[netCDF #SAI-630695]: ncgen mishandles character variables with NULs

Hi Chuck,

> We have long used arrays of strings in our files with NUL characters to
> delimit the strings. Here is an example, which can be created with the
> attached demo program:
> 
> netcdf simple_char {
> dimensions:
> n = 8 ;
> variables:
> char data(n) ;
> data:
> 
> data = "abc\000def" ;
> }
> 
> 
> This has always worked well for us. But it turns out that if
> ncgen/ncgen3 attempts to process this file, all of these character
> variables are truncated at the first NUL, rendering the new file corrupt:
> 
> netcdf simple_char {
> dimensions:
> n = 8 ;
> variables:
> char data(n) ;
> data:
> 
> data = "abc" ;
> }
> 
> 
> I've reproduced it with versions 3.6.2 and 4.2.1.1. I've read about
> related issues regarding attributes in the support forums, but it seems
> in cases like this ncgen should be symmetric with ncdump and faithfully
> record the embedded NULs. Any thoughts? We've rarely used ncgen, so this
> has not been a real issue. But lately we've had a couple cases where
> this limitation has caused some heartburn.
> 
> Thanks, and let me know if I can provide more information.

I just checked netCDF version 2.4.3 from 1996, and the ncgen then behaved
the same way, not preserving data beyond null bytes in char variables.

So it looks like a bug that's been around for at least 17 years, and I 
guess we ought to get around to fixing it :-) .  I'll enter a bug ticket
for it soon, unless someone can convince me it's a feature ...

--Russ

Russ Rew                                         UCAR Unidata Program
address@hidden                      http://www.unidata.ucar.edu



Ticket Details
===================
Ticket ID: SAI-630695
Department: Support netCDF
Priority: Normal
Status: Closed