python-cwlgen is now deprecated, please use
cwl-utils >= 0.4.
from cwl_utils import parser_v1_0 # You could alias this as cwlgen to simplify the migration from cwl_utils import parser_v1_0 as cwlgen
get_dict() → save()
parse_dict→ No super clear analogue, but loaded through
_UnionLoader((CommandLineToolLoader, ...workflow + other loaders)
- No more special field names, eg:
input_id| etc →
Classes aren't nested anymore, ie:
Take care if you're migrating to a newer spec, as some classes might have changed names (notably:
Don't forget to catch all references of cwlgen, as missing one (or using mismatch versions of the parser) will cause:
raise RepresenterError('cannot represent an object: %s' % (data,)) ruamel.yaml.representer.RepresenterError: cannot represent an object: <cwlgen.common.CommandInputArraySchema object at 0x1100a5780>
If you have issues with the migration, please see this thread or raise an issue on CWLUtils.
Python-cwlgen is a python library for the generation of CWL programmatically. It supports the generation of CommandLineTool, Workflow and DockerRequirement. The library works for both Python 2.7.12+ and 3.6.0.
Common Workflow Language
- Stop writing bash scripts for long complex jobs.
- Take pipelines anywhere (portability).
- Enforce reproducibility guidelines.
This python repository is a python wrapper for most of the classes (work in progress), allowing you to build the structure of the workflow in Python and have this module generate and export CWL for you.
Nb: This doesn't check the logic of Workflows or CommandLineTools for you.
CWLTool has a
--validate mode that you can use.
You can install python-cwlgen through pip with the following command:
pip install cwlgen
How it works?
This repository contains a number of python classes that mirror the CWL specifications (Workflow| CommandLineTool). In essence, each class's initializer has all of the properties it expects, which may be another object. The classes include the relevant docstrings to give you context of classes and their properties.
examples/ folder contains some simple examples, however in essence you simply initialize the class you're
trying to build. An initializer for a class has all of the properties it expects which may be another object.
Creating a CommandLineTool
import cwlgen tool_object = cwlgen.CommandLineTool(tool_id="echo-tool", base_command="echo", label=None, doc=None, cwl_version="v1.0", stdin=None, stderr=None, stdout=None, path=None) tool_object.inputs.append( cwlgen.CommandInputParameter("myParamId", param_type="string", label=None, secondary_files=None, param_format=None, streamable=None, doc=None, input_binding=None, default=None) ) # to get the dictionary representation: dict_to_export = tool_object.get_dict() # to get the string representation (YAML) yaml_export = tool_object.export_string() # print to console tool_object.export() # print to file tool_object.export("echotool.cwl")
CWL is developed by an informal, multi-vendor working group consisting of organizations and individuals aiming to enable scientists to share data analysis workflows. The CWL project is on Github.
SchemaDefRequirementdoesn't parse the
typessubfield into the specific types (
InputRecordSchema | InputEnumSchema | InputArraySchema), but leaves them as a simple dictionary.