Keep extra elements and attributes, mixed with predefined ones

[RobertoRoos]

I have a situation where I want to: load a massive XML, interact with a small set of elements and attributes and re-save the XML again.
There is no interest in most of the schema, defining a full model would just be a burden. Fortunately we can load most elements as raw XML (as described here: https://pydantic-xml.readthedocs.io/en/latest/pages/data-binding/raw.html).

However, this is only for entire elements. We still have some cases where an element is partially relevant. Take for instance this XML:

<company type="private" language="en">
    <contact>
        <email>info@business.nl</email>
    </contact>
    <product>Ice-cream</product>
    <product>Bread</product>
    <product>Milk</product>
</company>

With this model:

class Contact(BaseXmlModel):
    email: str = element()


class Company(BaseXmlModel, tag="company", arbitrary_types_allowed=True):
    type_: str = attr(name="type")

    contact: Contact
    products_raw: list[Element] = element(tag="product")

This loads and I can use the important entities as desired. But then when I export this again to XML the language="en" attribute is dropped.
Similarly, if the original XML has more child-elements that I didn't specifically call by name, they will have disappeared in the export.

How can I make sure extra entities like these will remain after the export?

[RobertoRoos]

The workaround now is to add extra="forbid" to the class, such that it will refuse to load unknown entities, to prevent from quietly disappearing after an export:

class Company(BaseXmlModel, ... , extra="forbid", ....):

Using extra="allow" does make the model pass validation, but that extra data is not really kept on. Is this as intended? I seems extra="ignore" and extra="allow" do the same thing now.

[RobertoRoos]

@dapper91 could you have a look at how this is going? Or maybe assign somebody else we could provide some insight?
It would be great if this has a future before I put more time in it. Thanks!

[RobertoRoos]

@dapper91 is there any interest in accepting PRs for this? Thanks!