fix(checks): dot-wildcard/multi-field JSONPath returns scalar vs list depending on match count

[chuenchen309]

Problem

_is_list_expression special-cases bracket wildcards (Slice, Union, Intersect) to force resolve() to return a list, but never checks jsonpath_ng.Fields — what dot-wildcards like trace.last.metadata.* (and comma-separated multi-field selectors) actually compile to.

As a result, resolve() falls through to its match-count heuristic (len(matches) > 1 or _is_list_expression(expression)) for these expressions: a wildcard matching exactly 1 key returns a bare scalar, while 2+ matches correctly returns a list. This breaks ComparisonCheck's match=any/all/none semantics (which require list/set/tuple) whenever the matched dict happens to have exactly one key — entirely dependent on incidental data shape, not the expression's own structure.

Fix

Add a Fields branch treating any multi-field selector (len(expression.fields) > 1) or wildcard ("*" in expression.fields) as a list-expression, matching the existing bracket-wildcard handling.

Testing

• Added TestResolveDotWildcard covering the single-match, multi-match, and non-wildcard-single-field cases.
• Confirmed red→green: reverting only extraction.py reproduces 'v1' == ['v1'] (bare scalar instead of a list) for the single-match wildcard case; reapplying passes.
• Full libs/giskard-checks/tests/ suite: 741 passed, 4 skipped (unrelated).
• ruff check/ruff format --check and basedpyright — 0 errors (pre-existing unrelated warnings elsewhere in the file untouched).
• xref: no open PR touches core/extraction.py.

AI-Generated disclosure

Found via an AI-assisted code review pass (Claude Code) over giskard-checks/src/giskard/checks/core/. I personally reproduced the match-count-dependent behavior with a minimal repro, verified the fix against single-match/multi-match/non-wildcard cases, and ran the full test suite plus lint/type checks before submitting.

[gemini-code-assist]

Code Review

This pull request updates the JSONPath extraction logic to recognize multi-field and wildcard selectors (using the Fields expression) as list expressions, ensuring they consistently resolve to a list. Unit tests were added to verify wildcard resolution. The reviewer suggested adding an additional unit test to explicitly cover the multi-field selector branch to ensure complete test coverage.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

While the new implementation in extraction.py correctly handles multi-field selectors (where len(expression.fields) > 1), there is currently no unit test covering this specific branch. Adding a test case for a multi-field selector (e.g., using [k1,k2]) will ensure full test coverage and prevent future regressions.

Suggested change

	def test_multi_field_selector_returns_a_list(self):
	trace = Trace[str, str](
	interactions=[
	Interaction(
	inputs="hi",
	outputs="hello",
	metadata={"k1": "v1", "k2": "v2", "k3": "v3"},
	)
	]
	)
	result = resolve(trace, "trace.last.metadata.[k1,k2]")
	assert isinstance(result, list)
	assert sorted(result) == ["v1", "v2"]

[kevinmessiaen]

LGTM, thanks for the fix!