Celery chain, group and chord: when to use each

# ======================================================================
# Q: Celery: chain, group, chord. When do you use them?
#
# A chord needs a result backend. I keep task results small (IDs, not large
# objects). For large workflows I prefer state in the database, so the workflow
# can continue after a failure.
# ======================================================================

from celery import chain, group, chord

chain(fetch_orders.s(day), build_report.s(), email_report.s(to))()   # sequential, passes the result along
group(sync_supplier.s(s.id) for s in suppliers)()                    # parallel
chord((price_for_region.s(r) for r in regions), merge_prices.s())()   # parallel, then a callback