Unable to omit 'Unrecognized extension field' warnings for fields in RDF schema

[bogdang989]

Hi,
The attachment contains test files to reproduce the issue (CWL and the xml schema file) - cwl-schemas.tar.gz

When running the tool with cwltool version 1.0.20180116213856 an 'Unrecognized extension field' warning is printed out. While running with an older version (1.0.20170413194156) there was no warning.
cwltool schemas-tool1.cwl -- --input foobar

Expected Behavior

After including $schemas in the CWL, cwltool should not print out a warning.

Actual Behavior

After including $schemas in the CWL, cwltool version 1.0.20180116213856 prints out a warning.

Workflow Code

class: CommandLineTool
cwlVersion: v1.0
id: schema
baseCommand:
  - echo
inputs:
  - id: input
    type: string?
    inputBinding:
      position: 0
outputs:
  - id: output
    type: File?
    outputBinding:
      glob: stdout.txt
label: schemas-tool1
requirements:
  - class: DockerRequirement
    dockerPull: 'ubuntu:16.04'
stdout: stdout.txt
$namespaces:
  sbg: 'http://sbgenomics.com/schema#'
$schemas:
  - sbg-schema.xml
'sbg:appVersion':
  - v1.0

Full Traceback

schemas-tool1.cwl:1:1: unrecognized extension field `http://sbgenomics.com/schema#appVersion`.  Did you include a $schemas section?

Your Environment

• cwltool version: 1.0.20180116213856
  Check using cwltool --version

[tetron]

Hmm, this looks correct, will have to investigate.

[tetron]

Also, this is actually a schema-salad bug. You should get a similar result using schema-salad-tool for validation:

schema-salad-tool common-workflow-language/v1.0/CommonWorkflowLanguage.yml schemas-tool1.cwl

[bogdang989]

Cool, thanks for the quick response.

[bogdang989]

Yes, actually when running with schema-salad-tool version 2.6.20171201034858 there is no warning

[tetron]

Perhaps you just need to update the version of schema salad you are using with cwltool?

[anton-khodak]

Confirming with cwltool 1.0.20180116213856, using schema-salad of the latest version does not help

[keiranmraine]

also #455 (discussed with mr-c on gitter 14-Dec)

[tetron]

This turn out to be a really stupid bug. The default for --skip-schemas is True so it would never load the schemas as intended. See #656 has the fix.

Separately there is a schema-salad bug in that it prints warnings even when extension schemas were skipped.

[mr-c]

@tetron I can confirm that the behaviour of --skip-schemas is fixed, however cwltool is still printing a false warning:
test.cwl:1:1: unrecognized extension field 'http://edamontology.org/topic_0003'. Did you include a $schemas section?

Here's a minimal reproducible version:

class: CommandLineTool
cwlVersion: v1.0
id: schema
baseCommand: [] 
inputs: []
outputs: []

$namespaces:
  edam: http://edamontology.org/

$schemas:
  http://edamontology.org/EDAM_1.16.owl

edam:topic_0003: edam:topic_3500

[tetron]

Here's what schema salad infers to be property (== RDF predicate) based on my understanding of RDFS and OWL:

• Anything rdf:type rdf:Property
• Anything defined as a subproperty of something else
• Anything listed as having a range of values
• Anything rdf:type owl:ObjectProperty

It doesn't look to me like edam:topic_0003 is any of those things. Is it mislabeled in the ontology or am I misunderstanding what should be considered a predicate in OWL?

        for s, _, _ in self.graph.triples((None, RDF.type, RDF.Property)):
            self._add_properties(s)
        for s, _, o in self.graph.triples((None, RDFS.subPropertyOf, None)):
            self._add_properties(s)
            self._add_properties(o)
        for s, _, _ in self.graph.triples((None, RDFS.range, None)):
            self._add_properties(s)
        for s, _, _ in self.graph.triples((None, RDF.type, OWL.ObjectProperty)):
            self._add_properties(s)

[mr-c]

@tetron Admittedly I just switched to EDAM as I don't have a copy of sbg-schema.xml.

Checking cwltool --validate https://raw.githubusercontent.com/common-workflow-language/workflows/master/tools/ucsc-liftOver.cwl (which uses schema.org) and I get no errors.

So I guess the issue remains that schemas-tool1.cwl:1:1: unrecognized extension field 'http://sbgenomics.com/schema#appVersion'. Did you include a $schemas section? is not a helpful error message.

[tetron]

Yes, the error message could be more helpful (like, if you did include $schemas section, it shouldn't lecture you about it). It could also list all the predicates it is expecting to find, although that could be long, and it's not nice to spam the user.

On EDAM, I think edam:topic_0003 is the "class of things which are topics in edam". I believe what you actually want is the predicate edam:has_topic, so see if this works:

edam:has_topic: edam:topic_3500

Which would express that the workflow was related to the topic of Zoology.