Behavior of set_run_options in composite experiments about qiskit-experiments HOT 6 CLOSED

The run options are completely determined by the parallel experiment. The run options of the sub-experiments are ignored.

from qiskit-experiments.

Options:

1. Do nothing, with the rationale that users are supposed to know that what counts is the run options of the parallel experiment. This option may confuse users who explicitly set say the number of shots of the sub-experiments (say to an equal number in all the sub-experiments), and don't even know that another number of shots was eventually picked (you have to look deeply inside the experiment data, at the counts, to see that they don't sum up to what you expect; so, if the number of shots is different than what you think, you may not even know it).
2. Raise a warning if run options of sub-experiments are different from the parallel experiment. This option, on the other hand, can be confusing to users who know that what counts is the parallel experiments, so will suffer from an over-triggering of warnings here.
3. Some hybrid approach, like raising a warning if the run options of sub-experiments is different from the default and from the parallel experiment - I'm not fond of this one either because it's complex.

What do you think?

from qiskit-experiments.

Same story also for transpile options

from qiskit-experiments.

For 1st release: option 2 and document expected behavior

from qiskit-experiments.

User should also be warned if their run options are being overridden by the provider

from qiskit-experiments.

Done in #159

from qiskit-experiments.