Allow unset default values for namelist variables

[jimmielin]

What is the feature/what would you like to discuss?

Currently there's incomplete infrastructure for namelist variables to have an "unset" default, i.e., if the namelist definition file (scheme_namelist.xml) does not provide a <values> option.

Already available

create_readnl_files.py generates namelist read code and correctly sets the following when namelist entries are missing:

• int: unset_int
• logical: .false.
• real: -HUGE(1.0_kind) or -HUGE(1.0)
• character: unset_str
  so the generated readnl code is complete.

What needs to be changed

@nusbaume wrote:

Is there a reason we aren't wanting a values XML tag at all as opposed to just providing a blank value entry and then just modifying the CAM-SIMA code to properly handle the "missing" value? So basically something like this:

<values>
  <value></value>  <!--default "missing" value -->
</values>

It seems like this will trigger errors in paramgen.py in CIME and atm_in_paramgen.py in SIMA, so some work is needed for this to work as well.

Alternatively,

1. CIME's entry_id_pg.xsd needs to be changed to allow <values> to be optional, by changing <xs:element ref="values"/> to <xs:element ref="values" minOccurs='0'/>
2. atm_in_paramgen.py should be changed to handle <values> being not present. Right now it just reads

val = self._data[nml_group][var]["values"]

could be changed to handle missing <value> through something like

if "values" in self._data[nml_group][var]:
  #Extract variable value(s):
  val = self._data[nml_group][var]["values"] #.strip()

Is there anyone in particular you want to be part of this conversation?

@peverwhee @nusbaume

Will this change (regression test) answers?

No

Will you be implementing this enhancement yourself?

Any Software Engineer can do this