Inconsistent serialisation of simple classes

[johnduffield]

I have an odd situation where a simple class containing just a string and a list of strings is serialised differently to other very similar classes.
The first string is not treated as an element unless I explicitly declare it, but only for that one case. Notice how MyClass2 is serialised differently to the others in the following example:

from dataclasses import dataclass, field

from xsdata.formats.dataclass.serializers.config import SerializerConfig
from xsdata.formats.dataclass.serializers import XmlSerializer

@dataclass
class MyClass1:
    first_string: str
    second_string: str        

@dataclass
class MyClass2:
    first_string: str
    list_of_stuff: list[str] = field(        
        metadata={"wrapper": "stuff_list", "name": "stuff", "type": "Element"}
    )                  

@dataclass
class MyClass3:
    first_string: str
    list_of_stuff: list[str] = field(        
        metadata={"wrapper": "stuff_list", "name": "stuff", "type": "Element"}
    )                  
    another_string: str

@dataclass
class MyClass4:
    first_string: str = field(metadata={"type": "Element"})
    list_of_stuff: list[str] = field(        
        metadata={"wrapper": "stuff_list", "name": "stuff", "type": "Element"}
    )                  

test1 = MyClass1("simple class with two strings","second string")                 
test2 = MyClass2("replacing the second string with a list makes the 'first_string' tag disappear", 
                       ["first string in list", "second string in list"])
test3 = MyClass3("Adding another string to the class fixes it", 
                       ["first string in list", "second string in list"], 
                       "another")
test4 = MyClass4("Explicitly declaring 'first_string' as an Element also works", 
                       ["first string in list", "second string in list"])

print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test1))
print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test2))
print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test3))
print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test4))

Output:

<?xml version="1.0" encoding="UTF-8"?>
<MyClass1>
  <first_string>simple class with two strings</first_string>
  <second_string>second string</second_string>
</MyClass1>

<?xml version="1.0" encoding="UTF-8"?>
<MyClass2>replacing the second string with a list makes the 'first_string' tag disappear<stuff_list>
    <stuff>first string in list</stuff>
    <stuff>second string in list</stuff>
  </stuff_list>
</MyClass2>

<?xml version="1.0" encoding="UTF-8"?>
<MyClass3>
  <first_string>Adding another string to the class fixes it</first_string>
  <stuff_list>
    <stuff>first string in list</stuff>
    <stuff>second string in list</stuff>
  </stuff_list>
  <another_string>another</another_string>
</MyClass3>

<?xml version="1.0" encoding="UTF-8"?>
<MyClass4>
  <first_string>Explicitly declaring 'first_string' as an Element also works</first_string>
  <stuff_list>
    <stuff>first string in list</stuff>
    <stuff>second string in list</stuff>
  </stuff_list>
</MyClass4>

[johnduffield]

After more investigation, it looks like this is an issue when classes have metadata on some, but not all members:

from dataclasses import dataclass, field

from xsdata.formats.dataclass.serializers.config import SerializerConfig
from xsdata.formats.dataclass.serializers import XmlSerializer

@dataclass
class MyClass1:
    first_string: str
    second_string: str        

@dataclass
class MyClass2:
    first_string: str
    second_string: str = field(metadata={"type": "Element"})        

@dataclass
class MyClass3:
    first_string: str = field(metadata={"type": "Element"})
    second_string: str         

test1 = MyClass1("simple class with two strings","second string")                 
test2 = MyClass2("Adding metadata to the second string makes the 'first_string' tag disappear", 
                "second string")

test3 = MyClass3("first string",
                 "Adding metadata to the first string makes the 'second_string' tag disappear",)

print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test1))
print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test2))
print(XmlSerializer(config=SerializerConfig(pretty_print=True)).render(test3))

Output:

<?xml version="1.0" encoding="UTF-8"?>
<MyClass1>
  <first_string>simple class with two strings</first_string>
  <second_string>second string</second_string>
</MyClass1>

<?xml version="1.0" encoding="UTF-8"?>
<MyClass2>Adding metadata to the second string makes the 'first_string' tag disappear<second_string>second string</second_string>
</MyClass2>

<?xml version="1.0" encoding="UTF-8"?>
<MyClass3>
  <first_string>first string</first_string>Adding metadata to the first string makes the 'second_string' tag disappear</MyClass3>

[tefra]

Since some fields don't have the metadata type, the library is trying to decide between an Element or a simple Text Node.

When there is only one undefined one, it defaults to Text Node, when there are more than one undefined it assumes they are Elements.

src