from datasets import DatasetInfo, GeneratorBasedBuilder, SplitGenerator, Split, Value, Features
import datasets
import csv

class FiveHTKiPrediction(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    def _info(self):
        return DatasetInfo(
            description="Curated dataset of serotonin receptor Ki values from PDSP Ki Database.",
            features=Features({
                "smiles": Value("string"),
                "ki": Value("float"),
                "receptor": Value("string"),
                "source": Value("string"),
            }),
            supervised_keys=None,
        )

    def _split_generators(self, dl_manager):
        data_path = dl_manager.download_and_extract("data/KiDatabase.csv")
        return [SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": data_path})]

    def _generate_examples(self, filepath):
        with open(filepath, encoding="utf-8") as f:
            reader = csv.DictReader(f)
            for idx, row in enumerate(reader):
                try:
                    yield idx, {
                        "smiles": row["SMILES"].strip(),
                        "ki": float(row["ki Val"]) if row["ki Val"] else None,
                        "receptor": row["Name"].strip(),
                        "source": row["source"].strip(),
                    }
                except Exception:
                    continue