Workflow Description Language static analysis toolkit for Python 3.6+
miniwdl is a library for parsing WDL documents into a type-checked abstract syntax tree (AST), providing a foundation for new runtime systems, developer tooling, and language experimentation. It also includes command-line tools supporting the WDL development cycle, including a "linter" to statically analyze WDL documents for errors and oversights, and a Cromwell wrapper to make it more convenient to test a workflow locally.
This project in alpha development; interfaces are liable to change somewhat. See the Releases for change logs. The Project board reflects the near-term roadmap.
pip: pip3 install miniwdl
conda: configure conda-forge and conda install miniwdl
miniwdl check /path/to/workflow.wdl
loads the WDL document and shows a brief outline with any lint warnings. Add --path /path/to/tasks/
with a directory to search for imported documents (one or more times). Example with HumanCellAtlas/skylab:
$ git clone https://github.com/HumanCellAtlas/skylab.git
$ miniwdl check --path skylab/library/tasks/ \
skylab/pipelines/smartseq2_single_sample/SmartSeq2SingleSample.wdl
SmartSeq2SingleSample.wdl
workflow SmartSeq2SingleCell
(Ln 14, Col 8) UnusedDeclaration, nothing references File gtf_file
call HISAT2.HISAT2PairedEnd
call Picard.CollectMultipleMetrics
call Picard.CollectRnaMetrics
call Picard.CollectDuplicationMetrics
call HISAT2.HISAT2RSEM as HISAT2Transcriptome
call RSEM.RSEMExpression
call GroupQCs.GroupQCOutputs
call ZarrUtils.SmartSeq2ZarrConversion
GroupQCs : GroupMetricsOutputs.wdl
task GroupQCOutputs
(Ln 10, Col 10) StringCoercion, String mem = :Int:
(Ln 11, Col 10) StringCoercion, String cpu = :Int:
(Ln 12, Col 10) StringCoercion, String disk_space = :Int:
HISAT2 : HISAT2.wdl
task HISAT2PairedEnd
task HISAT2RSEM
task HISAT2InspectIndex (not called)
task HISAT2SingleEnd (not called)
Picard : Picard.wdl
task CollectDuplicationMetrics
task CollectMultipleMetrics
task CollectRnaMetrics
RSEM : RSEM.wdl
task RSEMExpression
ZarrUtils : ZarrUtils.wdl
task SmartSeq2ZarrConversion
(Ln 36, Col 6) CommandShellCheck, SC2006 Use $(..) instead of legacy `..`.
(Ln 39, Col 9) CommandShellCheck, SC2006 Use $(..) instead of legacy `..`.
(Ln 39, Col 15) CommandShellCheck, SC2086 Double quote to prevent globbing and word splitting.
(Ln 40, Col 10) CommandShellCheck, SC2086 Double quote to prevent globbing and word splitting.
(Ln 40, Col 21) CommandShellCheck, SC2086 Double quote to prevent globbing and word splitting.
In addition to its suite of WDL-specific warnings, miniwdl check
uses ShellCheck, if available, to detect possible issues in each task command script. You may need to install ShellCheck separately, as it's not included with miniwdl.
If you haven't installed the PyPI package to get the miniwdl
entry point, equivalently PYTHONPATH=$PYTHONPATH:/path/to/miniwdl python3 -m WDL check ...
.
This tool provides a nicer command-line interface for running a workflow locally using Cromwell. Example:
$ cat << 'EOF' > hello.wdl
version 1.0
task hello {
input {
Array[String]+ who
Int x = 0
}
command <<<
awk '{print "Hello", $0}' "~{write_lines(who)}"
>>>
output {
Array[String]+ messages = read_lines(stdout())
Int meaning_of_life = x+1
}
}
EOF
$ miniwdl cromwell hello.wdl
missing required inputs for hello: who
required inputs:
Array[String]+ who
optional inputs:
Int x
outputs:
Array[String]+ messages
Int meaning_of_life
$ miniwdl cromwell hello.wdl who=Alyssa "who=Ben Bitdiddle" x=41
{
"outputs": {
"hello.messages": [
"Hello Alyssa",
"Hello Ben Bitdiddle"
],
"hello.meaning_of_life": 42
},
"id": "b75f3449-344f-45ec-86b2-c004a3adc289",
"dir": "/home/user/20190203_215657_hello"
}
By first analyzing the WDL code, this tool translates the freeform command-line arguments into appropriately-typed JSON inputs for Cromwell. It downloads the Cromwell JAR file automatically to a temporary location; a compatible java
JRE must be available to run it. The outputs and logs are written to a new date/time-named subdirectory of the current working directory (overridable; see --help
).
The tool supports shell tab-completion for the workflow's available input names. To use this, enable argcomplete global completion by invoking activate-global-python-argcomplete
and starting a new shell session. Then, start a command line miniwdl cromwell hello.wdl
and try double-tab.
The WDL
package provides programmatic access to the WDL parser and AST. The following example prints all declarations in a workflow, descending into scatter
and if
stanzas as needed.
$ python3 -c "
import WDL
doc = WDL.load('skylab/pipelines/optimus/Optimus.wdl',
path=['skylab/library/tasks/'])
def show(elements):
for elt in elements:
if isinstance(elt, WDL.Decl):
print(str(elt.type) + ' ' + elt.name)
elif isinstance(elt, WDL.Scatter) or isinstance(elt, WDL.Conditional):
show(elt.elements)
show(doc.workflow.elements)
"
String version
Array[File] r1_fastq
Array[File] r2_fastq
Array[File] i1_fastq
String sample_id
File tar_star_reference
File annotations_gtf
File ref_genome_fasta
File whitelist
String fastq_suffix
Array[Int] indices
Array[File] non_optional_i1_fastq
File barcoded_bam
Online Python developer documentation for the WDL
package:
(Read the Docs currently builds from the mlin/miniwdl fork of this repository.)
Locally, make doc
triggers Sphinx to generate the docs under docs/_build/html/
. Or, after building the docker image, copy them out with docker run --rm -v ~/Desktop:/io miniwdl cp -r /miniwdl/docs/_build/html /io/miniwdl_docs
.
Feedback and contributions are welcome on this repository. Please:
- Add appropriate tests to the automatic suite
- Use
make pretty
to reformat the code with black - Ensure compatibility with this project's MIT license
- Send pull requests from a dedicated branch without unrelated edits
The Project board is our up-to-date tracker.
To set up your local development environment,
git clone --recursive
this repository- Install dependencies as illustrated in the Dockerfile (OS packages + PyPI packages listed in
requirements.txt
andrequirements.dev.txt
) - Invoking user must be able to access local Docker daemon.
The Makefile has a few typical scripted flows:
make pretty
reformats the code with blackmake check
validates the code with Pylint and Pyremake
ormake test
runs the full test suite with code coverage report (takes several minutes)make qtest
runs most of the tests more quickly (by omitting some slower cases, and not tracking coverage)
Please disclose security issues responsibly by contacting [email protected].