Fixup partition file presence

[cbegeman]

Include graph partition file only when the ocean model is MPAS-Ocean so there aren't unnecessary files in Omega tests.

Checklist

• User's Guide has been updated
• Developer's Guide has been updated
• API documentation in the Developer's Guide (api.md) has any new or modified class, method and/or functions listed
• Documentation has been built locally and changes look as expected
• Testing comment in the PR documents testing used to verify the changes

[cbegeman]

Testing

PASS manufactured solution init and forward steps for MPAS-Ocean and Omega

PASS tests from all suites

[cbegeman]

@xylar Do you want to take a quick look and let me know if this is what you had in mind?

[xylar]

Looks good! I'd suggest a rename graph_filename --> graph_target to avoid confusion, see my detailed comments.

[xylar]

Let's also rename the parameter from graph_filename to graph_target to avoid confusion.

Then, self.graph_target doesn't need to be an attribute since it's not used except below to add the input file.

[cbegeman]

Looks good! I'd suggest a rename graph_filename --> graph_target to avoid confusion, see my detailed comments.

@xylar Sounds good! I'll make that change and retest

[xylar]

@cbegeman, it looks like my suggestion (not having the graph_target attribute) won't work. I didn't realize it was getting used in the setup() method, I thought it was later in the __init__() method. Sorry for the confusion.

I don't think we want to add new parameters to the setup() method because that will cause extra burden on classes that descent from OceanModelStep() (they will always need a setup()` method). Let's go back to how you had it. Sorry about that!

I still think renaming the parameter to graph_target is important. And then the corresponding graph_target attribute to the class needs to be added to the doc string for the class.

[cbegeman]

@xylar How about these code changes? It may be inconvenient sometimes to get the path from the base work dir rather than the relative path, but I think this is ok.

[cbegeman]

And yes, I'll fixup the doc strings once we settle on the code

[xylar]

@cbegeman, yes, I think this could work. One piece is that the default value (in 3 places) of graph_target='graph.info' presumably doesn't make sense anymore. I think we have to provide graph_target the way things are written and it shouldn't have a default value anymore. But the alternative would be to set self.make_graph = True if graph_target=='graph.info' and then not try set the target parameter in that situation when you add graph.info as an input file. We can talk about this more when we meet shortly.

[cbegeman]

I noticed that the default value didn't make sense but I didn't think of the make_graph approach. Maybe in the case where we make_graph we make the default graph_target=None? Do we need the ability to specify the file name of the graph?

[xylar]

Right, I think graph_target=None is the right default. And then we just need:

if self.graph_target is None:
    self.make_graph = True`

Adding the input file should work without further modification.

[xylar]

@cbegeman, This looks great!

I've run a whole bunch of tests. I'm still working my way through the spherical tests but no sign of trouble so far.

update: all the spherical tests passed.

I also set up and ran manufactured_solution with Omega. It ran successfully but then gave an error (as expected) on the missing output.nc file. Also, as expected, there is no graph.info file or partition file in the forward directory.