Fix/docker array secondaryfiles

[jeremiahsavage]

PR to allow this test to pass
common-workflow-language/common-workflow-language#292

[kmhernan]

@jeremiahsavage @mr-c hmm so with more investigation, it seems like for v1.0 at least within the if schema["type"] == "array" block there will be no secondary files passed to this would not work. I did get it to work by doing this around line 70:

   elif isinstance(schema["type"], dict):
            st = copy.deepcopy(schema["type"])
            if binding and "inputBinding" not in st and st["type"] == "array" and "itemSeparator" not in binding:
                st["inputBinding"] = {}
            #-- KMH FIX
            if 'secondaryFiles' in schema: st['secondaryFiles'] = schema['secondaryFiles']
            bindings.extend(self.bind_input(st, datum, lead_pos=lead_pos, tail_pos=tail_pos))

then back in the block where the commits are placed in this PR, I have:

            if schema["type"] == "array":
                for n, item in enumerate(datum):
                    b2 = None
                    if binding:
                        b2 = copy.deepcopy(binding)
                        b2["datum"] = item
                    #-- KMH FIX
                    if 'secondaryFiles' in schema:
                        bindings.extend(
                            self.bind_input(
                                {"type": schema["items"], "inputBinding": b2, "secondaryFiles": schema['secondaryFiles']},
                                item, lead_pos=n, tail_pos=tail_pos))
                    else:
                        bindings.extend(
                            self.bind_input(
                                {"type": schema["items"], "inputBinding": b2},
                                item, lead_pos=n, tail_pos=tail_pos))
                binding = None

Now, the secondaryFiles will be seen for v1.0. But I don't know if this is a bad approach? I have tested this using the 1.0.20160913171024 build of cwltool.

[jeremiahsavage]

@kmhernan I'll try testing your patch for v1.0

One problem there might be for v1.0, is that the spec removed mention of secondaryFiles in InputArraySchema, while it was present in the draft-3 spec.

See secondaryFile in draft-3:
5.4.1.1.2 InputArraySchema http://www.commonwl.org/draft-3/CommandLineTool.html#InputArraySchema
absent in v1.0:
5.4.2.2 InputArraySchema http://www.commonwl.org/v1.0/CommandLineTool.html#InputArraySchema

[tetron]

Thanks for looking into this. I have a fix pending in #211

[cwl-bot]

Can one of the admins verify this patch?

[tetron]

Thanks, I believe this is fixed so I'm closing the PR.