def load_dataset()

in evalbench/dataset/dataset.py [0:0]

• 20 lines of code
• 3 McCabe index (conditional complexity)

def load_dataset(dataset: Sequence[dict], config):
    input_items: dict[str, list[EvalInputRequest]] = {"dql": [], "dml": [], "ddl": []}
    for item in dataset:
        if not _item_meets_config_filters(item, config):
            continue
        eval_input = EvalInputRequest(
            id=item["id"],
            nl_prompt=item["nl_prompt"],
            query_type=item["query_type"].lower(),
            database=item["database"],
            dialects=_union_dialects(item["dialects"], config.get("dialects", [])),
            golden_sql=item["golden_sql"],
            eval_query=item["eval_query"],
            setup_sql=item["setup_sql"],
            cleanup_sql=item["cleanup_sql"],
            tags=item["tags"],
            other=build_normalized_other(item["other"]),
        )
        input_items[eval_input.query_type].append(eval_input)
    return input_items