feat(api_core): add universe and endpoint routing logic to gapic_v1 public helpers

[hebaalazzeh]

Centralizes generic routing, mTLS, and universe domain resolution helpers under google.api_core.gapic_v1.client_utils.

Exposes get_api_endpoint, get_default_mtls_endpoint, and get_universe_domain as public API helpers to be used by newly generated clients.

The unit tests added in test_client_utils.py are identical to the generator's original routing test cases in test_service.py.j2 (namely test__get_api_endpoint and test__get_universe_domain), modified only to call the public helpers directly:

• test__get_api_endpoint maps to the new get_api_endpoint helper.
• test__get_universe_domain maps to the new get_universe_domain helper.

[gemini-code-assist]

Code Review

This pull request introduces a new routing module to handle mTLS endpoint resolution and universe domain configuration, along with corresponding unit tests and a test fixture for mTLS environment isolation. It also includes minor refactoring in noxfile.py and test_bidi.py. The reviewer identified an issue with the regex-based endpoint conversion logic in routing.py and suggested a more robust string-based implementation to prevent potential security vulnerabilities.

[hebaalazzeh]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces a new routing module in gapic_v1 to handle API endpoint and universe domain resolution, along with comprehensive unit tests. The feedback suggests improving get_default_mtls_endpoint by making domain suffix checks case-insensitive to correctly handle mixed-case or uppercase endpoints, and adding corresponding test cases to verify this behavior.

I am having trouble creating individual review comments. Click here to see my feedback.

packages/google-api-core/google/api_core/gapic_v1/routing.py (36-47)

Domain names are case-insensitive. Performing case-sensitive checks like .endswith(".googleapis.com") can fail to recognize mixed-case or uppercase endpoints (e.g., foo.GoogleAPIs.com), resulting in a failure to resolve the correct mTLS endpoint. We should convert the endpoint to lowercase before performing suffix checks.

    if not api_endpoint or ".mtls." in api_endpoint.lower():
        return api_endpoint

    lowered_endpoint = api_endpoint.lower()
    if lowered_endpoint.endswith(".sandbox.googleapis.com"):
        # len(".sandbox.googleapis.com") == 23
        return api_endpoint[:-23] + ".mtls.sandbox.googleapis.com"

    if lowered_endpoint.endswith(".googleapis.com"):
        # len(".googleapis.com") == 15
        return api_endpoint[:-15] + ".mtls.googleapis.com"

    return api_endpoint

packages/google-api-core/tests/unit/gapic/test_routing.py (35-52)

Add test cases to verify that get_default_mtls_endpoint correctly handles case-insensitive domain suffixes (e.g., foo.GoogleAPIs.com).

def test_get_default_mtls_endpoint():
    # Test valid API endpoints
    assert get_default_mtls_endpoint("foo.googleapis.com") == "foo.mtls.googleapis.com"
    assert (
        get_default_mtls_endpoint("foo.sandbox.googleapis.com")
        == "foo.mtls.sandbox.googleapis.com"
    )
    # Test case-insensitivity
    assert (
        get_default_mtls_endpoint("foo.GoogleAPIs.com")
        == "foo.mtls.googleapis.com"
    )
    assert (
        get_default_mtls_endpoint("foo.Sandbox.GoogleAPIs.com")
        == "foo.mtls.sandbox.googleapis.com"
    )

    # Test endpoints that shouldn't be converted
    assert (
        get_default_mtls_endpoint("foo.mtls.googleapis.com")
        == "foo.mtls.googleapis.com"
    )
    assert get_default_mtls_endpoint("foo.com") == "foo.com"

    # Test empty/None endpoints
    assert get_default_mtls_endpoint("") == ""
    assert get_default_mtls_endpoint(None) is None

[nbayati]

I noticed that the logic in get_api_endpoint for handling use_mtls_endpoint closely overlaps with the existing logic in google.auth.transport.mtls.should_use_mtls_endpoint().

To maintain a single source of truth for evaluating the mTLS endpoint configuration and compliance with AIP-4114, consider refactoring get_api_endpoint to call should_use_mtls_endpoint() directly.

This might also allow you to remove the use_mtls_endpoint argument entirely from get_api_endpoint (since should_use_mtls_endpoint handles reading the GOOGLE_API_USE_MTLS_ENDPOINT environment variable natively), which would streamline the method signature and reduce duplicate logic in the generated GAPIC clients.

[daniel-sanche]

^ I gave similar feedback for read_enviornment_variables in the other PR. I didn't realize that code was copied in so many places

[nbayati]

Thanks for this refactoring! The mTLS logic looks solid. I have a few comments regarding some edge cases (specifically around is_proto3_optional and mock test coverage). I'll be OOO next week and I don't want to delay this PR, so there's no need to wait on my re-review. I'll defer to @daniel-sanche for the final review and sign-off!

---

1. packages/google-api-core/google/api_core/gapic_v1/requests.py (Line 68)
For proto-plus messages or other objects when is_proto3_optional=True, this fallback logic if not getattr(request, field_name, None): incorrectly overwrites an explicitly set empty string "" with a new UUID. The code should distinguish between missing fields and fields explicitly set to empty values, similar to how dictionaries are handled on line 55.

2. packages/google-api-core/tests/unit/gapic/test_requests.py (Line 30 and Line 47)
The __contains__ methods defined in MockRequest and MockValueErrorRequest are dead code. The setup_request_id function only uses the in operator for dict types, so these methods will never be executed.

3. packages/google-api-core/tests/unit/gapic/test_requests.py (Line 68)
There is a missing test case for is_proto3_optional=True with an explicit empty string for non-dict objects (e.g., (MockRequest(request_id=""), True, "")). Adding this would correctly expose the bug where setup_request_id overwrites explicitly set empty strings for proto-plus messages.

4. packages/google-api-core/tests/unit/gapic/test_requests.py (Line 113)
Using re.match allows trailing characters in the string because it only checks for a match at the beginning of the string. This should be changed to re.fullmatch(UUID_REGEX, value) (or the regex should be anchored with $) to guarantee strict UUID validation.