Use qualified rather than local names in generated model code

[MichaelClerx]

In generated code for:

• Inputs
• Outputs
• Update code_generation.py to add an output for every rdf term in pvar.outputs

In the protocol language for

• Lookups like sim:membrane_voltage would need some oxmeta bit in them

And then

• Some fallback mechanism so that the old syntax still works, or else a deprecation warning and manually updating a whole bunch of protocols?

[jonc125]

We can't use the qualified names as-is (something like oxmeta:membrane_voltage isn't a valid Cython identifier!) but could use e.g. oxmeta__membrane_voltage - is that what you mean?

This would I think just mean tweaking the variable_name function in code_generation.py.

[MichaelClerx]

No I was thinking about these lines: https://github.com/ModellingWebLab/weblab-fc/blob/master/test/code_generation/weblab_model.pyx#L187-L190

[MichaelClerx]

The model variable names are already unique, even if they're from different ontologies

[jonc125]

OK, I see. That would need changes in ModelWrapperEnvironment (and how it's set up) rather than in that dictionary, so that different prefixes map different subsets of the model variables. Currently as a hack (because pycml wasn't that clever) we have one environment pretending to be multiple ontologies. What you really want is one environment per ontology, then each has an independent set of local names.

[jonc125]

Not urgent though, because we don't have any protocols with conflicting local names yet!

[MichaelClerx]

I think it is urgent because I've updated the ModelInterface code to allow multiple rdf terms to annotate the same variable. I could hack in support so that it then extracts multiple local_name arguments from those rdf terms, but I'd much rather just do it right straight away?

[jonc125]

For now, just add a check that you don't get duplicate local names. That'll let #158 be merged, and we can address this separately.