Workflow Design and Logic
What to know
Workflows are DAGs of tools connected by data streams. Understand Filter (splits into True/False outputs), Formula with conditional logic, and control-of-flow tools. Know that tools execute when data reaches them, and that branches run independently. Design for readability: name tools, annotate, and avoid deep nesting of formulas where a Filter plus Join is clearer.
Traps
- Assuming a specific execution order across parallel branches; only data dependencies enforce order.
- Filter "False" output being empty when you expected rows (logic inverted).
- Formula referencing a field that does not exist yet downstream.
- Mixing data types in a conditional expression, causing coercion surprises.
- Forgetting that changing an upstream schema breaks downstream tools silently.
Rule of thumb
Model logic as explicit branches: Filter for binary paths, Formula for derived flags, and Join to recombine. Test each branch's row counts sum to the input count.
Worked example: A Filter on amount > 100 sends large orders to True and small to False. If downstream totals look wrong, verify the True and False row counts add up to the input; a null amount typically falls to False, which may be unintended.