How to write a workflow file

A workflow file is a YAML or JSON file with a list of steps. See Workflows for all keys of a workflow file.

Start from an example

Copy an example from the examples folder of the M3Resp source code, and change it:

Example	Use it for
examples/multimodal_example/multimodal.workflow.yaml	A short EIT, EMG and ventilator workflow
examples/eit_full_preprocessing/eit-full.workflow.yaml	All EIT steps
examples/emg_full_preprocessing/emg-full.workflow.yaml	All EMG and ventilator steps
examples/multimodal_full/multimodal-full.workflow.yaml	All steps of the three modalities

Give the input files

Put the file paths under inputs, and refer to them with @ in the steps:

inputs:
  eit_file: ../data/subject01.bin

steps:
  - uses: eit.load
    with: { file_path: "@eit_file", vendor: draeger }

M3Resp reads a relative path from the folder of the workflow file.

Add a step

To list the steps and their settings:

m3resp steps
m3resp describe eit.detect_breaths

Add the step with uses, and its settings with with. If the step must read another result than its default input, set it with in:

  - uses: eit.detect_breaths
    in: { signal: global_impedance }
    with: { min_duration_s: 0.6666666667 }

Run the same step twice

Each step writes its results under fixed names. To keep the results of both runs, rename the results of one run with out:

  - uses: eit.global_impedance
    in: { signal: raw_eit }
    out: { global_impedance: unfiltered_global_impedance }

  - uses: eit.global_impedance
    in: { signal: filtered_eit }

Choose the result files

Set the folder and the files in outputs:

outputs:
  dir: ../../output/my-analysis
  mode: automatic
  timestamped: true
  structured_export: true

With timestamped: true, each run writes to its own subfolder with the date and time. See The outputs section for all settings.

Use strict checks

Set schema_version: 1 at the top of the file. M3Resp then refuses unknown keys and settings that are not real booleans, so a typing error stops the run. Give each step an id, so other tools can find the step also after you add or move steps:

schema_version: 1

steps:
  - id: detect_breaths
    uses: eit.detect_breaths

Then check the file before you run it: How to check a workflow file.