generate_user_config_files is creating spurious masking entries in the config files

[mo-jareddrayton]

generate_user_config_files is creating spurious (and possibly erroneous?) masking entries.

[masking]
stream_ap5_cice-U = -1:None:None,180:None:None
stream_ap5_grid-V = -1:None:None,180:None:None
stream_ap6_cice-U = -1:None:None,180:None:None
stream_ap6_grid-V = -1:None:None,180:None:None
stream_onm_cice-U = -1:None:None,180:None:None
stream_onm_grid-V = -1:None:None,180:None:None

Looks like the problem is in the get_masking_attributes function.

CDDS/cdds/cdds/configure/user_config.py

Lines 215 to 225 in 574dd77

	maskings = OrderedDict()
	plugin = PluginStore.instance().get_plugin()
	grid_info = plugin.grid_info(model_id, GridType.OCEAN)
	masking_key_template = 'stream_{}_{}'
	masks = grid_info.masks
	for stream in streams:
	if masks:
	for grid, value in masks.items():
	key = masking_key_template.format(stream, grid)
	maskings[key] = value
	return maskings

[matthew-mizielinski]

The masking information is collected by sub-stream in the model

CDDS/cdds/cdds/common/plugins/cmip6/data/model/UKESM1-0-LL.json

Lines 211 to 234 in 574dd77

	"masked": {
	"grid-V": {
	"slice_latitude": [
	-1,
	null,
	null
	],
	"slice_longitude": [
	180,
	null,
	null
	]
	},
	"cice-U": {
	"slice_latitude": [
	-1,
	null,
	null
	],
	"slice_longitude": [
	180,
	null,
	null
	]

But there is no connection here to stream, i.e. so this applies to inm and ind, onm and ond as appropriate.

The strange masking entries do not appear to be doing anything untoward as far as I can tell, but this section (and the halo_removal section) does hold irrelevant information.

I don't think this will have any impact on output data, but does add some confusion to the config files. I think it worth altering the get_masking_attributes function to correct this, but this isn't urgent.

I think we would need to compare the masking information in the plugins to the variables within this grid and remove irrelevant entries going into the maskings dictionary.

Note that while I've mostly talked about substreams above, the masking can be used elsewhere -- in regional model output for example.

[matthew-mizielinski]

Need to check where cice-U information on masking is being used in MIP convert