Making wires the last positional argument for several classes

[PietropaoloFrisoni]

Context: Currently, as noticed in this issue, qml.GateFabric, qml.AllSinglesDoubles, and qml.BasisRotation do not have wires as the last positional argument, which is a requirement for pennylane operations. This might be confusing for users.

Description of the Change: In qml.GateFabric, we made init_state a kwarg instead of a positional argument, with an initial default state as the all-zeros state. For qml.AllSinglesDoubles and qml.BasisRotation we simply changed the argument order. Now, wires is the rightmost positional argument in all three cases.

Benefits: Now, the three Python classes above have an argument order in accordance with pennylane requirements. Additionally, now qml.AllSinglesDoubles also accepts a list as hf_state parameter (not necessarily a numpy array).

Possible Drawbacks: In the case of qml.AllSinglesDoubles and qml.BasisRotation, this is technically a breaking change since we changed the argument order:

import pennylane as qml
import numpy as np

weights = np.ones(3)
singles = [[0, 2], [1, 3]]
doubles = [[0, 1, 2, 3]]
hf_state = [1, 1, 0, 0]
wires=[0, 1, 2, 3]

# works fine
qml.AllSinglesDoubles(weights, hf_state, wires, singles=singles, doubles=doubles)

# works fine
qml.AllSinglesDoubles(weights=weights, wires=wires, hf_state=hf_state, singles=singles, doubles=doubles)

# fails since the second argument is now interpreted as hf_state
# qml.AllSinglesDoubles(weights, wires, hf_state, singles=singles, doubles=doubles)

Related GitHub Issues: #5521

[sc-61430]

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.68%. Comparing base (756dc62) to head (194f0a2).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5643      +/-   ##
==========================================
- Coverage   99.69%   99.68%   -0.01%     
==========================================
  Files         412      412              
  Lines       38619    38352     -267     
==========================================
- Hits        38500    38233     -267     
  Misses        119      119              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dwierichs]

Looks good to me, thanks for the quick fix @PietropaoloFrisoni ! 💯

I might have overlooked it, but is there any test where BasisRotation is instantiated with unitary_matrix as positional argument? Might be good to have that in as well, just by changing some arbitrary test. Maybe the test_standard_validity test :)

[PietropaoloFrisoni]

Thanks @dwierichs. Actually, I changed the order of the arguments in the tests as well (even where it was not technically necessary) just to be consistent. in test_standard_validity I noticed just now that there was a weights variable instead of unitary_matrix, so I replaced it.

[dwierichs]

Thanks @PietropaoloFrisoni I iterated my suggestion. It's good to go from my side either way.

[albi3ro]

This PR is looking good, and I'm sorry to bring this up right now, but I'd like to propose another option for AllSinglesDoubles.

We coudl instead default hf_state=None, and then if hf_state is None, we simply don't include the BasisState op in the decomposition.

[PietropaoloFrisoni]

@albi3ro No problem, I am fine with everything the @PennyLaneAI/algorithms is happy with

[PietropaoloFrisoni]

Closing the PR since it has been decided not to fix #5521