Hi John,
Thanks for this reply. Currently in ncWMS I check to see if the dataset
has ScaleMissingDeferred set with:
NetcdfDataset nc;
Boolean scaleMissingDeferred =
nc.getEnhanceMode().contains(Enhance.ScaleMissingDefer);
Then I have data-reading logic like this:
VariableDS var = grid.getVariable();
Array arr = ... // get a subset
float val = arr.getFloat(index.set(0, i - imin));
if (scaleMissingDeferred) {
val = (float) var.convertScaleOffsetMissing(val);
}
Will this generally be OK, do you think? (Maybe Ethan has an opinion
here since he's been delving into ncWMS and knows nj4 and THREDDS better
than me.)
> BTW, when you say "GridDatatype is being passed to me by "foreign"
code",
> how foreign is it? Can we assume they have set the
> NetcdfDataset.setDefaultEnhanceMode() ?
Well, the first goal was to get this code working with GridDatatypes
that are created by TDS. However, I'd like to think that the code might
be useful in other situations too. Are there any pre-conditions I
should know about?
Cheers, Jon
-----Original Message-----
From: John Caron [mailto:caron@xxxxxxxxxxxxxxxx]
Sent: 23 March 2010 00:43
To: Ethan Davis
Cc: Jonathan Blower; netcdf-java@xxxxxxxxxxxxxxxx
Subject: Re: [netcdf-java] Can I tell if a GridDatatype is "enhanced"?
Sorry for the late response.
A. yes, you do have to be careful about applying scale/offset twice.
1) as ethan said:
VariableDS vds = GridDatatype.getVariable();
this is not guaranteed to exist, so you need to check for null.
2) now we have to start breaking a lot of encapsolation, so probably we
should make public
EnumSet<NetcdfDataset.Enhance> enhanceMode =
VariableDS.getEnhanceMode();
and you want to see if
enhanceMode.contains( Enhance.ScaleMissingDefer);
if so, then scale/offset was not applied, but the scale/offset has been
calculated, so that
convertScaleOffsetMissing( Array)
should do the right thing.
Unfortunately, theres a few edge cases that might screw things up, as
you can see by looking at
VariableDS._read()
and
VariableDS.enhance().
Also, there are some changes coming in 4.2 that will handle things a bit
different.
But I think for your purposes, the above will work 99.44% of the time,
so I can make VariableDS.getEnhanceMode() public in next 4.1 release.
BTW, when you say "GridDatatype is being passed to me by "foreign"
code", how foreign is it? Can we assume they have set the
NetcdfDataset.setDefaultEnhanceMode() ?
B. GridDataset has:
/** @return underlying NetcdfFile, or null if none. */
public ucar.nc2.NetcdfFile getNetcdfFile();
which will give you the underlying NetcdfDataset, from which you can
call getFileType(). Its not guarenteed to be non-null, but so far all
implementations will have one.
Ethan Davis wrote:
> Hi Jon,
>
> Whoops! Sorry about that.
>
> A few possibilities:
>
> 1) It would be pretty easy to change the TDS to give you back a
> NetcdfDataset instead of a GridDataset. You could use that for the
> enhancement information and convert it to a GridDataset to get all the
> grid goodness out of it.
>
> 2) We could (probably should) expose the enhancement information in
> GridDataset.
>
> 3) Yuck! Dipping into implementation details. -- The only
implementation
> of ucar.nc2.dt.GridDataset is by ucar.nc2.dt.grid.GridDataset which
has
> a public getNetcdfDataset() method.
>
> Number 2 would be my first choice. Though I'm trying to think of any
> other reasons it would be useful for the WMS code to have a
> NetcdfDataset instead of GridDataset.
>
> Ethan
>
> On 2/11/2010 1:46 PM, Jonathan Blower wrote:
>> Hi Ethan,
>>
>> Thanks but there's no public getEnhanceMode() method on VariableDS
>>
(http://www.unidata.ucar.edu/software/netcdf-java/v4.1/javadoc/ucar/nc2/
>> dataset/VariableDS.html). There's a similar method on NetcdfDataset
>> though:
>>
http://www.unidata.ucar.edu/software/netcdf-java/v4.1/javadoc/ucar/nc2/d
>> ataset/class-use/NetcdfDataset.Enhance.html.
>>
>> But I can't get a handle to the NetcdfDataset if all I have is a
>> GridDataset (I think).
>>
>> Along similar lines, is there a way to get the underlying file type
of a
>> GridDataset/GridDatatype (so I can optimise data-reading algorithms
>> based on what sort of file I'm reading from)?
>> NetcdfDataset.getFileTypeId() works for a NetcdfDataset, but again I
>> can't get this if all I have is a GridDataset.
>>
>> In case you hadn't guessed, this is all part of my effort to refactor
>> ncWMS to be more integrable with TDS. From our previous
conversations I
>> believe that TDS will give me GridDataset objects, so that's why I'm
>> trying to work at this level, but if there's a way that the TDS
config
>> system could pass me higher-level objects that could work too. Any
>> thoughts?
>>
>> Cheers, Jon
>>
>>
>>
>> -----Original Message-----
>> From: Ethan Davis [mailto:edavis@xxxxxxxxxxxxxxxx]
>> Sent: 11 February 2010 18:34
>> To: Jonathan Blower
>> Cc: netcdf-java@xxxxxxxxxxxxxxxx
>> Subject: Re: Can I tell if a GridDatatype is "enhanced"?
>>
>> Hi Jon,
>>
>> Not very straight forward and there may be some loopholes I'm missing
>> that John can comment on when he returns.
>>
>> If you call getVariable() on your GridDataype, you will get the
backing
>> VariableDS. Call getEnhanceMode() on that and you'll get an
>> EnumSet<NetcdfDataset.Enhance>. If that Set contains
>> Enhance.ScaleMissing, the data you get back will have already had the
>> scale/offset/missing applied. If that set contains
>> Enhance.ScaleMissingDefer, you can directly use
>> VariableDS.convertScaleOffsetMissing() to convert the data.
>>
>> VariableDs also implements VariableEnhance.enhance(Set<Enhance>) but
the
>> comments say "DO NOT USE DIRECTLY. public by accident". I'll leave it
to
>> John to explain that and comment on the possibility of exposing any
of
>> this directly through the GridDatatype.
>>
>> Ethan
>>
>> On 2/8/2010 8:05 AM, Jonathan Blower wrote:
>>> Hi,
>>>
>>> I'm writing a method along the lines of:
>>>
>>> double[] extractDataFrom(GridDatatype grid);
>>>
>>> where the GridDatatype is being passed to me by "foreign" code out
of
>> my
>>> control. Do I have any way of telling whether the GridDatatype is
>>> "enhanced", i.e. whether scale/offset/missing attributes are
>>> automatically applied when I extract data?
>>>
>>> And does it matter? Can I program defensively to take care of both
>>> situations (enhanced and not enhanced)? I'm worried that simply
>>> applying convertScaleOffsetMissing() to a value from an enhanced
>>> GridDatatype might actually apply the enhancements twice and hence
>> give
>>> me the wrong answer.
>>>
>>> Thanks,
>>> Jon
>
> _______________________________________________
> netcdf-java mailing list
> netcdf-java@xxxxxxxxxxxxxxxx
> For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
