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, I'm still having issue with this even as of r14198 even though dusty has made some changes.

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.

---

[robe]

Bborie,

I'm guessing you won't have time to look at this within the next day or so, so pushing to 2.3. If you do feel free to push back. I'll push to 2.2.2 after release.

[dustymugs]

Wow I forgot about this… I'll take a look tonight.

[robe]

okay I'll switch back then. Hurry Hurry PostgreSQL 9.5.0 is already been stamped.

[robe]

Milestone renamed