On Dec 14, 2011, at 4:26 PM, Steve Hankin wrote: > Hi James, > > A CF perspective: Per CF the data types must match: > http://cf-pcmdi.llnl.gov/documents/cf-conventions/1.6/cf-conventions.html#missing-data > This is true even if the variable is "compressed" using add_offset and > scale_factor. > > Individual applications may be less fussy than the CF spec dictates, but in > general unmatched datatypes should be assumed to produce unpredictable > behavior in applications. Ncgen would be a particularly significant > application, but even ncgen would have to alter the _FillValue datatype to > ensure it matches the variable's datatype, in order to generate the file. Thanks for providing a definitive answer, or at least as much as one can. I figured this would be the case. James > > - Steve > > ============================= > > On 12/14/2011 1:24 PM, Gallagher James wrote: >> >> Does the type of the _FillValue attribute have to match the element type of >> the array to which it is bound? >> >> For example, suppose there's an array of Bytes, can the _fillValue be a >> Short? >> >> I'm asking because a question about DAP2 and promotion of DAP's (unsigned) >> Byte to (signed) Int16 has popped up again. >> >> Thanks, >> James >> -- >> James Gallagher >> jgallagher at opendap.org >> 406.723.8663 >> >> >> >> _______________________________________________ >> netcdfgroup mailing list >> netcdfgroup@xxxxxxxxxxxxxxxx >> For list information or to unsubscribe, visit: >> http://www.unidata.ucar.edu/mailing_lists/ -- James Gallagher jgallagher at opendap.org 406.723.8663
Attachment:
signature.asc
Description: Message signed with OpenPGP using GPGMail
netcdfgroup archives: