[raster] ST_MapAlgebraFct for two rasters

[dustymugs]

The sibling to 2-raster ST_MapAlgebraExpr. Instead of a set of expressions, use a function call.

1. st_mapalgebrafct(

rast1 raster, band1 integer, rast2 raster, band2 integer, userfunction regprocedure, pixeltype text DEFAULT NULL, extenttype text DEFAULT 'INTERSECTION', VARIADIC userargs text[] DEFAULT NULL

2. st_mapalgebrafct(

rast1 raster, rast2 raster, userfunction regprocedure, pixeltype text DEFAULT NULL, extenttype text DEFAULT 'INTERSECTION', VARIADIC userargs text[] DEFAULT NULL

)

3. st_mapalgebrafct(

rast1 raster, rast2 raster, userfunction regprocedure, VARIADIC userargs text[]

)

It'll be interesting to compare performance between this function and ST_MapAlgebraExpr.

[pracine]

I guess performance will be comparable only if you PREPARE the query parsing the expression in ST_MapAlgebraExpr. I did some test with the one raster version (which was not PREPARING the query at the moment) and ST_MapAlgebraFct was 7 time ST_MapAlgebraExpr for the same expression.

[dustymugs]

I'm suspecting that 2-raster ST_MapAlgebraFct will still be faster than 2-raster ST_MapAlgebraExpr. I'll be committing the code shortly. Development went quickly as both ST_MapAlgebraExpr and ST_MapAlgebraFct call the same underlying function, RASTER_mapAlgebra2.

[pracine]

Are you PREPARING the query in ST_MapAlgebraExpr?

[dustymugs]

That's just my hunch as to the performance. I've yet to compare the two as I'm just finishing up regression testing and valgrinding. All expressions are prepared in 2-raster ST_MapAlgebraExpr unless the expression has no parameters (which would cause the process to crash).