CWL: How to return int or string[] from CommandLineTool?

Question

[uagrif]

Hello guys,

I'm experimenting with scatter/gather feature of Common Workflow Language and I have troubles with return type that differs from string.

So, returning strings works good:

class: CommandLineTool
cwlVersion: "v1.0"
baseCommand: [bash, -c, "echo 42 > data.txt"]

inputs: []
outputs:
  value:
    type: string
    outputBinding:
      glob: data.txt
      loadContents: true
      outputEval: $(self[0].contents)

I run cwl in docker. The output for string evaluatuion is:

root@c9b274ac17d4:~/cwl/complex-out# cwl-runner complex.cwl
/usr/local/bin/cwl-runner 1.0.20181012180214
Resolved 'complex.cwl' to 'file:///root/cwl/complex-out/complex.cwl'
[job complex.cwl] /tmp/tmp557p7kqz$ bash \
    -c \
    'echo 42 > data.txt'
[job complex.cwl] completed success
{
    "value": "42\n"
}
Final process status is success

However, when I change output value type from string to int, cwl-runner reports permanentFail:

root@c9b274ac17d4:~/cwl/complex-out# cwl-runner complex.cwl
/usr/local/bin/cwl-runner 1.0.20181012180214
Resolved 'complex.cwl' to 'file:///root/cwl/complex-out/complex.cwl'
[job complex.cwl] /tmp/tmp8x40850b$ bash \
    -c \
    'echo 42 > data.txt'
[job complex.cwl] Job error:
Error validating output record. the `value` field is not valid because
  `'42\n'` is not int
 in {
    "value": "42\n"
}
[job complex.cwl] completed permanentFail
{}
Final process status is permanentFail

I also tried array of integers and array of strings; but I stuck event at the basic things.

I also tried to add JS requirements, but it didn't help.

requirements:
  InlineJavascriptRequirement: {}

Should I use ExpressionTool or SchemaDefRequirement/CommandOutputBinding to issue an integer from workflow processing unit?

Thanks!

Answer 1

[Michael R. Crusoe]

Hello uagrif,

The newline that echo outputs is the cause, so pass -n:

-n do not output the trailing newline

#!/usr/bin/env cwl-runner
class: CommandLineTool
cwlVersion: "v1.0"

requirements:
  InlineJavascriptRequirement: {}

baseCommand: [bash, -c, "echo -n 42 > data.txt"]

inputs: []
outputs:
  value:
    type: int
    outputBinding:
      glob: data.txt
      loadContents: true
      outputEval: $(parseInt(self[0].contents))

Or without the InlineJavascriptRequirement

#!/usr/bin/env cwl-runner
class: CommandLineTool
cwlVersion: "v1.0"

baseCommand: [bash, -c, "echo '{"value": 42}'"]

inputs: []

stdout: cwl.output.json
outputs:
  value: int

[uagrif]

Hello Michael,

Thanks for replying!

I've tried your code sample but 42 is still read as string.

(cwl-runner) root@28d532f49b21:~/cwl/complex-out# cwl-runner complex.cwl
/root/python-dfab/cwl-runner/bin/cwl-runner 1.0.20181102182747
Resolved 'complex.cwl' to 'file:///root/cwl/complex-out/complex.cwl'
[job complex.cwl] /tmp/tmp4qS0RC$ bash \
    -c \
    'echo -n 42 > data.txt'
[job complex.cwl] Job error:
Error validating output record. the `value` field is not valid because
  `u'42'` is not int
 in {
    "value": "42"
}
[job complex.cwl] completed permanentFail
{}
Final process status is permanentFail

[Michael R. Crusoe]

Thanks, I've fixed my answer and added a version without Javscript

[uagrif]

Thank you Michael, that works!

Here is sample code without InlineJavascriptRequirement:

#!/usr/bin/env cwl-runner
class: CommandLineTool
cwlVersion: "v1.0"

baseCommand: [echo, "{\"value\": 42}"]

# redirect stdout into file with predefined name
stdout: cwl.output.json

inputs: []
outputs:
  value: int

And another sample with string array output:

#!/usr/bin/env cwl-runner
class: CommandLineTool
cwlVersion: "v1.0"

# write json with content {value: ["one", "two", "three"]} ...
baseCommand: [echo, "{\"value\": [\"one\", \"two\", \"three\"]}"]
# .. into file with predefined name "cwl.output.json"
stdout: cwl.output.json

inputs: []
outputs:
  value: string[]