Two compile methods are redundant

[ajavadia]

What is the expected enhancement?

Currently we have:

1. wrapper.compile()

def compile(circuits, backend,
            config=None, basis_gates=None, coupling_map=None, initial_layout=None,
            shots=1024, max_credits=10, seed=None, qobj_id=None, hpc=None,
            skip_transpiler=False)

2. transpiler.compile()

def compile(circuits, backend,
            config=None, basis_gates=None, coupling_map=None, initial_layout=None,
            shots=1024, max_credits=10, seed=None, qobj_id=None, hpc=None,
            pass_manager=None)

The latter should just be removed. The first will have a pass_manager option.

Also see #978 about other simplifications to these function signatures.

[jaygambetta]

I almost agree but let’s keep it a top-level function. How much is skipped translation used.

[jaygambetta]

Ok I'm going to self-assign and work with @nonhermitian on the implementation and it answers his #1032

in the transpiler.compile you pass it circuits it returns compiled circuits

parallel_for():
       circuit -> dag
       dag -> dag
       dag -> circuit

wrapper.comple

compiled_circuits = transpiler.compile(circuits)
qobj = circuits2qobj([compiled_circuits])

Basically, there is no qobj in the transpiler.

[ss2165]

@ajavadia I see this issue has been closed but I don't see that compile in _compiler.py has as 'pass_manager' parameter, is this still to come?

The latter should just be removed. The first will have a pass_manager option.

[ajavadia]

@ss2165 yes that will replace the skip_transpiler, and if you want to skip the transpiler you just pass in an empty pass manager.
now there are a few follow up issues we are working on to make this all make sense (#1144, #1145, #1146, #1163).