cwl-runner

TLDR

Run a CWL workflow

$ cwl-runner [workflow.cwl] [inputs.yml]

Run with specific input values

$ cwl-runner [workflow.cwl] --[input_name] [value]

Run with Docker disabled

$ cwl-runner --no-container [workflow.cwl] [inputs.yml]

Run with custom output directory

$ cwl-runner --outdir [/path/to/output] [workflow.cwl] [inputs.yml]

Validate a CWL document

$ cwl-runner --validate [workflow.cwl]

Run in parallel

$ cwl-runner --parallel [workflow.cwl] [inputs.yml]

SYNOPSIS

cwl-runner [options] workflow.cwl [inputs]

cwl-runner executes workflows defined in Common Workflow Language (CWL), a specification for describing analysis workflows in data-intensive sciences. CWL workflows are portable across different execution environments.
The runner takes a CWL document and input parameters, resolves dependencies, and executes tools in the correct order. Steps can be run in containers (Docker/Singularity) for reproducibility, or natively if tools are installed locally.
Input parameters can be specified in a YAML/JSON file or as command-line arguments. Outputs are placed in the specified output directory with a manifest describing the results.

PARAMETERS

--validate

Validate the CWL document without running.

--outdir path

Output directory for results.

--no-container

Disable Docker/container execution.

--parallel

Run workflow steps in parallel.

--cachedir path

Directory for caching intermediate results.

--tmpdir-prefix path

Prefix for temporary directories.

--debug

Enable debug output.

--quiet

Only print warnings and errors.

--strict

Strict validation mode.

--leave-tmpdir

Don't delete temporary directories.

INPUT FILE EXAMPLE

$ # inputs.yml
fastq_file:
  class: File
  path: /data/sample.fastq
reference:
  class: File
  path: /data/reference.fa
threads: 8

CAVEATS

Default implementation (cwltool) may have performance limitations for large workflows. Some advanced features require specific runners (toil-cwl-runner, arvados-cwl-runner). Container requirements need Docker or Singularity installed. File paths in inputs must be accessible to the runner.

HISTORY

Common Workflow Language was developed by a community effort starting in 2014, with the specification reaching v1.0 in 2016. Led by Peter Amstutz and Michael Crusoe, CWL emerged from the need for portable, reproducible workflows in bioinformatics. cwltool is the reference implementation. CWL has been adopted by many scientific platforms and is now a standard for sharing computational pipelines.