OpenMDAO 1.x组连接中目标索引的便捷实现方法问询

Great question! In OpenMDAO 1.x, the built-in connect method doesn't natively support target indexing—meaning you can't directly implement an operation like b[1] = a via a simple connect parameter. As you noted, src_indices works the other way: self.connect('a', 'b', src_indices=[1]) effectively assigns b = a[1], pulling a specific index from the source into the full target variable.

To achieve the reverse (assigning a source value to a specific index of a target variable), you'll need to use a small intermediate component to handle the indexing logic. Here are two reliable approaches:

1. Use an Explicit Component for Target Assignment #

The cleanest and most maintainable way is to create a lightweight ExplicitComponent that takes your source variable a and outputs the target variable b with the desired index set. This component acts as a bridge between the source and target, giving you full control over the assignment.

Example code:

from openmdao.api import ExplicitComponent, Group, Problem

class TargetIndexSetter(ExplicitComponent):
    def setup(self):
        # Define input (source variable 'a')
        self.add_input('a', shape=1)
        # Define output (target variable 'b'—adjust shape to match your use case)
        self.add_output('b', shape=3)

        # Declare partial derivatives (required for explicit components)
        # We're only modifying index 1 of 'b' based on 'a'
        self.declare_partials('b', 'a', rows=[1], cols=[0])

    def compute(self, inputs, outputs):
        # Initialize the output variable (adjust default values as needed)
        outputs['b'] = 0.0
        # Assign the source value to the target's specific index
        outputs['b'][1] = inputs['a']

# Usage in a group
class MyGroup(Group):
    def setup(self):
        # Add your existing components here...
        self.add_subsystem('index_setter', TargetIndexSetter())
        
        # Connect the source 'a' to the component's input
        self.connect('a', 'index_setter.a')
        # Connect the component's output to your target 'b'
        self.connect('index_setter.b', 'b')

prob = Problem()
prob.root = MyGroup()
prob.setup()

2. Direct Variable Assignment (Limited Use Cases) #

If you're working in a scenario where the target variable b is already initialized and you just need to set a specific index during setup/configure, you could use set_val with variable references. However, this approach is less flexible and depends on execution order, so it's not recommended for most cases:

class MyGroup(Group):
    def configure(self):
        # Get references to the source and target variables
        a_val = self.get_val('a')
        # Set index 1 of 'b' to the value of 'a'
        self.set_val('b', indices=[1], val=a_val)

Keep in mind this method won't automatically update if a changes during the optimization/analysis loop, so the explicit component approach is far more robust for dynamic cases.

内容的提问来源于stack exchange，提问作者gbarter