Qiskit / qiskit-addon-utils / 22691633167

91%

push

github

Expectation-value indexing bugfix (#145)

* fix bug w observable axis (WIP)

* require mask arg for `_apply_postselect_mask`

Docstring claims returned bool_array and basis_dict have extra bits appended to indicate postselection/rejection, but this was not happening when input `postselect_mask` was `None`. Simplest, least-surprising consistent behavior seems to be just not allowing use of `_apply_postselect_mask` if no mask is given, i.e. making `postselect_mask` a required arg rather than optional kwarg.

* require pec signs for `apply_pec_signs`

Docstring claims returned bool_array and basis_dict have extra bits appended to indicate PEC negation, but this was not happening when input `pauli_signs` was `None`. Simplest, least-surprising consistent behavior seems to be just not allowing use of `_apply_pec_signs` if no pauli signs are given, i.e. making `pauli_signs` a required arg rather than optional kwarg.

* handle `None` arg values outside function calls

Prior commits changed `_apply_postselect_mask` and `apply_pec_signs` so their arguments are not optional but required.

Here, add logic to only call those functions if args are not `None` (else handle appropriately).

* avoid masking indexing bug

Was seeing an indexing error somewhere in the in-place, broadcast, boolean-masked division.

Simplified these lines, improved commenting. No longer see the indexing error.

* add TODO comment

* Fix lint bugs (#148)

* (WIP) unit tests for executor_expectation_values

Not passing yet. First trying to deal with edge cases e.g. args that can be lists or integers...

* quick fixes to bugs in tests

* (WIP) more tests passing

* Fix bug with axes into np.mean

* expval test cleanup

* Tidy up

* Add release note. Fix type hint

* fix lint

* fix reno reference

* Hold off on testing against sparse observable for now

---------

Co-authored-by: Caleb Johnson <caleb.johnson@ibm.com>

39 of 40 new or added lines in 1 file covered. (97.5%)

745 of 822 relevant lines covered (90.63%)

0.91 hits per line