Renderers

Renderers are a function that takes a task, which can be assumed to have a __workflow__ member of type Workflow, and return a YAML-serializable nested dict structure.

CWL

The default renderer is for the Common Workflow Language. It implements a very small subset of functionality, and is not yet strictly standards compliant. It assumes that all run names can be interpreted in the context of the workflow module's global scope.

An additional arg can be passed to the CWL renderer to allow the steps to be ordered to match the execution within the python script

workflow = construct(example_workflow(), simplify_ids=True)

rendered = render(workflow, sort_steps=True)

If the tasks within the script are listed as:

@workflow()
def example_workflow() -> int:
    """Test workflow with multiple tasks"""
    step1 = increment(num=1)
    step2 = increment(num=5)
    step3 = sum(left=step1, right=step2)
    return step3

Then the yaml will be returned as:

{
    "cwlVersion": 1.2,
    "class": "Workflow",
    "inputs": {},
    "outputs":
        { "out": { "label": "out", "type": "int", "outputSource": "sum-1-1/out" } },
    "steps":
        {
        "increment-1-2":
            {
            "run": "increment",
            "in": { "num": { "default": 1 } },
            "out": ["out"],
            },
        "increment-1-1":
            {
            "run": "increment",
            "in": { "num": { "default": 5 } },
            "out": ["out"],
            },
        "sum-1-1":
            {
            "run": "sum",
            "in":
                {
                "left": { "source": "increment-1-2/out" },
                "right": { "source": "increment-1-1/out" },
                },
            "out": ["out"],
            },
        },
}

The sort-steps can also be used in the CLI passing it as a flag:

python -m dewret workflows.py example_workflow --sort-steps

Custom

...