transform() grid-shift 2nd chance logic defective

[william....@…]

What steps will reproduce the problem?

select x(transform(pointfromtext('POINT (173 52)',4269),4267)), y(transform(pointfromtext('POINT (173 52)',4269),4267)); WARNING: transform: -38 (failed to load NAD27-83 correction file) ERROR: transform: couldn't project point: -38 (failed to load NAD27-83 correction file) <SELECT aborted; no output rows>

What is the output as corrected?

select x(transform(pointfromtext('POINT (173 52)',4269),4267)), y(transform(pointfromtext('POINT (173 52)',4269),4267)); WARNING: transform: -38 (failed to load NAD27-83 correction file) WARNING: transform: -38 (failed to load NAD27-83 correction file)

x | y

---

173 | 51.9999999990852

(1 row)

Please use labels and text to provide additional information.

The PostGIS transform() function contains logic that intends to substitute a '2nd chance' non-shifted transformation for an unavailable grid-shift transformation (e.g., NAD83 to NAD27 transformation of a point outside of the area covered by the NAD27 grid-shift table, or in the absence of that table). The logic handling this situation is located in the transform_point() function of lwgeom_transform.c but seems outdated with respect to how recent versions of proj.4 handle the situation.

This discord causes a 'SELECT ... transform(geom,srid) ... FROM mytab' to abort and be rolled back where even a single row of table mytab contains a geometry geom that lies outside the bounds of an applicable grid-shift table. Running on Postgres 8.3.5, PostGIS 1.3.3, and proj.4 4.6.1, such a SELECT aborts with the twin messages shown in the problem reproduction above.

Attached is a patch that seems to be a reasonable solution for this problem.

Although this patch works for me, please scrutinize it carefully. This is my first look at PostGIS and proj.4 source code. Worse, I code in C very rarely and am especially error-prone when using C pointers and indirection.

I have not been able to fabricate data that exercises every branch of this

patch, so it may still contain errors.

Thanks so much for PostGIS! It's a tremendous resource that contributes materially to the public good.