First version submitted by Kristian Nilssen, 2007
 3/2007 1st release version, changes by Wesley Ebisuzaki

 List of modifications by Sergey Varlamov, inverse time order
 
 See help file for "how to use" instructions and examples.
 
 Version 20.02.2008 (Sergey Varlamov):

  -corrected code bug in the grid_template processing, need to update
   from Version 31.01.2008, else all grids are treated as CF-1.0
   with zero x,y dimention variables added

  -code divided in 3 files:
     wgrib2nc.h - constant and the netcdf library-independant definitions,
     Netcdf_sup.c - netcdf library-independant functions,
     Netcdf.c - netcdf library-dependant, static-only functions,
                except f_netcdf() function (import it only).
                
 Version 31.01.2008 (Sergey Varlamov):

  -changed the way new netcdf file is created: it is created only if
   no errors and arrived data met all required criteria. It help to avoid
   generation of empty non-valid netcdf file that could cause
   an error if the 'next run' will go with an -append option

  -time info in netcdf is saved as double (NC_DOUBLE,
   was integer NC_INTEGER but no 64b integer type in netcdf 3.6.2).
   For int(4) type max valid range of dates is 1902-2037 or 1970-2037,
   depending on C library implementation. Replace mktime by other function?

  -changed here and in all wgrib2 subroutines presentation of
   latitude and longitude from float to double to preserve accuracy for
   high-resolution grids (especially ocean data)

  -added nc_time option [[+|-]{yyyymmddhhnnss}][:{dt}{mn|hr|dy}]
   where date could be used for alignment only (if negative),
   or it is assumed absolute data input start date.
   'dt' is time interval for data saved in the netcdf file.
   The 'normal' calender is assumed.
   Example:
     -nc_time -200001010300:6hr
   would be for data four times a day at 3,9,15 and 21 UTC.
   If data do not align to initial and time step - these data are
   silently ignored. Usefull if there is possible some
   irregularity in grib2 data timing.

  -fatal_error if the level type is unsupported by wgrib2
 
 Version 09.01.2008 (Sergey Varlamov):

  -with the help of Rich Signell added support for export
   of Polar (grid_template=20) and Lambert (grid_template=30)
   projected data under the CF-1.0 convention.
   Mappind is done using "coordinates" attribute and directly
   defining in the netcdf file the lat-lon values for all points.
   Other possibility would be to specify in the netcdf file the "grid_mapping"
   attributes (see CF convention) when mapping to lat-lon will be done
   by user processing software. It would extend the range of eligible
   for conversion grid templates in future.
   CF convention NetCDF files are not directly compatible with GrADS v1.9b4 but
   potentially could be used with user-provided "GrADS data description file (catalog)".
   Could be used with the IDL and MATLAB data analysis tools.
  
  -changed behavior for -nc_pack option: with 'float' packing option
   it does the check of data to be in valid data range, with the 'short' or 'byte'
   option the simple packing of data is also going. Data outside of valid range
   are writted as _FillValue values (missing). Originally it was an error
   if non-missing data were found out of 'shor' or 'byte' packing range and
   'float' option was ignored.

  -updated help file.

  -changed the _FillValue for the float non-packed data to the wgrib2 UNDEFINED,
   was slightly different. [Do not 'append' data to necdf created with the earlier
   version of wgrib2 if there could be undefined values - this attribute is not checked
   or changed after the file first was created (variable added)
   and new missing values could be treated as valid data].
  
 Version 13.07.2007 (Sergey Varlamov):

  -help file was updated; removed dublicating instructions from this code

  -corrected warnings with some compilers (like ecc) when printing size_t X values
   usind the %d format: now apply %lu format and (unsigned long)X conversion
   in print operations.

  -added some attributes to the generated netcdf file that could help to make it more
   "self-describing" as it is the strong feature of netcdf. These include time attributes
   like reference_time/date if it is possible to fix, time_step etc.
   The global attributes like center and sub-center information etc could be added.
   Does somebody need it?

 Version 18.06.2007 (Sergey Varlamov):

  -in the user defined table of GRIB2 to NETCDF conversion parameters
   it was added support for next optional directives, mainly for
   the 'advanced' users, see help:
    $nlev 5               Equivalent of '-nc_nlev'; has precedence over the command line option
    $levs 1000 500.0 ...  Vertical level values exported to the netcdf file, in netcdf units
    $lev_type 100:pressure:pressure level:mb:0.01
                          User-defined type of vertical level for 4D data.
    $grads 1              Equivalent of  '-nc_grads' if 1 or '-no_nc_grads' if 0; take precedence

  -the total number of fields successfully added or updated in the netcdf file
   is counted. If the user defined g2nc_table is used for data export to netcdf
   and no field were added/updated - warning message is issued at the cleanup stage
   as it could be caused by user error in filling the table. Error example:
   specify $levs in [mb], but left scaling 1,0. As pressure levels in grib file
   are default in [Pa] - no one level will fit to user table in [mb] and data
   will be skipped from output to the netcdf. Hope that warning could help in such cases.

  -corrected some found bugs. These include memory deallocation for the
   such shared objects as user-defined table for GRIB2 to NETCDF convertion
   and some others that could impact updating of existing partially undefined
   netcdf file (undefined vertical levels). Modified some Netcdf.c
   program variable names to be more self-explanatory,
   like g2nc_smlt -> g2nc_4Dlt etc.

