Add clifford gates to collect_cliffords

[ShellyGarion]

Summary

This PR adds some Clifford gates to the CollectClifford() transpiler pass.

Details and comments

This is an immediate bugfix which does not handle the following problems:

• When adding a gate to the Clifford class, it should be added in other places of the code,
  and there is no single list of gates that the Clifford class supports.
  

  qiskit/qiskit/transpiler/passes/optimization/collect_cliffords.py

  Line 27 in 6b57b78

  class CollectCliffords(CollectAndCollapse):

  
  and
  

  qiskit/qiskit/circuit/random/utils.py

  Line 290 in fa3d6df

  def random_clifford_circuit(num_qubits, num_gates, gates="all", seed=None):

• The current Clifford class can handle certain u gates (for certain angles), and this does not appear in the docs, neither in the code above:
  

  qiskit/qiskit/quantum_info/operators/symplectic/clifford_circuits.py

  Line 100 in fa3d6df

  if isinstance(gate, Gate) and name == "u" and len(qargs) == 1:

[qiskit-bot]

One or more of the following people are relevant to this code:

• @Qiskit/terra-core

[coveralls]

Pull Request Test Coverage Report for Build 9890871127

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 5 of 5 (100.0%) changed or added relevant lines in 2 files are covered.
• 50 unchanged lines in 5 files lost coverage.
• Overall coverage decreased (-0.001%) to 89.893%

---

Files with Coverage Reduction	New Missed Lines	%
crates/qasm2/src/expr.rs	1	94.02%
crates/qasm2/src/lex.rs	6	92.37%
crates/circuit/src/dag_node.rs	7	85.58%
qiskit/transpiler/passes/basis/basis_translator.py	7	97.7%
qiskit/quantum_info/operators/symplectic/pauli.py	29	85.98%

Totals
Change from base Build 9871528102:	-0.001%
Covered Lines:	65747
Relevant Lines:	73139

---

💛 - Coveralls

[alexanderivrii]

Thanks @ShellyGarion, this looks good! If we want a more general approach that supports whatever is supported by clifford_circuits, what do you think about changing the filter_fn to the following:

def _is_clifford_gate(node):
    """Specifies whether a node holds a clifford gate."""
    if getattr(node.op, "condition", None) is not None:
        return False
    is_clifford_gate = True
    try:
        Clifford(node.op)
    except QiskitError:
        is_clifford_gate = False
    return is_clifford_gate

or

def _is_clifford_gate(node):
    """Specifies whether a node holds a clifford gate."""
    if getattr(node.op, "condition", None) is not None:
        return False
    if node.op.name in clifford_gate_names:
        return True
    # test other gates
    is_clifford_gate = True
    try:
        Clifford(node.op)
    except QiskitError:
        is_clifford_gate = False
    return is_clifford_gate

My only concern with these suggestions is whether they are not prohibitively slow for large clifford circuits.

UPDATE: unfortunately the suggestions do make the pass slower: a quick experiment on 10 random circuits (created by circuits = [random_circuit(20, 1000, max_operands=4, seed=seed) for seed in range(10)]) gives

• creating circuits: 6.13 seconds
• name-based method: 15.97 seconds
• Clifford(gate)-based method: 82.83 seconds

[alexanderivrii]

another suggestion: can we use _BASIS_1Q and _BASIS_2Q defined at the end of clifford_circuits.py?

[ShellyGarion]

@alexanderivrii - thanks for your suggestion. I also fixed random_clifford_circuit in a similar way.
I think it's a good solution, although it still does not handle certain u gates (and other 1 and 2-qubit gates) which are clifford for certain angles.

[ShellyGarion]

should we also add "permutation" here?

[alexanderivrii]

I think it's a good idea. I have checked and we can indeed construct a Clifford from a circuit containing PermutationGate objects, so it makes sense for the collection pass to handle these as well

[alexanderivrii]

LGTM, thanks for doing this!