Differences in ParameterSet.calculate_static() and calculate_stochastic() causes problems with PBM

[aleksandra-kim]

Original report by Pascal Lesage (Bitbucket: MPa, ).

---

A ParameterizedBrightwayModel (pbm) has two methods for recalculating parameter values, e.g. if presamples are loaded using load_existing .

1) pbm.calculate_stochastic(), which calls the evaluate_monte_carlo method of ParameterSet. 2) pbm.calculate_static(), which calls the evaluate method of ParameterSet.

In both cases, the pbm can update amounts. The code is:

if update_amounts:
        for key, value in self.data.items():
            value['amount'] = result[key]

This works fine with calculate_stochastic, but not with calculate_static. The reason is this:

• When one calls pbm.load_existing( [presample_packages]) , all loaded presamples are copied to global_params, but are not removed from data. This is as expected.

• The parameter_set.calculate_static() does not add global parameters to result.

• Hence, when iterating through the pbm.data, all parameters that were loaded from existing are still in pbm.data but are missing from the result dict obtained from calculate_static, leading to KeyErrors.

There are multiple solutions, the two most evident being:
1) In  ParameterizedBrightwayModel.calculate_static(), catch and pass global parameters on update:

if update_amounts:
        for key, value in self.data.items():
            if key in self.global_params:
                pass
            else:
                value['amount'] = result[key]

This works, since global parameters are not updated by ParameterSet.evaluate().

2) In ParameterSet.evaluate(), add global params to results (as is done in evaluate_monte_carlo). This simply means replacing one line:

interpreter.symtable[key] = self.global_params[key]

becomes

interpreter.symtable[key] = result[key] = self.global_params[key]

Both work, but the latter adds consistency in ParameterSet, hence my dropping this here rather than changing the ParameterizedBrightwayModel.

[aleksandra-kim]

Original comment by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

---

Return global parameter values after caalculate_static. Fixes #4

[aleksandra-kim]

Original comment by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

---

Included in 0.6.6, available on pypi and conda.org.