Cannot ANALYZE table w/ valid features in it

[pramsey]

The features are admittedly a bit awful

 POINT ZM (1463702.128 517438.510000008 0 -1.79769313486232e+308)
 POINT ZM (1234886.324 454719.691000004 0 -1.79769313486232e+308)
 POINT ZM (1230629.645 468753.839000007 0 -1.79769313486232e+308)
 POINT ZM (1243683.988 466816.384999998 0 -1.79769313486232e+308)

however, there's about 600 of them, and they do have coordaintes, so no reason to not analyze them, this appears to be exercising some kind of shortcoming in the analyze logic.

[pramsey]

point shape file

[robe]

We can add this to regress tests I guess after you have tested.

CREATE TABLE test_awful(gid serial, geom geometry);
INSERT INTO test_awful(geom)
VALUES ( 'POINT ZM (1463702.128 517438.510000008 0 -1.79769313486232e+308)'::geometry),
 ('POINT ZM (1234886.324 454719.691000004 0 -1.79769313486232e+308)'::geometry),
 ('POINT ZM (1230629.645 468753.839000007 0 -1.79769313486232e+308)'::geometry ),
 ('POINT ZM (1243683.988 466816.384999998 0 -1.79769313486232e+308)'::geometry );

analyze test_awful;

I assume the notice:

NOTICE:  no non-null/empty features, unable to compute statistics

is what you are referring to.

[pramsey]

OK, so the computation works fine in 2D mode, but in 4D mode it runs afoul of this little code block:

 if ( ! gbox_is_valid(&gbox) )
  {
    POSTGIS_DEBUGF(3, " skipped infinite/nan geometry %d", i);
    continue;
  }

The super-huge M value gets bumped up to Inf in float space, so the box is, from the point-of-view of the stats system, not handle-able. I'm going to close this as won't-fix since hte workaround is just to force2d the crapping values off your geometries.

[strk]

I recall the stats gatherer had code to skip "hard deviants", could this be the case for the M value ?