Problem (UDUnits?) with hyperslab specs.

[Charlie Zender]

Hi Rorik,

I've add a new TODO, #261, which may be a problem with the UDUnits
implementation. It may not be, but I'd appreciate it if you would
verify this and implement a fix if it's UDUnits-related.
Specifying hyperslabs using floating point notation appears to
trigger the UDUnits algorithms, which can then fail in a variety of 
ways. The following should be legal (right?) and work but doesn't:

zender@dust:~$ ncks -H -d lon,0.63e-6 ~/nco/data/in.nc
var_id: 5, att_nm: units
nco_err_exit(): ERROR nco_inq_att
Attribute not found

Thanks,
Charlie

[Rorik Peterson]

Yeah, this is a UDUnits problem.  When testing for the dimesion specification type, it thinks the dash means a date string (i.e. 1999-1-1).  I'll fix it.

rorik

[Rorik Peterson]

Charlie,
I changed the scheme that determines the dimension specification limit type.  It works on your reported bug, but maybe check that sahara dust problem listed in the TODO

rorik

rorik@chabuku:~/nco$ src/nco/ncks -H -C -d time_udunits,"1999-12-08 12:00","1999-12-08 18:00" -v time_udunits data/in.nc
time_udunits[0]=876012
time_udunits[1]=876018

rorik@chabuku:~/nco$ src/nco/ncks -H -C -d wvl,"1 picometer","1 furlong" -v wvl data/in.nc
wvl[0]=5e-07
wvl[1]=1e-06

rorik@chabuku:~/nco$ src/nco/ncks -H -d lon,0.63e-6 -v lon data/in.nc
lon[0]=0

[Charlie Zender]

Hi Rorik,

Your patch fixes both problems.
I'm removing this from the TODO list.

Thanks!
Charlie