[CF-metadata] variable missing "coordinates" attribute

John Caron caron at unidata.ucar.edu
Wed Jan 26 15:12:13 MST 2005


Richard P Signell wrote:

>John,
>
>I happen to know this user ;-) and yes, he was just sloppy on this one.  It
>might be he didn't notice that "topo" was lacking the coordinates attribute
>because he was using IDV.   IDV is smart enough that if it finds the
>"coordinates" attribute specified for one variable, it uses those for other
>variables that have the same dimensions (in this case "j" and "i") yet
>don't have the "coordinates" attribute specified.   So it sort of cleans up
>after sloppy people.
>  
>
Hi Rch, sorry to blow your cover ;^)

The IDV is using the netcdf-java library to figure out these coordinate 
systems. Version 2.1 uses a "greedy" algorithm that assigns all possible 
coordinate axes to a variable, then tries to seperate them into coherent 
systems if needed.

In netcdf-java 2.2 im trying to do things right whenever possible, and 
CF is supposed to be our exemplary way of doing things, so i was trying 
to interpret strictly to CF, and thus removed the above algorithm that 
was responsible for catching the slop.

anyway, ill probably add back in something to deal with the case when 
the "coordinates" attribute is missing, but when its present, i will 
take it as deliberate on the part of the file writer.




More information about the CF-metadata mailing list