Custom Query (964 matches)

The NetCDF library version 4.1.1 or higher can now process OPeNDAP files when DAP (Data Access Protocol) support is activated. This is only possible for reading input NetCDF files. Writing to OPeNDAP NetCDF file is not possible.

Additional libraries associated with cURL are required during the linking process. This is set with the calling to the curl-config --libs script in any of the make configuration files:

 ifdef USE_DAP
             LIBS += $(shell curl-config --libs)
 endif

Here, the USE_DAP environmental variable can be activated in the makefile or build.sh/build.bash scripts. A call to this script will usually return:

% curl-config --libs
-L/usr/kerberos/lib64 -lcurl -ldl -lgssapi_krb5 -lkrb5 -lk5crypto -lcom_err -lidn -lssl -lcrypto -lz

You need to install the ​cURL libraries for your system.

A new function, find_file, was added to inp_par.F to check if an input file exists. Special steps are taken for OPeNDAP files which have a URL in the name, for example:

http://tashtego.marine.rutgers.edu:8080/thredds/dodsC/met/ncep-nam-3hour/swrad_nam_3hourly_MAB_and_GoM.nc

All the file name variables are now declared with 256 characters. Previously, all were 80 characters long. For example, now we have:

        character (len=256), dimension(Ngrids) :: GRDname

For reason that will become obvious in the future, the cpp option NETCDF4 was renamed to HD5. This option can be activated to create NetCDF-4/HDF5 format files. That is, an HDF5 format file.

Version 4.1.1 of the NetCDF library also offer support for third party parallel-netcdf library (also known as pnetcdf) developed by the Argonne National Lab and Northwestern University. I started to looking into that but the documentation about how to compile this is very limited. We are having problems compiling with PGI and gfortran. The ifort sort of compiles but I have not been able to run satisfatoryly a simple parallel I/O example. Stay tunned about this development...

In ​src:ticket:418, I allowed for the file name to a maximum of 256 characters long. This requires a change in inp_par.F, function decode_line, to declare several strings as 256 characters long. I knew that I needed to make this change but I forgot.

This also affects the include file for biology (ecosim_inp.h, fennel_inp.h, nemuro_inp.h, npzd_Franks_inp.h, npzd_Powell_inp.h, npzd_iron_inp.h) and sediment (sediment_inp.h) models. ROMS developers of new biological models need to take notice of this change. For example, you need to have in any routine included in inp_par.F:

      character (len=40) :: KeyWord
      character (len=256) :: line
      character (len=256), dimension(100) :: Cval

One of these days, I will include all these routines in inp_par.F inside a module. It is in my things-to-do list.

I am aware of many users being burn out because of the previous 80 character limit on file names. I think that 256 (a nice power of 2) is a decent length for any file, including URL from OPeNDAP.

It turns out that the changes introduced in ​src:ticket:416 were wrong. The previous adjoint-based codes for NPZD_IRON option were correct. This only affects files tl_npzd_iron.h, rp_npzd_iron.h, and ad_npzd_iron.h. Many thanks to Andy Moore for checking this correction.