Re: Vector subscripting on a file variable

Hi Dave,

I think the warning against named subscripting has to do with the
fact that it is usually used to reorder the dimensions of a variable.
Reordering the
dimensions of a variable as it is being read from a file can be very
slow because it can result in huge numbers of small reads scattered
throughout the file.

Your situation is quite different because you are using vector
subscripting on the 'slow' dimension. A complete lat /lon slice can
most likely be read from
contiguous elements in the file. I'm guessing the performance will be
nearly the same as if you were reading a fully contiguous chunk using
regular subscripting.
If you instead used vector subscripting on the lon dimension, an
impact on performance would be more likely.
  -dave

On Mar 3, 2008, at 5:43 PM, Dave Allured wrote:

> I would like to read selected records from a file variable. I
> would like to use vector subscripting to read selected records and
> ignore other records. Something like:
>
> f = addfile ("infile.nc", "r")
> record_nums = (/ 9, 10, 11, 21, 22, 23 /)
>
> data = f->precip(record_nums, :, :) ; time, lat, lon
>
> My question is whether using vector subscripting on a large file
> variable is reasonably efficient. NCL documentation warns against
> using named subscripting on a file variable, and this seems related.
>
> My index array "record_nums" is guaranteed to be monotonic and
> increasing. Also in this case, the first subscript is the actual
> Netcdf record dimension. Thank you for any advice.
>
> Dave Allured
> CU/CIRES Climate Diagnostics Center (CDC)
> http://cires.colorado.edu/science/centers/cdc/
> NOAA/ESRL/PSD, Climate Analysis Branch (CAB)
> http://www.cdc.noaa.gov/
> _______________________________________________
> ncl-talk mailing list
> ncl-talk_at_ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk

_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Mon Mar 03 2008 - 18:29:50 MST