how to chain first and last from subchains to chains properly in airflow?

Consider the following:

sod = (
  DummyOperator(task_id="sod")
  >> DummyOperator(task_id="sod_do_this")
  >> DummyOperator(task_id="sod_last")
)

(
  BranchPythonOperator(
    task_id="should_run_sod",
    python_callable=lambda: "sod", # no_sod
  )
  >> [sod, DummyOperator(task_id="no_sod")]
  >> DummyOperator(task_id="end")
)

The idea is to "connect" a subchain "sod" in between. Hoewever, __rshift__ returns the last operator in chain, so sod = DummyOperator(task_id="sod_last") and the stuff becomes mixed. should_run_sod is connected to sod_last, not to sod.

Can I write it in some simple way, other than assigning variables everywhere? I would want to get the same result as in the following, this however requires variables for last and first, which becomes more convoluted

sod = DummyOperator(task_id="sod")
sod_do_this = DummyOperator(task_id="sod_do_this")
sod_last = DummyOperator(task_id="sod_last")
sod >> sod_do_this >> sod_last 

should_run_sod = BranchPythonOperator(
  task_id="should_run_sod",
  python_callable=lambda: "sod", # no_sod
)
should_run_sod >> [sod, DummyOperator(task_id="no_sod")]

end = DummyOperator(task_id="end")
sod_last >> end
no_sod >> end

TaskGroup can be used to store your subchain.

with TaskGroup(group_id='sod') as sod:
    DummyOperator(task_id="sod") 
    >> DummyOperator(task_id="sod_do_this") 
    >> DummyOperator(task_id="sod_last")

(
  BranchPythonOperator(
    task_id="should_run_sod",
    python_callable=lambda: "sod", # no_sod
  )
  >> [sod, DummyOperator(task_id="no_sod")]
  >> DummyOperator(task_id="end")
)

There are two ways you could achieve this, 1) using the chain() method or 2), using a TaskGroup. The latter is cleaner.

Approach 1

from airflow.models.baseoperator import chain

with DAG(...):
    should_run_sod = BranchPythonOperator(
        task_id="should_run_sod",
        python_callable=lambda: "sod",  # no_sod
    )

    sod = [
        DummyOperator(task_id="sod"),
        DummyOperator(task_id="sod_do_this"),
        DummyOperator(task_id="sod_last"),
    ]

    should_run_sod >> [sod[0], DummyOperator(task_id="no_sod")]
    chain(*sod, DummyOperator(task_id="end"))

Approach 2

with DAG(...):
    should_run_sod = BranchPythonOperator(
        task_id="should_run_sod",
        python_callable=lambda: "sod",  # no_sod
    )

    with TaskGroup(group_id="sod") as sod_group:
        sod = EmptyOperator(task_id="sod")
        sod_do_this = EmptyOperator(task_id="sod_do_this")
        sod_last = EmptyOperator(task_id="sod_last")

        sod >> sod_do_this >> sod_last

    should_run_sod >> [sod_group, EmptyOperator(task_id="no_sod")]
    sod_group >> EmptyOperator(task_id="end")