[postgis-devel] [PostGIS] #253: Gist causes wrong result from ~=

[PostGIS]

#253: Gist causes wrong result from ~=
----------------------+-----------------------------------------------------
  Reporter:  nicklas  |       Owner:  pramsey      
      Type:  defect   |      Status:  new          
  Priority:  medium   |   Milestone:  postgis 1.4.1
 Component:  postgis  |     Version:  1.3.X        
Resolution:           |    Keywords:               
----------------------+-----------------------------------------------------
Comment (by nicklas):

 ok
 I don't understand everything from the discussion above, but I have two
 questions, that at least to me seems relevant :-)
 If the problem is this RECHECK thing, why isn't it working as expected
 from Paul's comment in r3287. I read it as he is putting that recheck flag
 there, instead of in the operator class definition. If the recheck doesn't
 work as expected I guess that's a good explanation of the unpredictable
 behavior of the bug from Paul's comment that it is the conversion from
 float8 to float4 that makes it nessecary. But at the other side I think
 that should only cause a little bit to many true-answers (because all
 boxes is rounded to be a little bigger) for the bbox-comparation. Why does
 it make the planner choose not to do the vertex to vertex comparasion.

 the second question, can this recheck-thing explain that there is
 different answer between select-part and where-part of the query. That
 part of the issue worries me a bit.

 and the third unannounced question.
 Do we (read you) have control that no other functions are affected so we
 don't just hide the symtom instead of finding the cause.

 /Nicklas

-- 
Ticket URL: <http://trac.osgeo.org/postgis/ticket/253#comment:26>
PostGIS <http://trac.osgeo.org/postgis/>
PostGIS