Report offending input in sequence module error messages

[jaewonyun1234]

Hi @a-corni, @HGSilveri,

Following on from #1095 and #1097, this covers the sequence module (38 messages: 35 in sequence.py, 2 in _schedule.py, 1 in _seq_drawer.py).

Error messages that only stated a rule now also report the value that broke it. _schedule.py reports the channel with no target and the duration that overflowed; _seq_drawer.py reports the channels that aren't in the rydberg basis. Qubit-id errors still list the declared ids — those sets are small and useful to see.

A few were more than wording:

• The 'Microwave' clash in declare_channel interpolated the channel object, so it printed the full spec (~236 characters) instead of the name the user chose. It now reports the name and id, e.g. got 'ryd' ('rydberg_global') with the declared {'mw': 'mw_global'}.
• estimate_added_delay had two adjacent string literals with no space, so it read ...if sequence orpulse is parametrized.
• _schedule.get_samples raises when qubits is missing for a DMM channel, but no test reached it. Added one.

Also aligned with the register review: 'device' must be an instance of 'BaseDevice'; an unavailable DMM lists the still-available ids; a local channel that exceeds max_targets reports the limit, the count, and the given targets.

Messages that describe a state rather than a rejected input are left unchanged (_in_ising transitions, qubit_info, register, get_measurement_basis, SLM on a mappable register, _target with no qubits, empty-sequence drawing).

Existing match= assertions now cover the full message (re.escape where needed). Partially addresses #1057.

[HGSilveri]

Nice going @jaewonyun1234 !
Here are a few minor comments, keep up the good work :)

[HGSilveri]

You can just iterate over the last.targets here

[jaewonyun1234]

last.targets is a set[QubitId], so iterating it directly builds the dict in hash order, which changes from run to run. Going through self._register.qubit_ids and filtering is what keeps the printed order stable.

With your version, test_phase fails on 7 of 8 PYTHONHASHSEED values. On the 28-qubit fixture register the same error prints as:

seed 0: got {'q9': 0.0, 'q0': 1.0, 'q7': 0.0, 'q15': 0.0, ...}
seed 1: got {'q17': 0.0, 'q9': 0.0, 'q11': 0.0, ..., 'q0': 1.0, ...}

q0 is the only qubit with a different reference, and it moves around each run.

That said, you're right that the loop doesn't show why it's written that way. Would you prefer a small helper, used here and at the two other spots you flagged, plus the same pattern in config_slm_mask?

def _in_register_order(self, qubits: Collection[QubitId]) -> list[QubitId]:
    """Qubit ids in the register's order, so error messages are reproducible."""
    return [q for q in self._register.qubit_ids if q in qubits]

Separately on this message: on a global channel it lists every qubit, so 27 of the 28 entries are the same 0.0 and only q0 differs. Would you rather it grouped them by reference value, or showed only the ones that stand out?

[jaewonyun1234]

Or, if a new method feels like too much for this, I can just add a one-line comment at each spot instead:

# Iterate the register, not the target set, so the order is reproducible

Happy either way — whichever you prefer.

[HGSilveri]

How about iterating over sorted(last.targets)?

[HGSilveri]

Same here

[jaewonyun1234]

Same reason as the thread above — these are sets, so the register loop is what keeps the printed order stable. Happy to switch all three to a helper if you prefer that.

[HGSilveri]

And here

[jaewonyun1234]

Same reason as the thread above — these are sets, so the register loop is what keeps the printed order stable. Happy to switch all three to a helper if you prefer that.