Actual Dag

An actual DAG

Next we will build a slightly more topologically complex DAG that demonstrates how dagster determines the execution order of solids in a pipeline:

.. image:: actual_dag_figure_one.png

.. literalinclude:: ../../../../dagster/tutorials/intro_tutorial/actual_dag.py :linenos: :caption: actual_dag.py

Again, it is worth noting how we are connecting inputs and outputs rather than just tasks. Point your attention to the solid_d entry in the dependencies dictionary: we declare dependencies on a per-input basis.

When you execute this example, you'll see that solid_a executes first, then solid_b and solid_c -- in any order -- and solid_d executes last, after solid_b and solid_c have both executed.

In more sophisticated execution environments, solid_b and solid_c could execute not just in any order, but at the same time, since their inputs don't depend on each other's outputs -- but both would still have to execute after solid_a (because they depend on its output to satisfy their inputs) and before solid_d (because their outputs in turn are depended on by the input of solid_d).

Try it in dagit or from the command line:

.. code-block:: console

$ dagster pipeline execute -f actual_dag.py -n define_diamond_dag_pipeline

What's the output of this DAG?

We've seen how to wire solids together into DAGs. Now let's look more deeply at their :doc:Inputs <inputs>, and start to explore how solids can interact with their external environment.