Generalized subroutine to read the state variables from the namelist (parse_variables) #783
Conversation
…nto a table and verify its contents; putting it in the default_model_mod
Updating the generic routine by including the type state_vars_type and storing the contents of the nml entry here Including a conditional use_clamping that tells whether or not the model's nml entry includes clamping values; executing a separate do loop over the rows if true to allow for reading in these values
its functionality. Adding checks against known values with fortran-test-anything
…riables: clamp and noclamp
…ing 2 distinct input.nml files
…del_mod This model is the perfect example of how we can have separate calls to get_state_variables for each file domain instead of having the domain included in the state_variables nml item.
|
@mjs2369 Do you have restart files so you can test the changed model_mods? |
There was a problem hiding this comment.
Hi Marlee,
Good work finding all the model_mods routines that can be switched out by this.
I've put some comments in. The thing to sort out first, is the get_state_variables routine.
- Does the overloading and having to set clamping = .true. make. sense? '
- Can you keep state_vars type private, so the model_mod does not access the internals (see the comments on add_domain)?
The 3d template model_mod needs updating to use the new routines, since this is what is used when the 'new_model.sh' script is run.
Documentation for routine also needs adding, maybe 'adding a new model' is the best place for this, I'm not sure.
I think the tests should comprehensively test the whole table read - if we are running the test, should test all the values.
Only for cam-fv and pop - do you want me to test that all models are bitwise? |
If you're changing the model_mod then yes, check that the new code does the expected. |
- Make the name of the function parse_variables - Remove interface to create two distinct functions: parse_variables and parse_variables_clamp - Remove use_clamping and max_state_vars arguments - Update the calls to the function in all edited model_mods
…riables is empty before calling add_domain (POP and CICE)
|
Marlee still testing this. |
|
Tested and ready to merge @hkershaw-brown Cam-fv: bitwise filter executions |
|
@mjs2369 can you point me to where your test restart files are on Derecho, thanks! |
| MAX_STATE_VARIABLE_FIELDS_CLAMP | ||
|
|
||
| type :: state_var_type | ||
| integer :: nvars |
There was a problem hiding this comment.
note for chat today
nvars = -1
@hkershaw-brown /glade/work/masmith/test_cases/ |
"big" for models currently in the repo
Default value for nvars = -1 (rather than unset) in that state_var_type
There was a problem hiding this comment.
Great work on this Marlee,
I added the error check for nvars vs. MAX_STATE_VARIABLES in parse_variables, and parse_variables_clamp, if you can double check those that would be great.
Feel free to release when you're ready!
Cheers,
Helen
Edit: parse_variables -> parse_variables_clamp
hkershaw-brown
added
release!
mjs2369
changed the title
Description:
There are many variations in the model_mods of the subroutine to take the table of state variables read in from the namelist and put the info into various data structures, which can then be passed into add domain. A new variation of this subroutine would need to be written every time a new model is added to DART.
This PR creates two generalized subroutines that convert the table of state variables to a state_var_type :
parse_variables reads netcdf variable name, dart qty, and update values from the nml
parse_variables_clamp reads netcdf variable name, dart qty, clamp_values, and update values from the nml
Alternate versions for this subroutine were replaced with get_state_variables in the models MOM6, wrf_hydro, aether_lat-lon, cam-fv, cam-se, POP, and cice
New dev test test_parse_variables added
Fixes issue
Fixes #448
Types of changes
Documentation changes needed?
Tests
Compiled all models with updated model_mods
Build_everything runs as expected with all compilers
Bitwise tested with CAM-FV
New dev test runs successfully and reports pass/fails correctly
Checklist for merging
Checklist for release
Testing Datasets