ST_Azimuth with geography type returns incorrect value

[francoisb]

It seems #4718 is a regression.

Perfect west 270° (3π/2 rad) - the only value in the regression test from tickets.sql which had a negative result in the previous code - is by chance correct, but all other expected results in the range 180°-360° (π-2π rad) are incorrect.

Second, a perfect north (value exactly 0) is now incorrectly returned as 180° (π rad), instead of 0.

The fix is simple. Expected formula in lwgeodetic.c line 2180 is:

        return az < 0 ? az + 2.0 * M_PI : az;

instead of:

        return az > 0 ? az : M_PI - az;

Regression test from tickets.sql should add additional values, say at 45° (Northeast, π/4) and at 315° (Northwest, 7π/4).

Documentation for ST_Azimuth could be extended to read:

North = 0; Northeast = π/4; East = π/2; Southeast = 3π/4; South = π; Southwest 5π/4; West = 3π/2; Northwest = 7π/4

instead of:

North = 0; East = π/2; South = π; West = 3π/2

[francoisb]

This is a duplicate of #4840, actually.

The documentation change suggestion could still be applied, though.

[pramsey]

Milestone renamed

[robe]

In prep for 3.1.3 release

[Regina Obe <lr@…>]

In cb8b6b8/git:

Update ST_Azimuth to include more defs suggested by Francois Bonzon. References #4880 for PostGIS 3.1.4

[Regina Obe <lr@…>]

In 60314f5/git:

Update ST_Azimuth to include more defs suggested by Francois Bonzon. Closes #4880 for PostGIS 3.2.0