[ncl-talk] Bug report, named dimensions and Netcdf-4

[David Brown]

Hi Dave,
Thanks for the report. We will look into it.
 -dave

On Mon, Dec 7, 2015 at 1:31 PM, Dave Allured - NOAA Affiliate
<dave.allured at noaa.gov> wrote:
> NCL team,
>
> I found this problem with named dimensions in NCL 6.3.0 on both Linux and
> Mac OS.  The dimension name reference operator "!" gets the wrong dim name
> when used with a Netcdf-4 file variable, under some circumstances.  Can you
> please fix this for the next release?  Thank you.
>
> Sample input file:
>
>> ncdump -hs test5.nc
> netcdf test5 {
> dimensions:
>     lat = 3 ;
>     lon = 4 ;
>     time = 2 ;
> variables:
>     float time(time) ;
>         time:_Storage = "contiguous" ;
>     float lon(lon) ;
>         lon:_Storage = "contiguous" ;
>     float lat(lat) ;
>         lat:_Storage = "contiguous" ;
>     float precip(time, lat, lon) ;
>         precip:_Storage = "contiguous" ;
>
> // global attributes:
>     :_Format = "netCDF-4" ;
> }
>
> Bug demonstration.  The correct answer, "time", is obtained with the second
> and third methods shown:
>
>  NCAR Command Language Version 6.3.0
> ncl 0> f = addfile ("test5.nc", "r")
> ncl 1> print (f->precip!0)
> (0)     lat                    ******** WRONG RESULT
> ncl 2> x = f->precip
> ncl 3> print (x!0)
> (0)     time
> ncl 4> dims = getfilevardims (f, "precip")
> ncl 5> print (dims(0)+"")
> (0)     time
>
> Notes:
>
> 1.  The problem seems specific to only the full Netcdf-4 format.  I did not
> see this problem with Netcdf-3 or Netcdf-4 classic.  My testing was limited.
>
> 2.  The problem depends on the order of dimensions as originally defined in
> the Netcdf-4 file.  In the demo above, notice that time is the first
> dimension in the data variable, but it is NOT the first defined dimension in
> the file.  When time is the first defined dimension, the problem goes away.
>
> 3.  It does not make any difference whether the storage mode is contiguous
> or chunked.
>
> --Dave
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>