Qiskit / qiskit-fermions / 39914000349

98%

push

github

feat(operators): add min_num_modes to the operator types (#395)

Call sites kept deriving the number of modes an operator needs by
building the support set or scanning the index buffers by hand, each
with its own handling of the edge and Majorana cases. The mappers, for
instance, bound-checked a term against `num_qubits` through a helper
that only the vertex operators could use, and the Majorana mappers
each divided by two on their own.

Core gains `max_mode` as a required `OperatorTrait` method, a single
allocation-free scan instead of a `HashSet` build and sort, and a
provided `min_num_modes` on top of it. The latter is deliberately the
public quantity: it is what sizes a register, it is 0 for an operator
acting on no mode (no Option or out-pointer needed in Python and C),
and it is not confused with `num_support`, which counts distinct modes.
`max_mode` stays core-only to keep the public API small.

`max_mode` is always a fermionic mode. Majorana index `m` belongs to
mode `m / 2`, so `MajoranaOperator::max_mode` applies that map itself
rather than leaving it to every caller. This is the one place where it
differs from the maximum of `get_support`, and it is documented there
alone. As a result `min_num_modes` is `max_mode + 1` for every type, a
Majorana operator can never overflow it, and for the others an index of
`u32::MAX` makes the count unrepresentable and panics, which is
acceptable for an absurd operator.

The Jordan-Wigner and ternary-tree mappers now use `max_mode`, which
makes the `max_paired_index` helper redundant.

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

93 of 93 new or added lines in 12 files covered. (100.0%)

14031 of 14285 relevant lines covered (98.22%)

457552.32 hits per line