Empty array/list parses as None

[apexo]

1. zeep version 3.2.0
2. WSDL: https://gist.github.com/apexo/f1ffbbb37764032bebb1936c3e8e8c25
3. Script: https://gist.github.com/apexo/372f69deee0a132458d506e78e8bea3d

So I have a WSDL with a request that contains an optional repeatable element <list-entry> (minOccurs=0, maxOccours=unbounded) that is wrapped in another element (<list-wrapper>) .

The sample script parses this document:

<env:Envelope xmlns:env="http://schemas.xmlsoap.org/soap/envelope/">
  <env:Body>
    <testRequest>
      <list-wrapper>
        <list-entry>1</list-entry>
      </list-wrapper>
    </testRequest>
  </env:Body>
</env:Envelope> 

as follows (kind of expected):

{
    'list-wrapper': {
        'list-entry': [
            '1'
        ]
    },
    'dummy': None
}

while

<env:Envelope xmlns:env="http://schemas.xmlsoap.org/soap/envelope/">
  <env:Body>
    <testRequest>
      <list-wrapper>
      </list-wrapper>
    </testRequest>
  </env:Body>
</env:Envelope> 

is parsed as

{
    'list-wrapper': None,
    'dummy': None
}

I find this a bit surprising. IMO, the latter should parse as

{
    'list-wrapper': {
        'list-entry': []
    },
    'dummy': None
}

Otherwise it is impossible to tell whether <list-wrapper> itself was present in the request or not (if <list-wrapper> were optional in the WSDL which it is not in the reduced sample). But whether the wrapper is present or not sometimes makes a semantic difference. And that's my problem. Thank you for your attention.

[apexo]

If I comment out lines 160-161 in xsd/types/complex.py

            #if allow_none and len(elements) == 0 and len(attributes) == 0:
            #    return

it works as expected.

[apexo]

This change breaks 3 test cases:

• test_any_optional
• test_build_sequence_with_optional_elements
• test_nested_complex_type_optional

For test_any_optional, I can't really tell what the expected behaviour could be.

For test_build_sequence_with_optional_elements, I'd say: works as expected if we want the wrapper element (item_2) to be significant. Also I think the test case is wrong, item_2 shouldn't be empty. AFAIK, xsi:nillable="true" doesn't imply minOccurs=0.

For test_nested_complex_type_optional, kind of the same as test_build_sequence_with_optional_elements. Also: item_2 itself shouldn't be empty? It should contain a sequence of elements, of which at least one (item_2b) is not optional.

[martinlehoux]

I have the same issue, I will try you solution @apexo