import json
import datasets


_DESCRIPTION = """\
This is a Japanese translated version of HumanEval, an evaluation harness for the HumanEval problem solving dataset described in the paper "Evaluating Large Language Models Trained on Code".
"""

_URL = "https://raw.githubusercontent.com/KuramitsuLab/jhuman-eval/main/data/jhuman-eval.jsonl.gz"

_LICENSE = "MIT"


class JHumaneval(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("0.1.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="jhumaneval",
            version=VERSION, 
            description=_DESCRIPTION,
        )
    ]

    def _info(self):
        features = datasets.Features(
            {
                "task_id": datasets.Value("string"),
                "prompt_en": datasets.Value("string"),
                "prompt": datasets.Value("string"),
                "entry_point": datasets.Value("string"),
                "canonical_solution": datasets.Value("string"),
                "test": datasets.Value("string"),
            }
        )

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            license=_LICENSE,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        data_dir = dl_manager.download_and_extract(_URL)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "filepath": data_dir, 
                },
            )
        ]

    def _generate_examples(self, filepath):
        """Yields examples."""
        with open(filepath, encoding="utf-8") as file:
            data = [json.loads(line) for line in file]
            id_ = 0
            for sample in data:
                yield id_, sample
                id_ += 1