Python: fix(redis): pass collection prefix as a list

[MukundaKatta]

Summary

• pass Redis IndexDefinition.prefix as a single-item list instead of a plain string
• update the manual test fixture to match the correct redis-py API shape
• add a regression test that asserts the connector builds the index definition with exactly one collection prefix

Closes #13894

Verification

• ran git diff --check
• added a focused unit test around ensure_collection_exists() prefix construction

Notes

• I did not run the Python test suite locally because the required Redis/Python test dependencies were not installed in this environment.

[github-actions]

Automated Code Review

Reviewers: 4 | Confidence: 95%

✗ Correctness

The production code fix (wrapping the prefix argument in a list) is correct — IndexDefinition._append_prefix calls len(prefix) and iterates over it, so passing a bare string like "test:" would produce PREFIX 5 t e s t : instead of the intended PREFIX 1 test:. However, the new test test_create_index_uses_single_collection_prefix has a bug: it parametrizes expected_index_type as "hash"/"json" strings, then passes them to IndexType(expected_index_type). Since IndexType is an Enum with integer values (HASH = 1, JSON = 2), calling IndexType("hash") raises ValueError: 'hash' is not a valid IndexType. The test will fail at assertion time, not because the code is wrong, but because the expected value cannot be constructed.

✓ Security Reliability

This is a correctness bug fix for the Redis connector. The IndexDefinition constructor's prefix parameter is iterable, and passing a bare string like "test:" causes redis-py to iterate over each character as a separate prefix (e.g., "t", "e", "s", "t", ":") instead of treating the whole string as one prefix. Wrapping it in a list ["test:"] ensures the prefix is correctly interpreted as a single entry. The new test properly validates this behavior by patching IndexDefinition and asserting it receives the list form. No security or reliability issues found.

✓ Test Coverage

The diff changes the prefix argument to IndexDefinition from a bare string to a single-element list, matching the redis-py API expectation. The existing test_create_index_manual test is updated to reflect the new list format, and a new parametrized test test_create_index_uses_single_collection_prefix is added to verify both hash and JSON collection types pass prefix=["test:"] to IndexDefinition. The new test properly patches IndexDefinition at the correct import path, asserts call arguments, and verifies the returned definition object is forwarded to create_index. Test coverage for this change is thorough.

✗ Design Approach

The core fix — changing prefix=f"{self.collection_name}:" (a string) to prefix=[f"{self.collection_name}:"] (a list) — is correct and necessary. IndexDefinition._append_prefix iterates with for p in prefix, so a bare string would iterate individual characters, producing 5 (or more) nonsense prefix entries instead of one. The accompanying test update to test_create_index_manual is also correct. However, the new parametrized test test_create_index_uses_single_collection_prefix contains a bug: it passes string values "hash" and "json" to IndexType(expected_index_type), but IndexType is a standard Enum with integer values (HASH=1, JSON=2). IndexType("hash") will raise ValueError: 'hash' is not a valid IndexType at test execution time, meaning the new test always fails and provides no coverage.

---

Automated review by MukundaKatta's agents

[github-actions]

Bug: IndexType("hash") and IndexType("json") will raise ValueError because IndexType is an Enum with integer values (HASH = 1, JSON = 2), not string values. The parametrize should use the actual IndexType enum members directly.

Suggested change

	@mark.parametrize(
	("fixture_name", "expected_index_type"),
	[
	("collection_hash", "hash"),
	("collection_json", "json"),
	@mark.parametrize(
	("fixture_name", "expected_index_type"),
	[
	("collection_hash", IndexType.HASH),
	("collection_json", IndexType.JSON),
	],
	)