"""African Music Dataset loading script for Hugging Face Datasets."""

import datasets
from pathlib import Path


_CITATION = """\
@dataset{spotify_africa_dataset_2025,
  title={Spotify-Africa Music Dataset: A Comprehensive Collection of African Music Metadata},
  author={Spotify-Africa Dataset Project},
  year={2025},
  publisher={Hugging Face},
}
"""

_DESCRIPTION = """\
A comprehensive dataset documenting African music from Spotify spanning 1,600+ tracks, 
650+ artists, and 67 years of musical history (1958-2025). Includes track metadata, 
artist information, temporal trends, regional summaries, and network relationships.
"""

_HOMEPAGE = "https://huggingface.co/datasets/electricsheepafrica/Spotify-Africa-Dataset"

_LICENSE = "cc-by-4.0"

_URLS = {
    "master_tracks": "data/datasets/master_tracks/*.parquet",
    "analysis_ready_tracks": "data/datasets/analysis_ready_tracks/*.parquet",
    "scaled_tracks": "data/datasets/scaled_tracks/*.parquet",
    "comprehensive_tracks": "data/datasets/comprehensive_tracks/*.parquet",
    "popular_tracks": "data/datasets/popular_tracks/*.parquet",
    "enriched_tracks": "data/datasets/enriched_tracks/*.parquet",
    "enriched_artist_summary": "data/datasets/enriched_artist_summary/*.parquet",
    "enriched_region_summary": "data/datasets/enriched_region_summary/*.parquet",
    "analysis_ready_artists": "data/datasets/analysis_ready_artists/*.parquet",
    "popular_artists": "data/datasets/popular_artists/*.parquet",
    "artist_summary": "data/datasets/artist_summary/*.parquet",
    "genre_analysis": "data/datasets/genre_analysis/*.parquet",
    "ml_training_popular": "data/datasets/ml_training_popular/*.parquet",
    "temporal_analysis": "data/datasets/temporal_analysis/*.parquet",
    "temporal_trends": "data/datasets/temporal_trends/*.parquet",
}


class AfricanMusicDataset(datasets.GeneratorBasedBuilder):
    """African Music Dataset from Spotify metadata."""

    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="master_tracks",
            version=VERSION,
            description="Unified dataset merging all collections with enriched features (1,217 tracks)",
        ),
        datasets.BuilderConfig(
            name="analysis_ready_tracks",
            version=VERSION,
            description="Clean, high-quality subset from top 30 artists (155 tracks)",
        ),
        datasets.BuilderConfig(
            name="scaled_tracks",
            version=VERSION,
            description="Large-scale collection via genre/market searches (979 tracks)",
        ),
        datasets.BuilderConfig(
            name="comprehensive_tracks",
            version=VERSION,
            description="Regional diversity focus (355 tracks)",
        ),
        datasets.BuilderConfig(
            name="popular_tracks",
            version=VERSION,
            description="Top tracks from leading artists (100 tracks)",
        ),
        datasets.BuilderConfig(
            name="enriched_tracks",
            version=VERSION,
            description="Tracks with regional, temporal, and popularity annotations",
        ),
        datasets.BuilderConfig(
            name="enriched_artist_summary",
            version=VERSION,
            description="Artist-level aggregations with hit ratios and recency",
        ),
        datasets.BuilderConfig(
            name="enriched_region_summary",
            version=VERSION,
            description="Regional roll-ups with volume and popularity metrics",
        ),
        datasets.BuilderConfig(
            name="analysis_ready_artists",
            version=VERSION,
            description="Artist metadata for top-tier acts",
        ),
        datasets.BuilderConfig(
            name="popular_artists",
            version=VERSION,
            description="Follower and popularity data for influential artists",
        ),
        datasets.BuilderConfig(
            name="artist_summary",
            version=VERSION,
            description="Legacy artist aggregations",
        ),
        datasets.BuilderConfig(
            name="genre_analysis",
            version=VERSION,
            description="Genre-tagged subset for classification tasks",
        ),
        datasets.BuilderConfig(
            name="ml_training_popular",
            version=VERSION,
            description="High-popularity tracks for supervised learning",
        ),
        datasets.BuilderConfig(
            name="temporal_analysis",
            version=VERSION,
            description="Year-level aggregations for trend studies",
        ),
        datasets.BuilderConfig(
            name="temporal_trends",
            version=VERSION,
            description="Time-series data from scaled collection",
        ),
    ]

    DEFAULT_CONFIG_NAME = "master_tracks"

    def _info(self):
        # Define features based on config
        if "artist" in self.config.name and "track" not in self.config.name:
            features = datasets.Features({
                "artist_id": datasets.Value("string"),
                "artist_name": datasets.Value("string"),
            })
        elif "region" in self.config.name:
            features = datasets.Features({
                "region": datasets.Value("string"),
            })
        elif "temporal" in self.config.name:
            features = datasets.Features({
                "release_year": datasets.Value("int64"),
            })
        else:
            # Track-level features
            features = datasets.Features({
                "track_id": datasets.Value("string"),
                "track_name": datasets.Value("string"),
                "artist_id": datasets.Value("string"),
                "artist_name": datasets.Value("string"),
                "album_id": datasets.Value("string"),
                "album_name": datasets.Value("string"),
                "release_date": datasets.Value("string"),
                "popularity": datasets.Value("int64"),
            })

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        config_name = self.config.name
        data_pattern = _URLS.get(config_name, _URLS["master_tracks"])
        
        # Download/locate files
        data_dir = Path(data_pattern).parent
        files = list(Path(dl_manager.download_config.cache_dir).parent.glob(data_pattern))
        
        if not files:
            # Fallback to local path
            files = list(Path(data_pattern).parent.glob("*.parquet"))

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepaths": files},
            ),
        ]

    def _generate_examples(self, filepaths):
        """Yields examples from Parquet files."""
        import pyarrow.parquet as pq
        
        idx = 0
        for filepath in filepaths:
            table = pq.read_table(filepath)
            df = table.to_pandas()
            
            for _, row in df.iterrows():
                yield idx, row.to_dict()
                idx += 1