def load_dataset_from_bird_format()

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

• 22 lines of code
• 2 McCabe index (conditional complexity)

def load_dataset_from_bird_format(dataset: Sequence[dict]):
    input_items: dict[str, list[EvalInputRequest]] = {"dql": [], "dml": [], "ddl": []}
    dialect = "sqlite"
    query_type = "dql"
    for item in dataset:
        eval_input = EvalInputRequest(
            id=item["question_id"],
            nl_prompt="".join([item["question"], item["evidence"]]).replace(
                "`", '"'
            ),
            query_type=query_type,
            database=item["db_id"],
            dialects=[dialect],
            golden_sql=item['SQL'],
            eval_query="",
            setup_sql="",
            cleanup_sql="",
            tags=[item["difficulty"]],
            other={}
        )
        input_items[eval_input.query_type].append(eval_input)
    return input_items