Build a discipline from a simple Python function¶

Let’s consider a simple Python function, e.g.:

def f(x=0., y=0.):
    """A simple Python function"""
    z = x + 2*y
    return z

Then, we can consider the AutoPyDiscipline to convert it into an MDODiscipline.

Create and instantiate the discipline¶

For that, we can use the create_discipline() API function with AutoPyDiscipline as first argument:

from gemseo import create_discipline
from numpy import array

disc = create_discipline('AutoPyDiscipline', py_func=f)

The original Python function may or may not include default values for input arguments, however, if the resulting AutoPyDiscipline is going to be placed inside an MDF, a BiLevel formulation or a BaseMDA with strong couplings, then the Python function must assign default values for its input arguments.

Execute the discipline¶

Then, we can execute it easily, even considering default inputs:

print(disc.execute())

which results in:

{'y': array([ 0.]), 'x': array([ 0.]), 'z': array([ 0.])}

or using new inputs:

print(disc.execute({'x': array([1.]), 'y':array([-3.2])}))

which results in:

{'y': array([-3.2]), 'x': array([ 1.]), 'z': array([-5.4])}

Optional arguments¶

Optional arguments are:

py_jac=None: The Python function to compute the Jacobian which must return a 2D numpy array,
use_arrays=False: if True, the function is expected to take arrays as inputs and give outputs as arrays,
grammar_type=MDODiscipline.GrammarType.JSON: The type of grammar to be used.

Here is an example of Jacobian function, returning a 2D matrix. The rows of the matrix correspond to the derivatives of the outputs, the columns correspond to the variables with respect to the outputs are derived.

def dzdxy(x=0., y=0.):
    """Jacobian function of z=f(x,y)"""
    jac = array((1,2))
    jac[0, 0] = 1.
    jac[0, 1] = 2.
    return jac