On 7/15/2011 2:09 PM, Russ Rew wrote:
If "this suggestion" means extending missing_value to permit multiple
values, then the only change it makes is changing a one-to-one mapping
into a one-to-many mapping. Since my example of breaking ncdump/ncgen
backward compatibility doesn't apply to the "missing_value" attribute,
that may be harmless, unless there is other existing software that
depends on "missing_value" having a single value. So I withdraw my
confused objection to using "missing_value".
i think that missing_value attribute has always been a vector, the CDM
treats it that way:
from
http://www.unidata.ucar.edu/software/netcdf/docs/netcdf/Attribute-Conventions.html:
"The missing_value attribute can be a scalar or vector containing values
indicating missing data".
CF Appendix A says:
"A value used to represent missing or undefined data (deprecated by the
NUG)"
Is there some other place that says it cannot be a vector?
(BTW, we should remove "(deprecated by the NUG)", apparently missed that
in ticket 58. i will submit defect ticket)
