Instantaneous CAM-SIMA history infrastructure

[peverwhee]

Overview of Changes

• Python:
  • cime_config/hist_config.py: New namelist parsing for history configurations
  • src/data/write_hist_file.py: Generates new physics history file from registry (bld/atm/obj/history/physics_history.F90)
• Fortran:
  • src/history/cam_history.F90: contains driving functions that are called by cam_comp.F90 (history_init_files, history_write_files, history_wrap_up), as well as history_out_field and history_add_field
  • history_init_files called at cam_init time
  • history_write_files and history_wrap_up called during cam_timestep_final
  • src/history/cam_hist_file.F90: contains namelist reading and history file handling (set-up, defining, writing, closing files)
  • src/control/cam_control_mod.F90: broken up to avoid circular dependency
  • src/utils/cam_grid_support.F90: modified to allow 1d output and to handle split history files
  • src/utils/cam_filenames.F90: renamed and moved from src/control/filenames.F90
• Misc:
  • src/data/registry.xml: updated to include XML tag with diagnostic name and usual flag
• Uses the modular history developed by @gold2718
  • Exists here right now: https://github.com/peverwhee/history_output
  • I propose we move it to an official repo when all the different accumulate types are implemented/flushed out (but I'm open to other suggestions/requests)

Documentation on design and configuration: Here

Testing

Fortran unit testing not yet implemented (see issue https://github.com/ESCOMP/CAM-SIMA/issues/273), but the following python unit tests have been added (in $SIMA/test/unit):

• test_hist_config.py: tests cime_config/hist_config.py, which is what parses the namelist and writes history NL groups to atm_in
• test_write_hist_file.py: tests data/write_hist_file.py, which generates history_add_field and history_out_field calls for all registry variables with a \<diagnostic> XML tag.

[cacraigucar]

Does it make sense to provide a template file for a physics parameterization's diagnostic package (both .F90 and .meta)? I realized that I'm probably going to copy the kessler_diagnostics example and add my own History_add_field and history_out_field. It would be nice to have a template to start the process from. Thoughts?

[peverwhee]

@cacraigucar I'm happy to make a template if you think that would be helpful! Would it be better or worse to just have the .F90 template and then use the metadata generator for the metadata once the fortran is populated?

[cacraigucar]

@cacraigucar I'm happy to make a template if you think that would be helpful! Would it be better or worse to just have the .F90 template and then use the metadata generator for the metadata once the fortran is populated?

Probably just having a template for the .F90 file and having the metadata generator run after the Fortran file is complete makes the most sense (and least amount of work for a user)

[cacraigucar]

@cacraigucar I'm happy to make a template if you think that would be helpful! Would it be better or worse to just have the .F90 template and then use the metadata generator for the metadata once the fortran is populated?

Probably just having a template for the .F90 file and having the metadata generator run after the Fortran file is complete makes the most sense (and least amount of work for a user)

I realized that one of the "pains" of filling in the metadata file is finding the correct standard name. Since this will already be in the add_field call, perhaps we could have a custom metadata generator for diagnostic files/metadata? This request, if folks think it makes sense, would be a new issue/PR.

[peverwhee]

@cacraigucar added the template here - https://github.com/peverwhee/atmospheric_physics/blob/diagnostics/cam_diagnostics/scheme_diagnostics_template.F90

[peverwhee]

@cacraigucar I'm happy to make a template if you think that would be helpful! Would it be better or worse to just have the .F90 template and then use the metadata generator for the metadata once the fortran is populated?

Probably just having a template for the .F90 file and having the metadata generator run after the Fortran file is complete makes the most sense (and least amount of work for a user)

I realized that one of the "pains" of filling in the metadata file is finding the correct standard name. Since this will already be in the add_field call, perhaps we could have a custom metadata generator for diagnostic files/metadata? This request, if folks think it makes sense, would be a new issue/PR.

Hmm. I do see that this is duplicated effort (putting the standard name in the add field call and also in the metadata), but I'm leaning towards it being overkill to make a new generator to avoid that step

[cacraigucar]

@cacraigucar added the template here - https://github.com/peverwhee/atmospheric_physics/blob/diagnostics/cam_diagnostics/scheme_diagnostics_template.F90

Perfect - I am using it now!