[postgis-devel] [PostGIS] #1536: [raster] There should be an option for certain variants of ST_Intersection(raster, raster)

[PostGIS]

#1536: [raster] There should be an option for certain variants of
ST_Intersection(raster, raster)
---------------------+------------------------------------------------------
 Reporter:  pracine  |       Owner:  pracine      
     Type:  defect   |      Status:  new          
 Priority:  low      |   Milestone:  PostGIS 2.0.0
Component:  raster   |     Version:  trunk        
 Keywords:           |  
---------------------+------------------------------------------------------

Comment(by pracine):

 #1537 fixed...

 Even if you restraint the extent to the intersection you must do something
 for when the provided raster do not have a nodata value defined. If I
 intersect a raster having nodata value with a circle, for example, I want
 to have some control over the nodata value assigned to the resulting
 raster. Doing ST_SetBandNoDataValue() afterward is not sufficient as all
 the nodata values, in this case, have been set to 0. I have no other way
 than calling ST_MapAlgebraExpr another time to convert all of them to the
 nodata value I wish.

 As for ST_Clip() ST_Intersection(raster, geometry) should let us provide a
 nodata value.

 The same logic applies to ST_Intersection(raster, raster). What if I want
 the two bands, the second raster has -10 as nodata value, many of then in
 the intersecting area and the first raster has no nodata value defined
 (and worse many pixel having the value -10)? What should be the nodata
 value assigned to the first raster?

 I think we should have a parameter for each raster to define the resulting
 nodata value. If I don't provide it then ST_MinPossibleValue() should be
 used (at the risk of transforming some withdata values into nodata
 values).

 I think ST_MapAlgebra provides all the tools for ST_Intersection to be
 able to define or redefine the nodata value. You have to set those values
 using the alternate expressions and then redefine the nodata value with
 ST_SetBandNodataValue().

 You might say then there should also be a nodata value paramater to
 ST_MapAlgebraExpr but no: such a parameter could be conflictual with the
 value set with the alternate expressions. It is the role of the caller to
 set the nodata value to the same defined with the alternate expressions.

 What do you think?

-- 
Ticket URL: <http://trac.osgeo.org/postgis/ticket/1536#comment:7>
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.