doc2wdl

Generate a WDL task wrapper from a tool's help text, using docopt for parsing.

See:

• https://github.com/openwdl/wdl
• https://github.com/broadinstitute/gatk/wiki/How-to-Prepare-a-GATK-tool-for-WDL-Auto-Generation
• https://support.terra.bio/hc/en-us/articles/360037120252
• http://docopt.org/

Usage

# Capture the help text from a tool
mycommand --help > mycommand-help.txt
# Generate a draft WDL task
doc2wrapper docopt -f mycommand-help.txt -o mycommand.task.wdl
# In one shot
cat example-help.txt | doc2wrapper docopt > example.task.wdl

# Python-specific introspection
doc2wrapper argparse -m cnvlib.commands -p AP -o example.wdl

The generated WDL will very likely need further editing to work well, but hopefully this approach will save you some time over writing each task from scratch.

Conceptual flow

• Readers (docopt, argparse) parse the given doc, transform contents, and populate the object model in memory.
• Writers (WDL, Nextflow) take the populated object model and use the static jinja2 template to generate a string document.

Known issues

• Option help text doesn't get transferred to param_meta automatically (yet).
• Type inference is not perfect.
• Arrayed types and other non-primitive types won't be detected, e.g. options that can be repeated on the command line.
• Only one output file from the command will be picked up in the task output.
• Quirks in the original command's help text may slip through; docopt doesn't attempt to catch everything (it really wasn't designed for this purpose).