ST_MapAlgebra with mask seems to hard-code distance 1 instead of inferring distance from mask matrix

[robe]

I've moved this from the now closed mask documentation #2603 to here

DROP TABLE IF EXISTS shapes;
CREATE TABLE shapes(rid integer, rast raster);  

INSERT INTO shapes(rid,rast)
VALUES ( 1, ST_AsRaster(
		ST_Buffer(
			ST_GeomFromText('LINESTRING(50 50,150 150,150 50)'), 10,'join=bevel'), 
			200,200,ARRAY['8BUI'], ARRAY[118], ARRAY[0]) );

 SELECT st_mapalgebra(rast,1,'st_mean4ma(double precision[], int[], text[])'::regprocedure,'{{1,0,1,1,1}, {1,0,1,1,1}, {1,0,1,1,1}, {1,0,1,1,1}, {1,0,1,1,1}}'::double precision[],false)
FROM shapes;

Gives error:

ERROR:  rt_pixel_set_array: mask dimensions 5 x 5 do not match given dims 3 x 3

It's my understanding based on dusty's comments that the neighborhood size should be inferred from the mask size, so there seems to me something wrong here.

Bborie's comments copied from other ticket

---

Distance is two pixels left, right, up, down from the pixel of interest...

2 2 2 2 2
2 1 1 1 2
2 1 0 1 2
2 1 1 1 2
2 2 2 2 2

So the distance x/y are correct. As for that error message, I'll need to dig.

---