types for input/output not consistent

[gijzelaerr]

cwlVersion: v1.0
class: Workflow

inputs:
 a: int
 b: int[]
 c:
   type: int?

outputs: {}

steps: {}

python:

IN: from cwl_utils.parser_v1_0 import load_document
IN: [i.type for i in load_document('test.cwl').inputs]
Out: 
['int',
 <cwl_utils.parser_v1_0.InputArraySchema at 0x7f9242121400>,
 ['null', 'int']]

Shouldn't the InputArraySchema object also be a string, for example array?

[mr-c]

The generated parsers embody all the flexibility of the schema-salad definition of CWL.

It is feasible to add an field to the schema format and use that to indicate the canonical representation of a field. Then one would enhance schema-salad-tool --codegen python to either normalize to that formulation on read in or on field access (either option would enable proper PEP 484 Python type hints as well).

@tetron , what do you think of my assessment?