Rename discipline variables

from __future__ import annotations

from numpy import array

from gemseo.core.discipline.data_processor import NameMapping
from gemseo.disciplines.analytic import AnalyticDiscipline

NameMapping is a DataProcessor to rename one or more variables of a Discipline.

In this example, we consider an AnalyticDiscipline computing the sum and the difference of two operands:

discipline = AnalyticDiscipline(
    {
        "sum": "first_operand+second_operand",
        "diff": "first_operand-second_operand",
    },
    name="SumAndDiff",
)
discipline.io.input_grammar.defaults = {
    "first_operand": array([1.0]),
    "second_operand": array([2.0]),
}

We want to use this discipline in a study to sum and subtract the input variables "x1" and "x2" and return the output variables "y1" and "y2". Unfortunately, its input names "first_operand" and "second_operand" and the output names "sum" and "diff" do not match the naming of our user study. For this simple example, we could have easily created a new AnalyticDiscipline with an expression dictionary using "x1", "x2", "y1" and "y2". but this solution is not generic because in practice, the discipline's _run() method manipulating the data from the variable names cannot be modified. To fix this problem, we can rename the input and output variables:

discipline.io.input_grammar.rename_element("first_operand", "x1")
discipline.io.input_grammar.rename_element("second_operand", "x2")
discipline.io.output_grammar.rename_element("sum", "y1")
discipline.io.output_grammar.rename_element("diff", "y2")

and set its DataProcessor to a NameMapping defined from a dictionary of the form {new_variable_name: variable_name}:

discipline.io.data_processor = NameMapping({
    "x1": "first_operand",
    "x2": "second_operand",
    "y1": "sum",
    "y2": "diff",
})

We can verify that the discipline can be executed correctly:

discipline.execute()
discipline.io.get_input_data(), discipline.io.get_output_data()

({'x1': array([1.]), 'x2': array([2.])}, {'y1': array([3.]), 'y2': array([-1.])})

This DataProcessor is compatible with the use of namespaces:

discipline.add_namespace_to_input("x1", "ns_in")
discipline.add_namespace_to_output("y1", "ns_out")
discipline.execute()
discipline.io.get_input_data(), discipline.io.get_output_data()

({'x2': array([2.]), 'ns_in:x1': array([1.])}, {'ns_out:y1': array([3.]), 'y2': array([-1.])})

Finally, we may also be interested in this example, which illustrates the use of rename_discipline_variables(). Given a collection of disciplines and a dictionary of translations generated either by hand or by a user-friendly interface, this function automates the process presented in the current example, namely renaming the input variables, renaming the output variables and using a NameMapping.