[python] Add mock API tests for XML scenarios from PR #10063 and fix XML serialization bugs (WIP)

[Copilot]

Adds Python mock API test coverage for the new XML handling scenarios introduced in packages/http-specs PR #10063, and fixes the underlying generator bugs that were blocking 4 of those tests.

Dependency update

• Bumps @typespec/http-specs 0.1.0-alpha.35-dev.4 → 0.1.0-alpha.35-dev.5 (first release containing the new scenarios)

New test cases (sync + async)

11 new scenarios added to generic_mock_api_tests/test_payload_xml.py and its async counterpart — all 26 tests pass:

Scenario	Status
ModelWithRenamedProperty — @name on scalar field	✅ passes
ModelWithNestedModel — nested model property	✅ passes
ModelWithRenamedNestedModel — nested model whose type has @Xml.name	✅ passes
ModelWithWrappedPrimitiveCustomItemNames — custom wrapper + item names on primitive array	✅ passes
ModelWithRenamedWrappedModelArray — custom wrapper name on model array	✅ passes
ModelWithRenamedWrappedAndItemModelArray — custom wrapper + item names on model array	✅ passes
ModelWithRenamedAttribute — @attribute with renamed XML name	✅ passes
ModelWithUnwrappedModelArray — unwrapped model array with custom item name	✅ passes
ModelWithRenamedUnwrappedModelArray — unwrapped model array with renamed item name	✅ passes
ModelWithNamespace — XML namespace on model	✅ passes
ModelWithNamespaceOnProperties — per-property XML namespaces	✅ passes

Generator bug fixes (model_base.py.jinja2)

Four bugs in the XML runtime template were fixed:

1. Unwrapped model array serialization: _get_element was ignoring itemsName from parent property metadata when serializing model items in array context, emitting <SimpleModel> instead of <items>. Fixed by using parent_meta.get("itemsName") when present.

2. Namespace key mismatch: DPG-generated models store namespace as "namespace" key in _xml dicts, but the runtime only read "ns". Deserialization now reads both keys with correct prefix-aware propagation semantics.

3. Default vs prefixed namespace propagation: XML default namespace (xmlns="...") is inherited by child elements; prefixed namespace (xmlns:smp="...") is not. Deserialization now only propagates the model namespace to property lookups when the model has no prefix (default namespace).

4. Python ET reserved prefix: ET.register_namespace("ns2", ...) raises ValueError because Python reserves ns\d+ patterns. The fix wraps the call in try/except and falls back to getattr(ET, "_namespace_map") to honour the intended prefix.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>[python] add test case for #10063
</issue_title>
<issue_description>follow skill https://github.com/microsoft/typespec/blob/main/.github/skills/python-sdk-spector-mock-api-tests/SKILL.md to write test case for #10063
</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes [python] add test case for https://github.com/microsoft/typespec/pull/10063  #10097

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/@typespec/http-client-python@10098

commit: 9540ee7

[github-actions]

All changed packages have been documented.

• ✅ @typespec/http-client-python

Show changes

@typespec/http-client-python - internal ✏️

Add mock API test cases for XML scenarios introduced in #10063, covering: renamed property, nested model, renamed nested model, wrapped primitive with custom item names, model array variants (wrapped/unwrapped/renamed), renamed attribute, namespace, and namespace-on-properties. Tests for unwrapped model array serialization and namespace handling are skipped pending generator bug fixes.

[azure-sdk]

You can try these changes here

🛝 Playground	🌐 Website	🛝 VSCode Extension

[msyyc]

@copilot Try to fix the bug so that we could remove all the skip of related test case of this PR.

[Copilot]

Fixed in 3bbc458. Four bugs were addressed in model_base.py.jinja2:

1. Unwrapped model arrays — _get_element now uses parent_meta.get("itemsName") when present, so <items> / <ModelItem> tags are emitted instead of the model's type name.
2. Namespace key mismatch — DPG-generated models use "namespace" key; deserialization now reads both "ns" and "namespace" (with prefix-aware propagation so default-namespace models still work).
3. Namespace propagation — model namespace is only propagated to child-element lookups when the model uses a default namespace (no prefix), matching actual XML semantics.
4. Python ET reserved prefix — ET.register_namespace rejects prefixes matching ns\d+ (e.g. "ns2"); the fix falls back to ET._namespace_map via getattr so the intended prefix is used when serialising.

All 26 mock API tests and 35 XML unit tests now pass with no skips.