Enable seperate ice_history.nml & cice_in.nml settings

[anton-seaice]

To enable simpler changes to the icefields namelist in cice, this change adds support for an ice_history.nml file which is spliced into the normal cice runtime settings (cice_in.nml) by payu before the model is run.

This follow the discussion in https://forum.access-hive.org.au/t/access-esm1-5-release-work-plan/1721/24

The contents of ice_history.nml can be anything which also would fit in cice_in.nml , but the intention is to move &icefields_nml to the ice_history file

Do we need to update docs ? Or tests ? There doesn't appear to be any of either at this point

[pep8speaks]

Hello @anton-seaice! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

• In the file payu/models/cice5.py:

Line 56:8: E114 indentation is not a multiple of four (comment)

• In the file test/models/test_cice.py:

Line 92:28: E201 whitespace after '{'
Line 92:41: E202 whitespace before '}'

Comment last updated at 2024-08-15 04:15:42 UTC

[anton-seaice]

@aidanheerdegen @blimlim - This is ready for review.

To test, make a ice/ice_history.nml file and put some valid namelist values in it (move something from ice/cice_in.nml) and the payu run.

[coveralls]

coverage: 54.145% (+1.9%) from 52.281%
when pulling 64e356c on ACCESS-NRI:esm1.5-patch-cice-history-nml
into 83001a2 on payu-org:master.

[aidanheerdegen]

Do we need to update docs ? Or tests ? There doesn't appear to be any of either at this point

That would be great.

We have an issue about the lack of driver docs

#470

And there isn't an example to leverage off at this point I'm afraid.

But just adding some tests for this functionality in a driver test suite would be fantastic. There are some examples you could base that on here

https://github.com/payu-org/payu/blob/master/test/models/

[anton-seaice]

The expected behaviour for this change is :

• ice_history.nml overrides and adds to any values also in cice_in.nml
• there is no change if ice_history.nml doesn't exist
• ice_history.nml is not moved to output* folders, as its content is in cice_in.nml

[anton-seaice]

But just adding some tests for this functionality in a driver test suite would be fantastic. There are some examples you could base that on here

I think that would require splitting the new lines in a function ? e.g. patch_nml(self)

It feels a bit like overkill but maybe its sensible ?

[blimlim]

Awesome, really glad that this is working!

Changes look good to me. I've just added a comment for an optional change, but am not sure that it's necessarily better.

Running on gadi with the history parameters separated into a separate namelist worked for me.

[aidanheerdegen]

I have an opinion or two ...

[anton-seaice]

I added a test, and moved the 'dump_last' to the cice5 driver. I didn't add ice_history.nml to optional_config_files because I think its potentially confusing (cice_in.nml is the source of truth).

[aidanheerdegen]

LGTM