Pull the all referenced software container images

cwl-docker-extract is useful to cache or pre-pull all software container images referenced in a CWL CommandLineTool or CWL Workflow (including all referenced CommandLineTools and sub-Workflows and so on).

The default behaviour is to use the Docker engine to download and save the software container images in Docker format.

cwl-docker-extract path_to_my_workflow.cwl

cwl-docker-extract --dir DIRECTORY path_to_my_workflow.cwl

Or you can use the Singularity software container engine to download and save the software container images and convert them to the Singularity format at the same time.

cwl-docker-extract --singularity --dir DIRECTORY path_to_my_workflow.cwl

Print all referenced software packages

cwl-cite-extract prints all software packages found (recursively) in the specified CWL document.

Currently the package name and any listed specs and version field are printed for all SoftwareRequirement s found.

cwl-cite-extract path_to_my_workflow.cwl

Replace CWL Expressions with concrete steps

cwl-expression-refactor refactors CWL documents so that any CWL Expression evaluations are separate steps (either CWL ExpressionTools or CWL CommandLineTools.) This allows execution by CWL engines that do not want to support inline expression evaluation outside of concrete steps, or do not want to directly support CWL’s optional InlineJavascriptRequirement at all.

cwl-expression-refactor directory/path/to/save/outputs path_to_my_workflow.cwl [more_workflows.cwl]

Split a packed CWL document

cwl-graph-split splits a packed CWL document file into multiple files.

Packed CWL documents use the $graph construct to contain multiple CWL Process objects (Workflow, CommandLineTool, ExpressionTool, Operation). Typically packed CWL documents contain a CWL Workflow under the name “main” and the workflow steps (including any sub-workflows).

cwl-graph-split --outdir optional/directory/path/to/save/outputs path_to_my_workflow.cwl

Pack a CWL document

cwl-pack packs a CWL document and all of its referenced documents into a single JSON document. Unlike cwl-normalizer, cwl-pack does not upgrade the CWL version and does not refactor expressions; it is a thin CLI wrapper around cwl_utils.pack.pack().

cwl-pack path_to_my_workflow.cwl
cwl-pack --outfile packed.json path_to_my_workflow.cwl

Normalize a CWL document

cwl-normalizer normalizes one or more CWL document so that for each document, a JSON format CWL document is produces with it and all of its dependencies packed together, upgrading to CWL v1.2, as needed. Can optionally refactor CWL Expressions into separate steps in the manner of cwl-expression-refactor.

cwl-normalizer directory/path/to/save/outputs path_to_my_workflow.cwl [more_workflows.cwl]

Generate for Workflow Parameters from a CWL document

cwl-inputs-schema-gen generates a JSON Schema for workflow input parameters from a CWL document.

cwl-inputs-schema-gen path_to_my_workflow.cwl

Using the CWL Parsers

from pathlib import Path
from ruamel import yaml
import sys

from cwl_utils.parser import load_document_by_uri, save

# File Input - This is the only thing you will need to adjust or take in as an input to your function:
cwl_file = Path("testdata/md5sum.cwl")  # or a plain string works as well

# Import CWL Object
cwl_obj = load_document_by_uri(cwl_file)

# View CWL Object
print("List of object attributes:\n{}".format("\n".join(map(str, dir(cwl_obj)))))

# Export CWL Object into a built-in typed object
saved_obj = save(cwl_obj)
print(f"Export of the loaded CWL object: {saved_obj}.")

Choosing a JavaScript engine

CWL expressions are evaluated with QuickJS when a usable qjs executable is found on the PATH, and with Node.js otherwise (falling back to a node:slim software container if Node.js is not installed either). QuickJS starts quickly enough that every evaluation runs in its own short-lived subprocess; there is no software container fallback for QuickJS. Setting the CWL_JS_ENGINE environment variable to node or quickjs forces a specific engine.

A specific engine can also be installed programmatically, which takes precedence over CWL_JS_ENGINE:

from cwl_utils.sandboxjs import QuickJSEngine, set_js_engine

set_js_engine(QuickJSEngine())

Regenerate parsers

To regenerate install the schema_salad package and run:

cwl_utils/parser/cwl_v1_0.py was created via schema-salad-tool --codegen python https://github.com/common-workflow-language/common-workflow-language/raw/codegen/v1.0/extensions.yml --codegen-parser-info "org.w3id.cwl.v1_0" > cwl_utils/parser/cwl_v1_0.py

cwl_utils/parser/cwl_v1_1.py was created via schema-salad-tool --codegen python https://github.com/common-workflow-language/cwl-v1.1/raw/codegen/extensions.yml --codegen-parser-info "org.w3id.cwl.v1_1" > cwl_utils/parser/cwl_v1_1.py

cwl_utils/parser/cwl_v1_2.py was created via schema-salad-tool --codegen python https://github.com/common-workflow-language/cwl-v1.2/raw/codegen/extensions.yml --codegen-parser-info "org.w3id.cwl.v1_2" > cwl_utils/parser/cwl_v1_2.py

Release

To release CWLUtils, bump the version in src/cwl_utils/__meta__.py, and tag that commit with the new version. The gh-action-pypi-publish should release that tag.