[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Java netCDF 2.2 / Structures





Norman Fomferra wrote:
Hi John,

Thank you very much for your help.
Is a support for "real" HDF5-like structures planned? And if so, when?

I assume you mean when can you write Structures? That would be when the 
netcdf-4 file structure is final, which is ...

waiting for the HFD5
version 1.8 release before the format will be supported.  The alpha
release uses some format artifacts that will not be required when HDF5
1.8 is released from NCSA, and we don't want to continue to support
those artifacts.

HDF5 version 1.8 is expected soon, maybe in April if there are no
unanticipated problems.

For more details on netCDF-4, a paper is available here in PDF or Word
form:

  http://www.unidata.ucar.edu/software/netcdf/papers/2006-ams.pdf
  http://www.unidata.ucar.edu/software/netcdf/papers/2006-ams.doc

and a 15-minute recorded presentation given in January is available
here:

  http://ams.confex.com/ams/Annual2006/techprogram/paper_104931.htm

There will probably be another month or so after that to get the changes into 
the Netcdf-Java library


In general, we prefer the use of NetCDF over HDF because of its pure
Java API. We even still have some unsolved problems with the native
HDF libs. So hopefully, UNIDATA will continue working on the NetCDF
specs and Java implementation.

Yes we will!!


There is also still room for runtime performance improvements.
Especially the frequent reallocation of arrays, lists of ranges and
indexes in the API can make up to 20% of processing time besides pure
java.io calls (-Xprof).

For sure. I hope to do a "performance improvement" release, but it probably has 
to wait for a bunch of other stuff, including netcdf-4.

Please feel free to let me know of any specific slowdowns, and i will 
investigate when i can.

Thanks for your feedback!