Support controlled ZYZ decomposition of composite operators

[Mandrenkov]

Context:

Attempting to run the preliminary example from the QSVT in Practice demo with the Lightning Qubit device

import pennylane as qml
from pennylane import numpy as pnp

dev = qml.device("lightning.qubit", wires=[0, 1])

A = pnp.array([[0.1, 0.2], [0.3, 0.4]])
phase_angles = pnp.array([0.0, 1.0, 2.0, 3.0])


@qml.qnode(dev)
def my_circuit(phase_angles):
    qml.qsvt(A, phase_angles, wires=[0, 1])
    return qml.state()

print(my_circuit(phase_angles))

yields the error

  File "/Users/.../PennyLane-0.35.0.dev0-py3.11.egg/pennylane/ops/op_math/controlled_decompositions.py", line 188, in ctrl_decomp_zyz
    phi, theta, omega = target_operation.single_qubit_rot_angles()
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'Prod' object has no attribute 'single_qubit_rot_angles'

Description of the Change:

• Updated ctrl_decomp_zyz() to support Operators that are not Operations.
• Trimmed trailing whitespace from the changelog (done automatically by VS Code).

Benefits:

• The QSVT in Practice demo can be executed on the Lightning Qubit device.

Possible Drawbacks:

None.

Related GitHub Issues:

None.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (e166f7e) 99.68% compared to head (dc5bf45) 99.67%.
Report is 3 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5242      +/-   ##
==========================================
- Coverage   99.68%   99.67%   -0.01%     
==========================================
  Files         399      399              
  Lines       36848    36567     -281     
==========================================
- Hits        36731    36449     -282     
- Misses        117      118       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[timmysilv]

since we were already doing the try-except thing, do you know if it hurts to instead just update the except to catch (NotImplementedError, AttributeError)?

[Mandrenkov]

Thanks for taking a look, @timmysilv!

since we were already doing the try-except thing, do you know if it hurts to instead just update the except to catch (NotImplementedError, AttributeError)?

That also works! The main difference is that that doing so would also catch AttributeErrors inside the implementation of single_qubit_rot_angles() which might not be preferable. I opted for the if-else approach because:

1. It explicitly documents the assumption that the Operator is an Operation (in that branch).
2. In my opinion,AttributeError is a bit too broad of an exception to catch. A lot of silly mistakes can get caught here which are better left exposed (and subsequently fixed).

[Mandrenkov]

Thanks for the review, @albi3ro!

[albi3ro]

Thanks for the updates :)