Fix conjugated_by of PauliString. #7065
Conversation
| def _has_stabilizer_effect_(self) -> Optional[bool]: | ||
| if self._is_parameterized_(): | ||
| return None | ||
| return self.exponent % 1 == 0 |
There was a problem hiding this comment.
ISwapPow gate need to have _has_stabilizer_effect_ method just like SwapPowGate, fix it here, otherwise, unit tests would fail in pauli_string_test.py.
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #7065 +/- ##
==========================================
- Coverage 98.17% 98.15% -0.02%
==========================================
Files 1089 1089
Lines 95177 95253 +76
==========================================
+ Hits 93438 93494 +56
- Misses 1739 1759 +20 ☔ View full report in Codecov by Sentry. |
babacry
force-pushed
the
conj
branch
5 times, most recently
from
77f1722 to
a39e761
Compare
| @@ -377,7 +377,10 @@ def __init__(self, *, _clifford_tableau: qis.CliffordTableau) -> None: | |||
| # ZI [ 0 0 | 1 0 | 1 ] | |||
| # IZ [ 1 0 | 1 1 | 0 ] | |||
| # Take the third row as example: this means the ZI gate after the this gate, | |||
| # more precisely the conjugate transformation of ZI by this gate, becomes -ZI. | |||
| # more precisely the conjugate transformation of ZI by this gate, becomes -ZI: | |||
There was a problem hiding this comment.
Conjugation definitions are not uniform in Cirq, it turns out the conjugation definition here is opposite with the conjugation definition in PauliString.conjugated_by.
Here, in CliffordGate, in terms of matrix conjugation it is ---C^{-1}----P-----C.
In PauliString, the definition in terms of matrix conjugation is -----C----P------C^{-1}.
Thus I added comments here to clarify the difference.
|
Thanks Nour for the inputs! Updated. PTAL. |
Rewrite PauliString.conjugated_by.
Context: previous conjugated_by relies on decompose_into_cliffords which crashes in for some cliffords: #6946.
Followup: note
pass_operations_over()andinplace_after()still rely on_decompose_into_cliffords_and_pass_operation_over_, I will clean them up in the followup PRs. (tried to clean them up in the same pr, but too many nit updates are needed in both the class itself and the tests)