fix(dataset): raise a clear error when custom dataset_info is not a list

[he-yufeng]

What

register_dataset_info (used by --custom_dataset_info xxx.json) loads the JSON and iterates it with for d_info in dataset_info: _register_d_info(d_info). When a single entry is written as a top-level object {...} instead of the documented one-element list [{...}], for d_info in dataset_info iterates the dict keys (strings), and _preprocess_d_info then runs d_info['preprocess_func'] = ... on a str, raising a cryptic TypeError: 'str' object does not support item assignment with no hint that the JSON shape is wrong.

The built-in dataset_info.json is a list and the docs state it "is a list of metadata for datasets", so a top-level non-list is a user mistake. Dropping the outer [] when copying a single-entry example is an easy one to make.

Fix

Validate that the loaded dataset_info is a list and raise a helpful ValueError otherwise. Valid list inputs are unaffected.

Verification

Reproduced the dispatch logic standalone (no swift deps needed for the type check):

input	before	after
[{...}] (documented)	ok	ok
{...} (top-level object)	TypeError: 'str' object does not support item assignment	ValueError: dataset_info should be a JSON list ...

test_register_dataset_info_rejects_non_list writes a temp .json holding a top-level object and asserts register_dataset_info raises ValueError.

[gemini-code-assist]

Code Review

This pull request introduces a validation check in register_dataset_info to ensure that dataset_info is a list, raising a ValueError with a clear message if it is not. It also adds a corresponding unit test to verify this behavior. The review feedback highlights two important issues: first, a potential crash in the logging logic of register_dataset_info where .keys() might be called on the list, along with an inaccurate type hint; second, a resource leak in the new test because the temporary file created with delete=False is never cleaned up.

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]

Since dataset_info is now strictly validated to be a list, the logging logic on line 117 will crash with an AttributeError: 'list' object has no attribute 'keys' if log_msg is None and len(dataset_info) >= 5. This is because dataset_info.keys() is called on a list.

To fix this, please update line 117 to slice the list instead of calling .keys(). For example:

log_msg = dataset_info if len(dataset_info) < 5 else dataset_info[:5]

Additionally, the type hint for dataset_info on line 84 (Union[str, List[str], None]) is misleading because dataset_info is actually a list of dictionaries (dataset entries), not a list of strings. It should be updated to Union[str, List[Dict[str, Any]], None].

[he-yufeng]

Fixed in 79f3cc5. You're right — once the list validation is in place, list(dataset_info.keys()) (a leftover from when dataset_info was a dict) crashes for 5+ entries whenever log_msg is None (a JSON-list string or a list passed directly). Switched it to dataset_info[:5] and tightened the annotation to Union[str, list, None] since the entries are dicts, not str.

[gemini-code-assist]

The temporary file created with delete=False is never cleaned up after the test runs. We should wrap the test execution in a try...finally block to ensure the temporary file is deleted, preventing resource leaks in the test environment.

Suggested change

	with tempfile.NamedTemporaryFile('w', suffix='.json', delete=False, encoding='utf-8') as f:
	json.dump({'ms_dataset_id': 'swift/self-cognition'}, f)
	path = f.name
	try:
	register_dataset_info(path)
	except ValueError as e:
	assert 'list' in str(e)
	else:
	raise AssertionError('expected ValueError for a non-list dataset_info')
	import os
	with tempfile.NamedTemporaryFile("w", suffix=".json", delete=False, encoding="utf-8") as f:
	json.dump({"ms_dataset_id": "swift/self-cognition"}, f)
	path = f.name
	try:
	try:
	register_dataset_info(path)
	except ValueError as e:
	assert "list" in str(e)
	else:
	raise AssertionError("expected ValueError for a non-list dataset_info")
	finally:
	if os.path.exists(path):
	os.remove(path)

[he-yufeng]

Fixed in 79f3cc5 — wrapped the assertion in an inner try/except/else and added an outer finally: os.remove(path) so the delete=False temp file is always cleaned up.