[postgis-users] ST_Mean4ma , ST_StdDev4ma rounding down

[Pierre Racine]

The two NOTICE are because there is actually two rasters created by in the dummy_rast table and the first one has no bands.

Values are rounded by ST_MapAlgebra() because, by default, ST_MapAlgebra() will try to produce a raster with the same pixel type as the input rasters. If you want the resulting raster to receive float values you must tell ST_MapAlgebra() by passing it '32BF' as pixel type.

Try this:

select (st_dumpvalues(st_mapalgebra(array[row(rast, 1), row(rast, 2), row(rast, 3)]::rastbandarg[], 'st_mean4ma(double precision[], int[], text[])'::regprocedure, '32BF'))).* from (select st_union(rast) as rast from dummy_rast) as foo;

Pierre

> -----Original Message-----
> From: postgis-users [mailto:postgis-users-bounces at lists.osgeo.org] On
> Behalf Of Severin Thaler
> Sent: Saturday, February 20, 2016 10:26 PM
> To: PostGIS Users Discussion
> Subject: [postgis-users] ST_Mean4ma , ST_StdDev4ma rounding down
> 
> Hi all,
> 
> here’s the problem.
> after creating the dummy raster : http://postgis.net/docs/manual-
> 2.1/RT_reference.html
> i ran:
> 
> 
> select (st_dumpvalues(rast)).* from dummy_rast;
> 
> and get:
> 
> NOTICE:  Raster provided has no bands
> NOTICE:  Raster provided has no bands
>  nband |                                                    valarray
> -------+-------------------------------------------------------------------------------------------------
> ----------------
>      1 |
> {{253,254,253,254,254},{253,254,254,253,249},{250,254,254,252,249},{25
> 1,253,254,254,253},{252,250,254,254,254}}
>      2 |
> {{78,98,122,173,209},{96,118,180,249,254},{99,112,169,245,254},{89,99,1
> 22,176,229},{79,88,97,112,135}}
>      3 |
> {{70,86,100,135,161},{80,108,162,227,250},{90,108,175,251,254},{77,86,1
> 09,164,203},{62,69,76,86,101}}
> (3 rows)
> 
> which by itself is kinda weird and that is why i unionized first, i.e.
> 
> 
> select (st_dumpvalues(st_union(rast))).* from dummy_rast;
> 
> and now get a more proper output:
> 
>  nband |                                                    valarray
> -------+-------------------------------------------------------------------------------------------------
> ----------------
>      1 |
> {{253,254,253,254,254},{253,254,254,253,249},{250,254,254,252,249},{25
> 1,253,254,254,253},{252,250,254,254,254}}
>      2 |
> {{78,98,122,173,209},{96,118,180,249,254},{99,112,169,245,254},{89,99,1
> 22,176,229},{79,88,97,112,135}}
>      3 |
> {{70,86,100,135,161},{80,108,162,227,250},{90,108,175,251,254},{77,86,1
> 09,164,203},{62,69,76,86,101}}
> (3 rows)
> 
> however, my real concern is that when i compute the average over the 3
> bands:
> 
> select (st_dumpvalues(st_mapalgebra(array[row(rast, 1), row(rast, 2),
> row(rast, 3)]::rastbandarg[], 'st_mean4ma(double precision[], int[],
> text[])'::regprocedure))).* from (select st_union(rast) as rast from
> dummy_rast) as foo;
> 
> 
> i get
> 
> 
>  nband |                                                    valarray
> -------+-------------------------------------------------------------------------------------------------
> ----------------
>      1 |
> {{133,146,158,187,208},{143,160,198,243,251},{146,158,199,249,252},{13
> 9,146,161,198,228},{131,135,142,150,163}}
> (1 row)
> 
> i.e. the actual means but rounded down, similarly for the standard
> deviation callback function.
> is this a bug because it shouldn’t be rounded down, we should have the
> decimals too.
> 
> thanx for feedback