Skip to content

CCPP Framework Meeting Minutes 2019 01 31

Jump to bottom
goldy edited this page Mar 9, 2020 · 1 revision

Discussion of new metadata standard and new functionality

  • no auto-flipping of vertical direction yet (later this year; not needed by any of the groups right now)
  • Cheryl's chemistry model still uses old metadata, new Kessler suite developed using metadata

How to most effectively convert old to new metadata and use new metadata in FV3/SCM? Or: discussion on new metadata standard and conversion/parsing tools

  • test_capgen.py is a useful tool to convert and parse and shows the basic usage
  • units: agreed to not convert dimensionless units to 1 in order to preserve the information (just don't output to CF in future)
  • missing dimension information for variables: need to allow having ":" instead of guessing it; warnings are useful and should be saved during the conversion process to inspect/fix problems afterwards
  • any new property for metadata is going to be optional (with reasonable default values)
  • proposed changes to current PR (https://github.com/NCAR/ccpp-framework/pull/172):
    • instead of keeping new metadata in place, write to separate file with ending ".md" (i.e., xxx.F90 -> xxx.md)
    • will require a converter from .md to a format that doxygen can pretty-print (on a per-argument-table level) and that can be included in the source code via doxygen include statements; example:
      • xxx.md -> xxx_arg_table_name_of_subroutine1.xml, xxx_arg_table_name_of_subroutine2.xml, ...
      • in Fortran code: !! \htmlinclude xxx_arg_table_name_of_subroutine1.xml
    • Steve will make changes to the PR that enable the first step (write converted metadata to separate file)
    • whoever has time to look into the .md to documentation conversion and inclusion in the source code should do so and update the issue related to this PR (https://github.com/NCAR/ccpp-framework/issues/171) - Grant was volunteered to do this
    • also need to manually convert GFS_typedefs.F90 and other host-model metadata for NEMSfv3gfs and SCM (Dom, Grant?)
  • any proposed changes and discussion that is relevant should be added to https://github.com/NCAR/ccpp-framework/issues/171
  • timeline: we want to have the NEMSfv3gfs metadata converted and parsed by the end of February (similar/shortly afterwards for SCM)

Other discussions

  • need a better plan to work with optional arguments (right now, pretty much static in host-model dependent prebuild config, should be suite-dependent)
  • in general, more information inside suite definition files rather than host-model dependent
  • machine.F issue (non-standard kind definitions, https://github.com/NCAR/ccpp-physics/issues/194) needs to be addressed soon

CCPP Framework wiki

Clone this wiki locally