Hi John,
>From what I understand, parallel I/O is a netcdf-4 specific feature, and
that would have to be the big one I'm interested in.
I have worked in 2007 on a project with distributed runs whose outputs
were then collated from many (text!) files. I/O was the clear bottleneck
in several modelling steps. I am thus interested in better techniques to
handle parallel I/O in write mode.
Chunking attracted my attention but I do not have a clear idea whether
this would alleviate known issues and how. Other features missing in
write mode in the Java implementation (e.g. groups if I get it right)
sound nice but do not address main issues (maybe only because I am not
aware of them).
The choice of Java vs C or C++ is based on partiality and familiarity so
far: I needed to reuse some C# code, and the netcdf Java jar files
converted to .NET assemblies with IKVM (www.ikvm.net) worked well so
far. Unless there is an incentive to use the C libs (performance,
feature set, etc.) over Java I think most of my colleagues would prefer
the latter.
Jean-Michel
-----Original Message-----
From: John Caron [mailto:caron@xxxxxxxxxxxxxxxx]
Sent: Tuesday, 12 August 2008 6:45 AM
To: Perraud, Jean-Michel (CLW, Black Mountain)
Cc: netcdf-java@xxxxxxxxxxxxxxxx
Subject: Re: [netcdf-java] Roadmap or intent for netcdf-java to write
netCDF4
Hi Jean-Michel:
Can you specify exactly what features you are interested in being
available in C library? Context such as why C vs Java would be helpful.
John
Jean-Michel.Perraud@xxxxxxxx wrote:
> Hi,
>
> I have been testing netcdf-java 4.0 alpha to read spatio-temporal
'legacy' data and write it to netcdf3. After a bit of trial and error
starting from the tutorial it worked great and I liked the 'feel' of the
API, hence an interest for longer term use in other projects.
>
> Is there an update on the possibility of having the same features as
the C API notably writing to netCDF4? Without elaborating I think I'll
need it for some purposes. I can fathom the difficulty involved, as
conveyed in
http://www.unidata.ucar.edu/mailing_lists/archives/netcdf-java/2006-Sept
ember/000212.html.
>
> I'd appreciate if someone at Unidata could give me their best guess on
that feature, if possible. No strings attached with the answer, of
course.
>
> Cheers
> Jean-Michel Perraud
> CSIRO Land and Water
>
>
> _______________________________________________
> netcdf-java mailing list
> netcdf-java@xxxxxxxxxxxxxxxx
> For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
