Pattern: Ordering-Only Dependencies

When you need one step to run after another but do not need the upstream step’s data. Use the _ prefix convention to signal ordering-only intent.

The right way

from barca import task

@task
def migrate_db():
    run_migrations()

@task(inputs={"_migrate": migrate_db})
def seed_data():
    insert_seed_records()

The _migrate parameter name starts with _, which tells barca to establish the DAG edge (ensuring migrate_db finishes before seed_data starts) and pass None to the function instead of the upstream value. The function body never references _migrate.

Why this works

• Intent is explicit. Anyone reading the code immediately sees that the dependency is for ordering, not data flow. The parameter exists in the signature to satisfy static analysis, but the _ prefix signals “I don’t use this value.”
• Value is not passed. The function receives None for _-prefixed parameters, making it clear the dependency is structural. The upstream artifact is still materialized and cached as normal – the _ prefix only affects what the downstream function sees.
• DAG is still correct. The edge is still present in the execution plan. Barca will still schedule seed_data in a later tier than migrate_db.

Common mistakes

Using a normal parameter name and ignoring it

# Works but unclear intent
@task(inputs={"migrate": migrate_db})
def seed_data(migrate):  # never used, but barca still passes the value
    insert_seed_records()

This is functionally correct but unclear to readers – the parameter looks like it carries data. Use the _ prefix to signal that the dependency is for ordering only and the value is not needed.

Trying to use after=

# Wrong -- after= was removed
@task(after=[migrate_db])
def seed_data():
    insert_seed_records()

Early prototypes of barca had an after= keyword for ordering-only edges. This was removed in favor of the _ prefix convention on inputs=, which keeps a single mechanism for all dependency types. If you see after= in old examples, replace it with inputs={"_name": upstream}.

Naming convention

The _ prefix was chosen deliberately to signal “ordering-only” to barca. This intentionally overlaps with Python’s convention for unused parameters — if barca won’t pass a value, you shouldn’t use the parameter anyway.

If you have a linter warning about unused _ parameters, add a # noqa: ARG001 comment or configure your linter to allow _-prefixed params.