Generated CWL reference¶

These pages are generated with Transpiler-Mate cwl2markdown. Output directories follow the source path so workflows with the same ID do not overwrite one another.

CWL source	Workflow reference
`cwl-workflow/app-water-body.cwl`	Staged acquisition
`cwl-workflow/app-water-body-cloud-native.cwl`	Single cloud-native acquisition
`cwl-workflow/app-water-bodies-cloud-native.cwl`	Multiple cloud-native acquisitions, nested detection workflow

Regenerate¶

Run from the repository root using Python 3.12 or later:

python -m venv /tmp/mastering-cwl-docs
/tmp/mastering-cwl-docs/bin/python -m pip install -r scripts/requirements-cwl-docs.txt
/tmp/mastering-cwl-docs/bin/python scripts/generate-cwl-docs.py

The script invokes the plugin on all 12 tracked .cwl files. The runtime parses each document and validates its Schema.org metadata. The nine standalone CommandLineTool documents have no workflows, which the plugin reports as an unsupported input; the script reports these as skips. Command-line tools within the workflow graphs appear in the generated workflow pages.

The script supplies a compatibility fallback for cwl2markdown 0.2.1, which asks for index.md but packages index.md.jinja, and renames rendered .md.jinja output to .md.

Metadata and output limitations¶

The creation date and illustrative Jane/John Doe author records come from the training repository's codemeta.json. They are example metadata, not verified authorship. The license follows LICENSE.md and the README (CC-BY-SA-4.0); CodeMeta currently declares a different license. Existing workflow versions are preserved; standalone tools use CodeMeta's application version.

The generated pages retain upstream template output, including generation timestamps, links for optional fields rendered as None, and references to UML and OGC API schema images. The corresponding images are generated separately with cwl2puml and rendered locally with PlantUML.

EOAP validation¶

Validate both water-bodies and detect_water_body in the multi-workflow document explicitly using their fragments. Package, metadata, and staging profiles use these reviewed applicability configurations:

Entrypoint	Staging configuration
`app-water-body.cwl`	`validation/staging-staged.json`
`app-water-body-cloud-native.cwl`	`validation/staging-cloud-native.json`
`app-water-bodies-cloud-native.cwl#water-bodies`	`validation/staging-cloud-native.json`
`app-water-bodies-cloud-native.cwl#detect_water_body`	`validation/staging-detection.json`

The staged workflow accepts acquisition directories at its entrypoint and in the crop and catalog tools. Cloud-native inputs are remote STAC references; their internal raster files are intermediate products, not staged acquisition directories. The three top-level packages expose a Directory-based STAC catalog produced by the catalog tool. The nested detection workflow returns an intermediate raster file and declares no package staging boundary.

The validator always flags EOAP.REQ14.COVERAGE for declared staged outputs: execution evidence is needed to prove that all produced EO files are collected. These review findings remain even when Directory type checks pass.

Standalone command-line tools have no package Workflow entrypoint. Use cwltool --validate <tool.cwl> for their CWL validity; their metadata is also checked by the EOAP validator, but package and staging checks are blocked by its workflow-selection requirement.

Container declarations now use mandatory DockerRequirement entries rather than advisory hints, so runners must honor the specified images.

Typed EOAP inputs¶

The CWL documents selectively adopt the semantic typing from feature/enhancements: imported GeoJSON Polygon and string-format URI records, an EPSG enum (4326), and spectral-band enums (green, nir, nir08). Staged acquisition inputs retain their Directory type.

Use cwl-workflow/typed-cloud-native-inputs.yaml as an input-shape example:

cwltool --podman cwl-workflow/app-water-body-cloud-native.cwl \
  cwl-workflow/typed-cloud-native-inputs.yaml

The example retains the tutorial's STAC reference; its availability is not verified by static validation. For the multi-acquisition workflow, supply stac_items as a list of records with a value field instead of item.

The imported Polygon schema requires both coordinates and bbox. The crop binding serializes the full Polygon as JSON; pixels outside the polygon are masked. URI records are unwrapped through self.value. These changes require updating jobs that previously supplied plain AOI and STAC strings. The imports reference the schemas repository's main branch.

Workflow enum definitions are inline in each CWL document under SchemaDefRequirement, keeping the Application Package self-contained. Named enum references allow cwltool to bind array inputs and preserve identical types across workflow/tool boundaries.