Re: nj22 and conventions


Thanks much for looking into this.

John Caron wrote:

Hi Roland:

GDT is a precurser to CF, so its likely it would be a straighforward extension of existing classes. IRIDL I think is Benno's, and Ive asked him about it.

This jibes with my understanding GDT. It occurs to me now that these particular files that I'm working might read just fine as geo-referenced files if they were re-labeled with CF-1.0 since they appear to be pretty simple. I will check this out as a first step. Extending the CF code is the right long-term answer.

Benno and I had a long talk about the IRIDL "convention" at AGU, That conversation in part prompted my message. A great many of the gridded files served by the IRI data library appear to be COARDS (some are not -- they have mixed up the dimension ordering or are clearly non-COARDS-ian for other more complicated reasons), but none of them are not labeled COARDS. It seems to me a great loss not to be able to read those files which are clearly COARDS compatible with the enhanced netCDF dataset nj22 code.

Benno provides a "COARDS filter", but it's not hooked up to his THREDDS catalog as far as I know. I want to be able to descend the THREDDS tree and find all the datasets with grids that can be geo-referenced so it seems that being able to read the IRIDL convention is the shortest path for me. I'll be interested to hear what Benno has to say on the topic.

The ideal case is that it is easy enough for others to add these Conventions parsers (aka "coordinate system builders"). I will work on documentation and examples.

All javadoc is now at

Thanks. I'll study this code.

Roland Schweitzer wrote:

Hi Folks,

I recently encountered netCDF data sources with the following conventions attributes:

:Conventions = "GDT 1.3" ;
Conventions "IRIDL"

which I would like to read using nj22.

I set about looking for information about the known conventions in nj22 and how to create and register conventions parsing classes. I can see a bunch of classes in ucar.nc2.dataset.conv in the source code, but I don't find these classes documented in the on-line javadoc ( I haven't read source files in detail, but it appears that GDT is not supported though it's registered on the Unidata page. Perhaps the CF1 class could be extended to read GDT.

So my questions and requests are these:

Please put ucar.nc2.dataset.conv in the on-line javadoc (or point me to it if I missed it).

Has anybody messed GDT using nj22?  If so, can you please share?

Is IRIDL going to be a recognized convention by future versions of nj?

Assuming I can create the javadoc myself (or it appears on-line), and that I have read user's documentation Word file, where else should I go seeking information about conventions and nj22?