`CliffordSimp` does not simplify consecutive swaps

[daniel-mills-cqc]

For example

from pytket import Circuit
from pytket.passes import CliffordSimp

circuit = Circuit(3).CX(0,1).CX(1,0).CX(0,1).CX(1,0).CX(0,1).CX(1,0)
circuit_orig = circuit.copy()
CliffordSimp(allow_swaps=False).apply(circuit)
circuit_orig.n_2qb_gates(), circuit.n_2qb_gates()

demonstrates that the gate count before and after are the same. The compiled circuit should be the identity, with no swaps. I imagine swaps are added during the optimisation, even though the final circuit does not need them, so I'm not sure if this is expected behaviour. CliffordResynthesis does work when swaps are not allowed.

[willsimmons1465]

Yeah, this is expected because you are calling it with allow_swaps=False. The procedure for recognising this identity would go via identifying that each pair of successive CXs cancel to one CX and a swap, which requires us to be able to generate implicit swaps to begin reducing this down. The docs for the pass link to the paper that the pass replicates, which is simple pattern matching of two CXs at a time (with 1q Cliffords sandwiched in between).

Is this a case where ouor documentation needs to be more clear or are you happy that this is just the intended behaviour, @daniel-mills-cqc ?

[daniel-mills-cqc]

Thanks @willsimmons1465, if that's the intended behaviour then that's all fine.
As a user I intuitively expected allow_swaps to apply only to the final circuit generated but your explanation is clear.