Easy creation of reproducible computational workflows with SciWIn-Client

Computational workflows are used in many different disciplines to automate complex, multi-step data analysis procedures. To ensure reproducibility, scalability and efficiency in scientific research, lots of effort has been put into domain specific languages that formalize and standardize computational scientific work. The Common Workflow Language (CWL)[1] is among the most powerful and established workflow languages. While CWL is an extremely flexible language, is also quite verbose. To facilitate the creation of CWL files, SciWIn-Client, which is written in the Rust programming language, offers a simple command line interface. SciWIn-Client is the first building block of the Scientific Workflow Infrastructure which is a key part of the FAIRagro consortium's service portfolio. CWL workflows are structured as directed acyclic graphs where individual steps are connected through their inputs and outputs. A CWL CommandLineTool represents a single executable step in a workflow. A CWL Workflow acts as a container for multiple steps, connects multiple tools and manages dependencies between steps. SciWIn-Client creates CWL CommandLineTools by capturing and analyzing the CLI command-line used by the researcher in their usual routine to carry out computational tasks. Steps represented by CWL CommandLineTools can be connected into complex workflows by using simple commands to build the directed acyclic graph edge by edge. Designed as an unobtrusive lean CLI program, SciWIn-Client integrates seamlessly into the researcher's toolbox providing for provenance management what Git delivers for version control. SciWIn-Client thus helps to implement FAIR data- and code handling practices right at the digital workbench. Next to the reproducibility and traceability of computational processes, scientists gain the ability package their intermediate and final results into FAIR Digital Objects (FDOs). We develop multiple export options, including Workflow RO-Crates[2] that can be published on workflowhub.eu, ARCs[3] that are compatible with DataPLANT's PLANTdataHUB[3], and Provenance Run Crate[4] as the most advanced FDO type. Furthermore, the CWL workflows created by SciWIn-Client can be executed by remote execution engines such as Reana[5] and thus provide easy access to high performance compute infrastructure.