Add iterator-access methods to SparseObservable
#13370
Conversation
jakelishman
added
on hold
|
One or more of the following people are relevant to this code:
|
jakelishman
force-pushed
the
sparse-observable/iterators
branch
from
6389829
to
3ec4289
Compare
Pull Request Test Coverage Report for Build 11684850034Warning: 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.
Details
💛 - Coveralls |
jakelishman
force-pushed
the
sparse-observable/iterators
branch
from
3ec4289
to
6b45a62
Compare
jakelishman
force-pushed
the
sparse-observable/iterators
branch
from
6b45a62
to
ac46ea1
Compare
|
Rebased and no longer on hold. |
This adds a new related class, `SparseObservable.Term` to be the item type of the iteration methods, and imbibes it with the simplest mathematical binary relations with `SparseObservable` as the other operand. For some reason, in my head, the methods to convert terms and observables to their corresponding Pauli basis are related to iteration? The `SparseObservable.pauli_bases` and `SparseObservable.Term.pauli_base` methods perform a similar bitwise trick to do this reduction as exemplified in the documentation, but return the other `quantum_info`-native classes `PauliList` and `Pauli`, respectively. These methods are expected to be used by the primitives.
jakelishman
force-pushed
the
sparse-observable/iterators
branch
from
ac46ea1
to
0aa0080
Compare
|
Rebased again, since it had a minor conflict with #13372. This also uses the new |
| @@ -697,7 +757,7 @@ impl From<LabelError> for PyErr { | |||
| /// observable generate only a small number of duplications, and like-term detection has additional | |||
| /// costs. If this does not fit your use cases, you can either periodically call :meth:`simplify`, | |||
| /// or discuss further APIs with us for better building of observables. | |||
| #[pyclass(module = "qiskit.quantum_info")] | |||
| #[pyclass(module = "qiskit.quantum_info", sequence)] | |||
There was a problem hiding this comment.
Are len and __getitem__ the only two methods we need to add to qualify this as a Sequence? 🙂
There was a problem hiding this comment.
Python actually has two concepts of "sequence" - one at the low level (which is this one) and collections.abc.Sequence. The requirements of the C API sequence is just to have a __getitem__ that acts on sequential integers and a correct __len__. Here's the glossary term in their docs with a brief explanation: https://docs.python.org/3/glossary.html#term-sequence.
The C-API-level "sequence" (as here) implies that a few functions are filled in on the corresponding Python type object as C-level function pointers, rather than needing Python-space look-ups to do. That can make it quite a bit more efficient to interact with the object from the C level. A bit more detail here: https://docs.python.org/3/c-api/typeobj.html#c.PySequenceMethods.
There was a problem hiding this comment.
Ok I see! And a dumb question: in the glossary it requires __len__ but here we're adding len, is that an issue?
There was a problem hiding this comment.
I see us as adding __len__?
There was a problem hiding this comment.
Lol me too, I don't know what happened, maybe my brain parsed the double underscore into len 😂
Summary
This adds a new related class,
SparseObservable.Termto be the item type of the iteration methods, and imbibes it with the simplest mathematical binary relations withSparseObservableas the other operand.For some reason, in my head, the methods to convert terms and observables to their corresponding Pauli basis are related to iteration? The
SparseObservable.pauli_basesandSparseObservable.Term.pauli_basemethods perform a similar bitwise trick to do this reduction as exemplified in the documentation, but return the otherquantum_info-native classesPauliListandPauli, respectively. These methods are expected to be used by the primitives.Details and comments
Built on top of #13298.
There's still one more part of
SparseObservableto go -apply_layout,composeandevolve- then potentially a follow-on PR that reorganises the class into two really separate components; a pure Rust core for Python-free consumption, and apyclasswrapper around it.