observe has unexpected behavior with an if statement and mz

[mmvandieren]

Required prerequisites

• Consult the security policy. If reporting a security vulnerability, do not report the bug using this form. Use the process described in the policy to report the issue.
• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
• If possible, make a PR with a failing test to give us a starting point to work on!

Describe the bug

The mz gives an error message if it is in an unexecuted if statement when using observe. However, if instead the kernel refers to that mz command by calling a sub kernel then no error message is flagged. See the working and broken code examples.

use case: the if statement was prudent if I wanted to have just one kernel that I needed to sample a subset of qubits, but I also wanted to compute expectation values for. for instance in a variational quantum random walk with auxiliary qubits

Steps to reproduce the bug

import cudaq

@cudaq.kernel()
def broken_kernel(with_measure : int):
q=cudaq.qvector(2)
if with_measure == 1:
mz(q)

hamiltonian = cudaq.spin.z(0) + cudaq.spin.z(1)

#the next line breaks the code
result = cudaq.observe(broken_kernel, hamiltonian, 0).expectation()

#but if we nest the mz command into a separate kernel, we don't get an error message
@cudaq.kernel()
def measure(qubits: cudaq.qview):
mz(qubits)

@cudaq.kernel
def working_kernel(with_measure : int):
q=cudaq.qvector(2)
if with_measure == 1:
measure(q)

result = cudaq.observe(working_kernel, hamiltonian, 0).expectation()

Expected behavior

expected that both versions of the kernel would either both give errors or both work ok (both working ok would be preferred)

Is this a regression? If it is, put the last known working version (or commit) here.

Not a regression

Environment

• CUDA-Q version: cu12-0.10.0
• Python version: 3.10.12
• C++ compiler:
• Operating system: Mac OS

Suggestions

No response

[mmvandieren]

oops, here's the code properly formatted

@cudaq.kernel()
def broken_kernel(with_measure : int):
    q=cudaq.qvector(2)
    if with_measure == 1:
        mz(q)

hamiltonian = cudaq.spin.z(0) + cudaq.spin.z(1)

#result = cudaq.observe(broken_kernel, hamiltonian, 0).expectation()

@cudaq.kernel()
def measure(qubits: cudaq.qview):
    mz(qubits)
    
@cudaq.kernel
def working_kernel(with_measure : int):
    q=cudaq.qvector(2)
    if with_measure == 1:
        measure(q)


result = cudaq.observe(working_kernel, hamiltonian, 0).expectation()```

with error message
```---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
Cell In[8], [line 11](vscode-notebook-cell:?execution_count=8&line=11)
      7         mz(q)
      9 hamiltonian = cudaq.spin.z(0) + cudaq.spin.z(1)
---> [11](vscode-notebook-cell:?execution_count=8&line=11) result = cudaq.observe(broken_kernel, hamiltonian, 0).expectation()
     13 @cudaq.kernel()
     14 def measure(qubits: cudaq.qview):
     15     mz(qubits)

File /opt/nvidia/cudaq/cudaq/runtime/observe.py:83, in observe(kernel, spin_operator, shots_count, noise_model, num_trajectories, execution, *args)
     81 validityCheck = cudaq_runtime.isValidObserveKernel(kernel)
     82 if not validityCheck[0]:
---> [83](https://vscode-remote+attached-002dcontainer-002b7b22636f6e7461696e65724e616d65223a222f636f6f6c5f63757272616e222c2273657474696e6773223a7b22636f6e74657874223a226465736b746f702d6c696e7578227d7d.vscode-resource.vscode-cdn.net/opt/nvidia/cudaq/cudaq/runtime/observe.py:83)     raise RuntimeError('observe specification violated for \'' +
     84                        kernel.name + '\': ' + validityCheck[1])
     86 spin_operator = to_spin_op(spin_operator)
     87 if isinstance(spin_operator, list):

RuntimeError: observe specification violated for 'broken_kernel': kernels passed to observe cannot have measurements specified.```