Qiskit / qiskit-addon-slc / 37622536777

96%

push

github

Test suite improvements (#118)

* fix: two bugs reachable through the public bounds API

`compute_local_scales` read the sampling cost off the last selected term, which
raised `IndexError` when no term was selected at all. That is reachable through
each of the three conditions making up the selection: a sampling cost budget
below the cheapest single term, a bias tolerance already above the total bias
bound, or a set of commutator bounds that are all zero. None of those are
malformed inputs. Mitigating nothing is a meaningful outcome, so the sampling
cost overhead is now reported as 1.0 rather than raising; the other two return
values were already correct for this case.

`time_evolved_norm_forward` defaulted `evolution_max_terms` to the largest
unsigned integer. The propagation routine treats that argument as a capacity to
pre-allocate rather than as an upper limit, so calling the function with its own
defaults asked the allocator for terabytes and aborted the process. All four
call sites now share one constant, keeping the public functions' existing and
already realistic default.

Closes #115.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* test: cover bound merging and speed-limit tightening directly

Both modules were previously reached only by the end-to-end test, which runs the
whole workflow on a 50-qubit circuit and compares against pickled fixtures. That
makes it a good regression test but a poor unit test: it exercises one path
through each module and cannot say which part of a change broke it.

These tests drive both modules on circuits small enough to reason about, so the
behaviour itself can be asserted rather than compared against a fixture: that
tightening never loosens a bound and leaves its input untouched, that an error
outside the observable's light cone is tightened to zero, that every accepted
single-term observable type gives the same answer, and that the documented error
paths raise.

Co-Authored-By: Claude Opus 5 <norep... (continued)

7 of 8 new or added lines in 4 files covered. (87.5%)

1278 of 1335 relevant lines covered (95.73%)

0.96 hits per line