CWL 1.0 improvements over sbg:draft-2

Estimated reading time: 5 minutes

Overview

Learn about the differences of CWL v1.0 from CWL sbg:draft-2.

No more $job for BiX8

$job does not exist in JS expressions. Available contexts are $inputs and $runtime.

sbg:draft-2 CWL v1.0

$job.inputs.input_bam.path

$(inputs.input_bam.path)

$job.allocatedResources.cpu

$(runtime.cores)

Entering a JS expression

Expressions are denoted by the syntax $(...) or ${...}. A code fragment wrapped in the $(...) syntax is used as a one-line expression. A code fragment wrapped in ${...} behaves like expressions in sbg:draft-2.

sbg:draft-2 CWL v1.0

$job.inputs.input_bam.path + ‘.vcf’

$(inputs.input_bam.path).vcf

{ return $job.inputs.input_bam.path + ‘.vcf’ }

${ return inputs.input_bam.path + ‘.vcf’}

Commands relating to file paths (basename, dirname, nameroot, nameext)

Experience improved commands relating to file paths. For instance, use .nameroot to get the input basename. Learn more from CWL’s documentation.

Expression Return

$inputs.input_bam.path

/path/to/file.sorted.bam

$inputs.input_bam.basename

file.sorted.bam

$inputs.input_bam.dirname

/path/to

$inputs.input_bam.nameroot

file.sorted

$inputs.input_bam.nameext

bam

One-line expressions

If the input is a BAM and the output is a VCF, you can define the output name as $(inputs.input_bam.nameroot).vcf.

No expressions in the base command

All the expressions must be inserted via arguments.

This means that if you want to insert pre-commands (such as un-TAR reference files), you need to define the whole command through arguments and leave the base command empty.

Expressions in secondary files

Secondary files can now be defined with an expression.

For example, if the input is VCF or VCF.GZ, the secondary file is either .idx or .tbi.: $(self.nameext == 'gz' ? "tbi" : "idx").

InitialWorkDirRequirement

You can create a file in workdir on runtime or make a file available in the workdir on runtime using InitialWorkDirRequirement.

Create literal content file example


requirements:

  - class: InitialWorkDirRequirement
    listing:
    - entry: $(JSON.stringify(inputs))
      entryname: cwl.inputs.json

Create expression content file example:


requirements:

  - class: InitialWorkDirRequirement
    listing:
    - entry: |-

         Some contents

      entryname: cwl.inputs.json

Stage an input example:


requirements:

 - class: InitialWorkDirRequirement
   listing:
   - "$(inputs.bam)"
   - "$(inputs.fastq_list)"

Example with staging inputs and creating files


requirements:

 - class: InitialWorkDirRequirement
   listing:
   - entry: $(JSON.stringify(inputs))
     entryname: cwl.inputs.json
   - entry: |-
     ${
        return JSON.stringify(inputs)
     }
     entryname: '${return "_1_cwl.inputs.json"}'
   - "$(inputs.bam)"
   - "$(inputs.fastq_list)"

ExpressionLibRequirement

Use ExpressionLibRequirement to write a JS function in one place and use it in multiple places.

Define a function as:


requirements:
 - class: InlineJavascriptRequirement

   expressionLib:
     - var ext = function(){
       var x = inputs.vcf.nameext == 'gz' ? "tbi" : "idx";
       return x
     };

And call it in a different place:


secondaryFiles:
 ${

    return ext()

  }

SchemaDefRequirement

Custom input and output structures can be defined in SchemaDefRequirement.


class: SchemaDefRequirement
types:
- name: FileRecord
  type: record
  fields:
    - name: file
      type: File
    - name: metadata
      type: map   
      values: string

shellQuote

Keep this as false.

EnvVarRequirement

Instead of starting a command with export MY_CUSTOM_VARIABLE=DzoniJovanovic, use EnvVarRequirement.


requirements:
 EnvVarRequirement:
   envDef:
     FILIP: $(inputs.last_name)

ExpressionTool

Similar to command-line tool except it has no command line and does not start a Docker container. The only purpose of this is to reshape stuff.

Example:

https://github.com/NCI-GDC/gdc-dnaseq-cwl/blob/master/tools/decider_bwa_expression.cwl

https://github.com/NCI-GDC/gdc-dnaseq-cwl/blob/master/tools/sort_scatter_expression.cwl

Input/Output Type Directory

The input can be type directory.

Input/Output Union Type

An input or output can be defined to be one or more possible types.

Example: Intervals input string or BED file

Input/Output Format Ontology


inputs:
 aligned_sequences:
   type: File
   label: Aligned sequences in BAM format
   format: edam:format_2572

$namespaces:
 edam: http://edamontology.org/
$schemas:
 - http://edamontology.org/EDAM_1.18.owl

SoftwareRequirement


hints:
 SoftwareRequirement:
   packages:
     interproscan:
       specs: [ "https://identifiers.org/rrid/RRID:SCR_005829" ]
       version: [ "5.21-60" ]