v.extract: where letting all pass on double prec

[hamish]

Hi,

I'm trying to clear out some vector points where the depth_m double prec attribute column is set with null= 9999.9. so I try

v.extract in=nav_points out=nav_points_clean_depths \
   where='depth_m < 9999'

but I still get all 129k points (confirmed by v.univar). v.db.select where= | wc -l shows the correct (cleaned) number, so I could cleanse it though v.db.select|v.in.ascii as a work-around.

A quick test with v.random -z zmin= zmax= with 1000 points seemed to work fine.

the highest real point in it is around 2300, and even with < 500 many fewer points are written but still some 9999.9s get through. ??! index off by one error??

seen in 6.4.0 and 6.5svn, haven't tested in trunk.

data avail privately upon request.

?, Hamish

[hamish]

v.edit tool=delete where='depth_m = 9999.9' gets 1393 features deleted, but on the orig file I only see 113:

g.copy v=nav_points,nav_points_clean_depths

v.edit nav_points_clean_depths tool=delete where='depth_m = 9999.9'
Selecting features...
1393 of 129097 features selected from vector map
<nav_points_clean_depths@user1>
 100%
1393 features deleted
...

but:

cut -f3 -d'|' nav.points.dat | grep -c '9999.9'
113

v.db.select nav_points | grep -c '|9999.9|'
113

v.db.select -c nav_points where='depth_m = 9999.9' | wc -l
113

?!?, Hamish

[hamish]

the wierdness was due to non-unique cat numbers among the points.

Hamish