[netCDF #QYC-628947]: NF90_GET_VAR not working with user defined type variable

Hello,

I will try to answer as best I can; I am not a fortran programmer, however, but 
hopefully that won't matter.

1.  The input for NF90_GET_VAR is in fact the file ID number -or- the group ID 
number (when dealing with netCDF 4 files).
2. I'm glad you were able to get the information about compound files working.  
 If this were C code, you would use nc_get_var() to read the values of the 
compound variable type into an array of structs, which you would then inspect 
one at a time.  You could also use nc_get_var1() to read in a single element 
from the compound type data, but you would still have to inspect the components 
of the data type individually.  nc_get_var(), which is invoked by NF90_get_var, 
returns the entire compound variable in a struct, and does not provide direct 
access to the individual components.

I'm sorry I can't provide any information in a Fortran context; hopefully this 
information is useful.

-Ward

> 
> 
> Dear Ward
> 
> Your recommendation sounds good. However, the user
> guide is not clear about what is really required BEFORE using the
> function NF90_GET_VAR.
> 
> I can read and display every piece of
> information on my test nc file, incluing the cases when variables are
> compound user defined types with several fields, each field with a
> different type, inside a group, which may exist inside other groups. I
> am attaching a text file showing that I can extract information about
> every single variable of any type and dimension, on the root group (what
> I name level 1 group on my attached output file) or on other groups of
> level 2 or deeper.
> 
> The user guide does not explain HOW the information
> of compound user type, namely the information on the respective fields,
> should be made available to the funtion NF90_GET_VAR.
> 
> Indeed, the user
> guide says that the INPUT to that function is
> 
> (1) the file ID number
> (I suspect it should be the groupd ID number, not the file ID number...)
> 
> 
> (2) the variable ID number, as provided by a previous call to the
> function NF90_INQ_VARID, which requires as input the file ID (I suspect
> it should also be the group ID...) and the variable name
> 
> Thus I am
> somehow lost, because I do not know HOW the function NF90_GET_VAR has
> information about the names, types, sizes of the fields to be read.
> 
> It
> requires the variable ID only.
> 
> Shoud I work with the function
> NF90_GET_VAR for every single field each time, NOT to the whole compound
> type variable? If so, then the variable ID should be the field ID...
> 
> I
> would appreciate the kindness of your attention and help, particularly
> for clarification of the previous item,
> 
> Kind regards
> 
> Arlindo Arriaga
> 
> 
> PS: the variable I want to read has has the name "sst" and it is
> composed by two fields, each field is a real with 4 bytes.
> 
> Such a
> structure is allocated with the dimension 4320. I represent that
> variable in F90 as follows:
> 
> TYPE binDataType
> real(kind=4) :: sum
> 
> real(kind=4) :: sum_squared
> end type binDataType
> TYPE(binDataType),
> allocatable, dimension(:) :: sst_data
> 
> 
> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> 
> On
> Thu, 14 Apr 2016 10:20:38 -0600, Unidata netCDF Support wrote:
> 
> >
> Hello,
> >
> > In order to read data stored in a compound type, you must
> use `NF90_INQ_COMPOUND` to learn about the compound data structure. The
> user guide for netCDF-Fortran NF90 may be found here:
> >
> > *
> http://www.unidata.ucar.edu/software/netcdf/documentation/historic/netcdf-f90/Compound-Types.html#Compound-Types
> [1]I hope this helps,
> >
> > -Ward
> >
> >> Dear Lady or Sir I would
> appreciate if you could tell me how to read data from nc files with
> netCDF-4 format using F90. The Fortran library function NF90_GET_VAR
> works well with integer or real arrays, but it can not compile if I want
> to read a compound user defined type variable (as for example, sst on a
> MODIS nc file with netcdf-4 format). Please let me know what may your
> services suggest as an alternative to that F90 library function ? I
> thank you for the kindness of your attention. Kind regards Arlindo
> Arriaga, PhD Meteorology UW-Madison 1991
> >
> > Ticket Details
> >
> ===================
> > Ticket ID: QYC-628947
> > Department: Support
> netCDF
> > Priority: Normal
> > Status: Closed
> 
> 
> 
> Links:
> ------
> [1]
> http://www.unidata.ucar.edu/software/netcdf/documentation/historic/netcdf-f90/Compound-Types.html#Compound-Types
> 
> 


Ticket Details
===================
Ticket ID: QYC-628947
Department: Support netCDF
Priority: Normal
Status: Closed