[Kst] extragear/graphics/kst/src/datasources/netcdf

Brisset, Nicolas Nicolas.Brisset at eurocopter.com
Fri Aug 25 09:53:46 CEST 2006

I forgot to mention 2 things:
1) you can get example netCDF files to play with from http://www.unidata.ucar.edu/software/netcdf/examples/files.html (I used GOTEX.C130_N130AR.LRT.RF06.PNI.nc but others should work as well)
2) netCDF files have global attributes for files and also attributes for variables (vectors, matrices etc). It seems we currently have only file metadata and not variable metadata (even though metadata are accessed from the RMB of a vector, which is in fact a bit confusing!). Gathering metadata for all variables to put it into the global file metadata area is feasible but sounds like a bad idea (as users mostly use a subset of the complete data available in a file, it would be a waste of resources and make the relevant information difficult to find). I think we should have plans for variable metadata as well, because in many cases units (for example) are stored in that way and I believe it would be interesting to be able to access them. 

I have the feeling we should try to add a design document for metadata in devel-docs/ and discuss this a bit more (I know, that's a post-1.3.0 topic :-)), as it would add a *very* professional and useful touch to kst (who has never lost track of the context in which a plot was made, making it unusable ?).


> -----Message d'origine-----
> De : Nicolas Brisset [mailto:nicolas.brisset at eurocopter.com] 
> Envoyé : mercredi 23 août 2006 22:09
> À : kde-commits at kde.org
> Cc : kst at kde.org
> Objet : [Kst] extragear/graphics/kst/src/datasources/netcdf
> SVN commit 576326 by brisset:
> Add initial support for metadata from netCDF files from 
> global attributes. For multi-valued attributes, only the 
> first value is used (but strings are complete and certainly 
> the most interesting kind of metadata). If there are requests 
> for getting more values it should be easy to add. Also change 
> kdDebug() to kstdDebug() and comment out debugging traces. 
> Only failures still print out something. Now George can test 
> this and try to make metadata strings usable in labels :-) I 
> am committing directly as it is a very isolated change that 
> won't affect users without netCDF support, and it is easy to 
> revert in case of problems.
>  M  +18 -13    netcdf.cpp  

More information about the Kst mailing list