Re: 980121: NetCDF-3.3.1 and fortran compiler option -U

Randy,

>Date: Mon, 16 Mar 1998 16:25:27 -0500 
>From: address@hidden (Randy Zagar)
>Organization: . University of Delaware/College of Marine Studies
>To: address@hidden
>Subject: Re: 980121: NetCDF-3.3.1 and fortran compiler option -U 
>Keywords: 199803162125.OAA27405 

In the above message, you wrote:

> 1.    A *long* while back, I mentioned that I/we were having problems
>       compiling a CFD model with NetCDF because its' Makefile insisted
>       on using a '-U' option to preserve the case of all the symbols...
> 
>       The recommended course of action we came up with was to take all
>       the symbol names in "netcdf.inc" and make them lower-case to
>       match what was in the library.

This has also been done in the latest netCDF release (3.4).

>       I can confirm that this works and, furthermore, does not seem
>       to interfere with anything else I've written...  The URL for
>       the "netcdf.inc" I am now using is:
> 
>               ftp://newark.cms.udel.edu/pub/users/zagar/netcdf.inc
> 
> 
> 2.    Onto another issue entirely... I recently solved a problem I've
>       been having with 'packed' variables.  I'm writing Perl programs
>       to convert datafiles to NetCDF, and in some cases it's appropriate
>       to store the data as 'BYTE' or 'SHORT' and have the 'scale_factor'
>       and 'add_offset' attributes as 'FLOAT'.
> 
>       The documentation says that 'valid_min', 'valid_max', and
>       'valid_range' should be the same type as the packed variable,
>       but I find it convenient to make them the same type as the
>       'scale_factor' and 'add_offset' attributes...
> 
>       From what you've seen, which way do other NetCDF developers
>       handle this issue ?

I haven't seen very much of the work of other netCDF developers, so I
don't feel qualified to comment on this issue.  You might try posting
this question to the netcdf mailing-list.

>       Am I begging for trouble if I continue doing it this way ?
> 
>       Lastly, the documentation doesn't address the '_FillValue'
>       attribute in the context of 'packed' variables.  It has taken
>       me two days to discover that '_FillValue' *must* be the same
>       type as the unscaled variable and *NOT* the same as that for
>       'scale_factor' and 'add_offset'.  It wouldn't be difficult to
>       mention this in the docs.
> 
> 
> 3.    Adding a -perl option to 'ncgen' would be *wonderful*...

I'll pass that on.

> 
> 
> Anyway, NetCDF is great stuff.  You guys are doing great work and I
> just wanted to know that I *really* appreciate your efforts.

Thanks.

> 
> 
> -Randy
> 
> 
> 
> p.s.  While composing this e-mail, it occurred to me that the whole
>       concept of a products' learning curve is a lot like Sissiphus
>       in Greek mythology.  By the time you get the rock up the hill,
>       they've changed the standard on you and you have to do it all
>       again...  Thankfully, NetCDF isn't written by Microsoft.
> 
> ===
> 
> Randy Zagar                          E-Mail: address@hidden
> Sr. Scientific Programmer            E-Mail: address@hidden
> College of Marine Studies             Voice: (302) 831-1139
> University of Delaware                  FAX: (302) 831-6838
> Newark, DE 19716

--------
Steve Emmerson   <http://www.unidata.ucar.edu>