 Version 13.06.2007 - changes by Sergey Varlamov:

   -in the user defined table of GRIB2 to NETCDF conversion parameters added
    keyword 'ignore'. New records format could be:

    wgib2_name:wgrib2_level|*:nc_name|ignore[:ignore|no|float|short|byte[:min:max]]

    Min and max values are significant for the short or byte packing only.
    Both could be omitted; it means that automatic scaling for the short
    or byte packing will be estimated from the first entered wgib2_name data
    at the wgrib2_level or at first level in case of '*' as level value.

    If the keyword 'ignore' is found as a netcdf variable name or
    as a packing type value, the corresponding data are ingnored
    and do not written to the netcdf file.
    'Ignore' keyword is recommended if the data from the same
    grib2 file are exported in number of output files (netcdf or other)
    by the same wgrib2 process when the same decoded data could be passed
    to output in other file of any supported type.
    Doing export to single netcdf file using 'ignore' keyword is not recommended
    as corresponding data are first decoded and after that skipped from writing
    to the netcdf file.

   -compare existing variables definition in open netcdf file
    with new data attributes before updating or adding data.

 Version 11.05.2007 (Sergey Varlamov):

   -added posibility to export data to netcdf as 4D data. Now supported are
    4D data, 3D data, or mixed 3D and 4D data in one netcdf file.
    4D data are defined in {TIME,LEV,LAT,LON} space.
    Grib2 types of vertical levels eligible for export to the netcdf as 4D data
    are included in nc_4Dlt (this file) and now include types
    20,100,104,105,107, and 160 of GRIB2 code.
    First found eligible type is used, error is generetad if other
    eligible type data are met in the input stream.
    To activate 4D data output plese use '-nc_nlev' option followed by
    integer 'max_number_of_vertical_levels'.
    The 'max_number_of_vertical_levels' defines vertical dimension of
    4D data exported to the netcdf file and do not apply to the data
    at the 'non-eligible' vertical levels (like mean sea level etc)
    or at levels included in to the user-defined
    GRID2 to NETCDF conversion table explicitly (see description below).
    Such 'invariant' level data are treated as 3D data
    defined in {TIME,LAT,LON} space with possible level information
    included into the variable names as were coded by Wesley Ebisuzaki.
    If 'max_number_of_vertical_levels' is zero - all coming data are threated as 3D,
    with level information included in variable names.
    When existing netcdf file is updated in the '-append' mode (see description below)
    the value of 'max_number_of_vertical_levels' must not exceed initial value provided
    when file was first created (defined).
    First creating netcdf file, vertical level values are not fixed (are undefined)
    and these are defined one-by-one when data at new level are added
    to the netcdf file, up to the 'max_number_of_vertical_levels'.
    It is user responsibility to define monotonically changing sequence
    of vertical levels, as it is required by the COARDS convention.
    Error is generated if new vertical level is specified in non-regular
    order.
    Example1: sequence of data at: 1000 mb, 850 mb, 950 mb - generate error for 950 mb data
    Example2: sequence of data at: 1000 mb, 950 mb, 850 mb, 1000 mb, 700 mb...
    is appropriate as 1000 mb level was already defined and data would be placed
    correctly in to the netcdf file.

   -it is possible to pack data in netcdf using '-nc_pack' option or user defined table
    of GRIB2 to NETCDF conversion parameters (see below).
    With '-nc_pack' option folowed by X=min:max:byte|short] all NEW input variables
    would be packed in short or byte with corresponding offset and scale fitting
    given range in specified data type with possible loss of precision.
    packed=(unpacked-offset)/scale; default no packing    
    NEW means that if some variable was already defined in the netcdf file
    and now is appended to it (in -append mode, see below), the initially defined
    packing parameters are used.
    Both zero min and max values activate 'auto' packing when scale and offset are defined
    from first entered field. When packing, fitting is checked, error if data
    do not fit packing limits.
    Packing in 'byte' is not recognized by GrADS v1.9b4.

   -added possibility to provide user defined table of GRIB2 to NETCDF
    conversion parameters.  It would be read from the user file with name provided
    following the '-nc_table' option.

   -option '-nc_grads' is introduced. GrADS (version 1.9b4) do not support
    non-constant data time stepping and silently generate wrong time stamps for such
    netcdf or opendap data sets when creating nice graphics. With this option an error
    is raised if time step is changing when writing data to the netcdf file. It is possible
    for the sequence of forecasts as example. Packing to byte also is not supported
    by GrADS v1.9b4.

   -added support for the '-append' option; it works fine with netcdf files
    created by wgrib2 as it uses dimension, variable and attribute names
    from the netcdf file generated by this utility.
    The existing netcdf file could be expanded in time as well as new variables
    at invariant vertical levels or in the range of fixed
    vertical levels could be added. For the netcdf (as 'direct access' data set)
    '-append' option really is also 'overwrite if same' if new data come with the
    variable name, level and verftime being same as already recorded in the file.
    As for all other output types - do not forget to delete invalid netcdf file
    if it was left from some errorneous runs. Else you could get unpredictable
    result!

   -changed float "_fillValue" attribute name to data type "_FillValue";
    No problems with GrADS to recognize this missing value,
    although not all is clear for me as in COARDS:
    "In cases where the data variable is packed via the scale_value attribute
     this implies that the missing_value flag is likewise packed. The same holds
     for the _FillValue attribute."...

   -check error code on verftime as in other Wesley Ebisuzaki files
   -many other small changes...