it saved a bit
 in those functions calls where we know there is no nodata because those
 type of functions would naturally use the index.  If you change then they
 have to do both an ST_Envelop and an ST_ConvexHull to utilize index.

 Other issue before which I don't think is relevant now is that ST_envelope
 was float4 rather than float4.  I wish Pierre was here to explain.  I
 don't recall the long discussion of why this makde ST_Envelope kinda
 useless.

-- 
Ticket URL: <http://trac.osgeo.org/postgis/ticket/2777#comment:17>
PostGIS <http://trac.osgeo.org/postgis/>
The PostGIS Trac is used for bug, enhancement & task tracking, a user and developer wiki, and a view into the subversion code repository of PostGIS project.