Implementing the xs:all tag

[Artanys]

[plq]

jenkins, run tests

[plq]

hey, thanks for the patch!

one of your tests are failing, can you have a look?

[plq]

What's the point of having a group name for the xs:all tag? What happens when you pass different values to xml_all_group ?

[Artanys]

Will fix tests, pushed this out a bit late last night :)

I modeled the group name after the implementation of xs:choice, with the addition that xs:all is a direct subelement of complexType instead of in a sequence. I will add an additional failure case for when you specify two "all groups" in the same element

[Artanys]

  <xs:complexType name="AbcAll">
    <xs:all>
      <xs:element name="abc" type="xs:string"/>
      <xs:element name="bcd" type="xs:string" minOccurs="0"/>
    </xs:all>
    <xs:all>
      <xs:element name="cde" type="xs:string" minOccurs="0"/>
    </xs:all>
  </xs:complexType>

gets generated if I specify something like

class AbcAll(ComplexModel):
    abc = Unicode(xml_all_group="all1", min_occurs=1)
    bcd = Unicode(xml_all_group="all1")
    cde = Unicode(xml_all_group="all2")

This then fails lxml schema validation in

Traceback (most recent call last):
  File "/home/artanys/dev/dvenv/spyne/spyne/interface/xml_schema/_base.py", line 223, in build_validation_schema
    self.validation_schema = etree.XMLSchema(etree.parse(f))
  File "src/lxml/xmlschema.pxi", line 86, in lxml.etree.XMLSchema.__init__

I can add an explicit check for this in the code and throw a reasonable error

[plq]

It's been some time since I looked at the xml schema spec so please correct me if I'm wrong here.

xsd:choice was implemented that way because there can be multiple choice tags. this is not the case for the xsd:all tag (is it?).

Instead of throwing an error, we have two options depending on whether you can have both <xsd:all> and <xsd:sequence> inside one complexType element.

If it's possible, we could change xml_all_group:str to xml_main_group:exactly_one_of('all', 'sequence') so you eliminate the source of the error in its root.

so your example becomes:

class AbcAll(ComplexModel):
    abc = Unicode(xml_main_group='all', min_occurs=1)
    bcd = Unicode(xml_main_group='all')
    cde = Unicode(xml_main_group='sequence')
    def = Unicode(xml_main_group='sequence')

which would generate:

  <xs:complexType name="AbcAll">
    <xs:all>
      <xs:element name="abc" type="xs:string"/>
      <xs:element name="bcd" type="xs:string" minOccurs="0"/>
    </xs:all>
    <xs:sequence>
      <xs:element name="cde" type="xs:string" minOccurs="0"/>
      <xs:element name="def" type="xs:string" minOccurs="0"/>
    </xs:sequence>
  </xs:complexType>

If having both all and sequence elements inside the complexType is not possible, then the main group should be implemented as a class attribute rather than an element attribute. So your example would become:

class AbcAll(ComplexModel):
    class Attributes(ComplexModel.Attributes):
        xml_main_group='all'

    abc = Unicode(min_occurs=1)
    bcd = Unicode
    cde = Unicode

which would generate

  <xs:complexType name="AbcAll">
    <xs:all>
      <xs:element name="abc" type="xs:string"/>
      <xs:element name="bcd" type="xs:string" minOccurs="0"/>
      <xs:element name="cde" type="xs:string" minOccurs="0"/>
    </xs:all>
  </xs:complexType>

It seems both could be implemented with minor modifications to your current patch.

So, what does the spec say?

[Artanys]

Schema says they cannot coexist. Will look into making it a class attribute