Hi all! I've been having issues with op_selection for nested graphs. Should all_node_events be empty for the following example?

from dagster import job, op, graph


@op
def return_fifty():
    return 50.0


@op
def add_thirty_two(number):
    return number + 32.0


@op
def multiply_by_one_point_eight(number):
    return number * 1.8


@op
def log_number(context, number):
    <http://context.log.info|context.log.info>(f"number: {number}")


@graph
def celsius_to_fahrenheit(number):
    return add_thirty_two(multiply_by_one_point_eight(number))


@job
def all_together_nested():
    log_number(celsius_to_fahrenheit(return_fifty()))


x = my_job.execute_in_process(op_selection=["*celsius_to_fahrenheit.add_thirty_two"])

x.all_node_events

Hi @Harry James

op_selection

in the

execute_in_process

api doesn’t support the * or + selection syntax yet, which is available in Dagit. Here’s a tracking issue for this feature parity w/ dagit. As a stopgap, you can do:

x = all_together_nested.execute_in_process(
    op_selection=[
        "celsius_to_fahrenheit.add_thirty_two",
        "celsius_to_fahrenheit.multiply_by_one_point_eight",
    ],
    run_config={"ops": {"celsius_to_fahrenheit": {"inputs": {"number": {"value": 1}}}}}, // fill out missing inputs via run_config
)

Created issue at: https://github.com/dagster-io/dagster/issues/7828

@yuhan thank you for the reply and for creating the issue! The * selection syntax does however work for the non-nested example found here https://docs.dagster.io/concepts/ops-jobs-graphs/job-execution#specifying-op-selection

from dagster import job, op


@op
def return_one():
    return 1


@op
def add_two(i: int):
    return i + 2


@op
def multi_three(i: int):
    return i * 3


@job
def my_job():
    multi_three(add_two(return_one()))
    
x = my_job.execute_in_process(op_selection=["*add_two"])

x.all_node_events