ST_Area(geography) returning wrong answer

[pramsey]

Reported in the mailing list

WITH shapes as (
  select 1 as rownum, 
    ST_GeogFromText('POLYGON((-144 -63,-144 -54,-126 -54,-126 -63,-144 -63))') as shape
  union
  select 2 as rownum, 
    ST_GeogFromText('POLYGON((-90 -63,-90 -54,-72 -54,-72 -63,-90 -63))') as shape
)
select 
  rownum, 
  ST_Area(shape) as spheroid, 
  ST_Area(shape, false) as sphere,
  ST_Area(shape) - ST_Area(shape, false) as delta
from shapes;

The meridian crossing code of the current implementation is tripping.

[pramsey]

Try trunk at r10407

[pclark]

This works for me using r10420; hadn't tried r10407. The results aren't identical, but they're close enough for what I care about - the discrepancy between the two polygons is on the order of 500 square cm, as opposed to twice the area of California like they were in postgis 2.0.1.

 WITH shapes as (
  select 1 as rownum, ST_GeogFromText('POLYGON((-144 -63,-144 -54,-126 -54,-126 -63,-144 -63))') as shape
  union
  select 2 as rownum, ST_GeogFromText('POLYGON((-90 -63,-90 -54,-72 -54,-72 -63,-90 -63))') as shape
  )
  select rownum, ST_Area(shape) as spheroid, ST_Area(shape, false) as sphere,
        ST_Area(shape) - ST_Area(shape, false) as delta
  from shapes;

returns

  rownum |     spheroid     |      sphere      |      delta       
 --------+------------------+------------------+------------------
       1 | 1040833443426.79 |  1035406049943.1 | 5427393483.68726
       2 | 1040833443426.79 | 1035406049943.05 | 5427393483.73877
 (2 rows)