Support XML Schema Definition Language (XSD) import

[multimeric]

Adds an XsdImportEngine, with tests.

XSD doesn't map cleanly to LinkML, because:

• XML has both attributes and child elements. I treat both as slots, but tag each with a keyword that indicates which one it is
• The root element is a pseudo class, because it may enforce a specific structure. I resolve this by adding a RootElement class where necessary

[sierra-moxon]

Thank you @multimeric! I think there is a lot of interest in an XML -> LinkML importer.

[cmungall]

Thank you!!

Re lack of isomorphism. See also https://stackoverflow.com/questions/191536/converting-xml-to-json-using-python which mentions a "standard" (for instance-level). If there is nothing more up to date, I suggest being consistent with this, such that xml --[xmltodict]--> json validates via [xml schema]-->linkml

Or at least an option to do this - for now just marking this in the docs is sufficient

[multimeric]

Hmm that's an interesting suggestion. However the proposed solution is to prepend @ to attribute names such that:

<p id="1">text</p>

Becomes

{
  "p": {
    "@id": 1,
    "$": "text"
  }
}

I think this is a bit ugly, and also perhaps easily confused with JSON-LD, but I will implement it that way if you prefer.

I consider this scenario a bit different from standard XML to JSON conversion because here we have a schema that can describe which slots are attributes directly. So I'm wondering fi there's a good field in the LinkML SlotDefinition that would capture this (current I'm using keywords).

[cmungall]

And now I reflect this you are right, this would make things quite complicated. Having the @ at the schema level isn't permitted in linkml, so there would need to be a mapping preserved... all quite ugly.

So scratch that, let's keep it in mind for future extensions.

Formally the way to do this in linkml would be to use conforms_to or instantiates

slots:
  id:
    conforms_to: xsd:attribute
  description:
    conforms_to: xsd:attribute

or

slots:
  id:
    instantiates: [xsd:attribute]
  description:
    instantiates: [xsd:attribute]

The former is stronger and could be used for validation if we later implement metaclasses for xml schema (see https://linkml.io/linkml/schemas/annotations.html#validation-of-annotations).

(a to be defined xml schema metamodel):

class:
  XmlAttribute:
    is_a: slot
    class_uri: xsd:attribute
    description: If a slot instantiates this then it should be mapped to an attribute when serializing to XML

However, if you want to keep your method flexible such that you can pass in a profile on the command line (e.g. using keywords) and satisfy your use case first that's valid!