On 7/2/2013 5:52 PM, Alex Fraser wrote:
Hi John,
Thanks for the detailed explanation.
----- Original Message -----
From: "John Caron" <caron@xxxxxxxxxxxxxxxx>
[...]
3) a valid range is applied to the underlying data, before anything
else happens. [...]
Is the same true for _FillValue? In that case I think I can use the following
rules for unsigned variables:
- When specifying a fill value, use the same data type
as the variable and allow it to be written signed;
nc4 IOSP demands this.
- When specifying a valid range, promote the data type
(e.g. byte -> short).
For example, the following should be equivalent:
byte Band1(y, x);
Band1:_Unsigned = "true";
Band1:_FillValue = -1b; // byte
byte Band2(y, x);
Band2:_Unsigned = "true";
Band2:valid_range = 0s, 254s; // short
Is that right?
Cheers,
Alex
Hi Alex:
Yes, that should work.
We may need to describe this in the user manual, so other libraries
implementing scale/offset/missing values will do the same thing.
If anyone on these lists have further opinions, they would be welcome
before we standardize on it.
John
