Using language to define workflows: How do we specify inputs to run scientific workflows?

by Dr Alexander Goscinski (PSI/LMS)

Wednesday 12 Feb 2025, 15:00 → 16:00 Europe/Zurich

OVGA/200

In this talk, I will explore common methods for specifying inputs when defining a workflow for a scientific experiment. To illustrate these approaches, we will examine three packages developed in the Laboratory for Materials Simulations (LMS) — koopmans, Sirocco, and AiiDA-WorkGraph — and analyze how they define their workflow inputs. We will show that koopmans and Sirocco adopt a data-model approach which can leverage strongly typed schemas for user input validation, enhancing robustness and enabling user-friendly error handling. However, we will also discuss the limitations of this approach, particularly when applied to writing generic program logic, using Common Workflow Language (CWL) as an example. To address these challenges, we propose an alternative source-code approach, which proves more effective for certain workflow definitions. We will introduce theoretical concepts from the formal language theory required for this approach and demonstrate its benefits through a prototype implementation based on AiiDA-WorkGraph.