import pandas as pd
import datasets
import numpy as np
from PIL import Image

class InvasivePlantsConfig(datasets.BuilderConfig):

    def __init__(self, **kwargs):
        super(InvasivePlantsConfig, self).__init__(**kwargs)
        
class InvasivePlantsDataset(datasets.GeneratorBasedBuilder):

    BUILDER_CONFIGS = [
        InvasivePlantsConfig(
            name="full",
            version=datasets.Version("1.0.0"),
            description="Full dataset, combining both systematically and opportunistically sampled leaves"
        ),
        InvasivePlantsConfig(
            name="systematic",
            version=datasets.Version("1.0.0"),
            description="Subset containing only systematically sampled leaves"
        ),
        InvasivePlantsConfig(
            name="opportunistic",
            version=datasets.Version("1.0.0"),
            description="Subset containing only opportunistically sampled leaves"
        )
    ]

    DEFAULT_CONFIG_NAME = "full"

    def _info(self):
        #expert_other_insect,expert_mechanical_damage,expert_confidence,expert_other_remarks,expert_notes,sampling_type
        return datasets.DatasetInfo(
            description="Invasive Plants Datasets",
            features=datasets.Features(
                {
                    "session": datasets.Value("string"),
                    "filename": datasets.Value("string"),
                    "sampling_type" : datasets.Value("string"),
                    "image": datasets.Image(),
                    "site": datasets.Value("int32"),
                    "day": datasets.Value("int32"),
                    "plant": datasets.Value("string"),
                    "level": datasets.Value("string"),
                    "view": datasets.Value("string"),
                    "healthy": datasets.Value("string"),
                    "rust" : datasets.Value("string"),
                    "leaf_miner" : datasets.Value("string"),
                    "other_insect" : datasets.Value("string"), 
                    "mechanical_damage" : datasets.Value("string"),
                    "other_remarks" : datasets.Value("string"),
                    "expert_healthy" : datasets.Value("string"),
                    "expert_rust" : datasets.Value("string"),
                    "expert_leaf_miner" : datasets.Value("string"),
                    "expert_other_insect" : datasets.Value("string"),
                    "expert_mechanical_damage" : datasets.Value("string"),
                    "expert_confidence" : datasets.Value("string"),
                    "expert_other_remarks" : datasets.Value("string"),
                    "expert_notes" : datasets.Value("string")
                }
            ),

            supervised_keys=None,
            homepage="https://huggingface.co/datasets/imageomics/invasive_plants_hawaii",
            license="MIT",
        )
    
    def _split_generators(self, dl_manager):
        data_dir = "images"
        if self.config.name == "full":
            full_csv = "metadata/full_dataset.csv"
            dorsal_csv = "metadata/full_dataset_dorsal.csv" 
            ventral_csv = "metadata/full_dataset_ventral.csv"
        elif self.config.name == "systematic":
            full_csv = "metadata/full_systematic_dataset.csv"
            dorsal_csv = "metadata/systematic_dataset_dorsal.csv" 
            ventral_csv = "metadata/systematic_dataset_ventral.csv"
        elif self.config.name == "opportunistic":
            full_csv = "metadata/full_opportunistic_dataset.csv"
            dorsal_csv = "metadata/opportunistic_dataset_dorsal.csv" 
            ventral_csv = "metadata/opportunistic_dataset_ventral.csv"
        else:
            raise ValueError(f"Unknown config: {self.config.name}")

        return [
            datasets.SplitGenerator(
                name="dorsal",
                gen_kwargs={
                    "metadata_path": dorsal_csv,
                },
            ),
            datasets.SplitGenerator(
                name="ventral",
                gen_kwargs={
                    "metadata_path": ventral_csv,
                },
            ),
            datasets.SplitGenerator(
                name="both",
                gen_kwargs={
                    "metadata_path": full_csv,
                },
            ),
        ]
    
    def _generate_examples(self, metadata_path):
        img_dir = "images"
        df_metadata = pd.read_csv(metadata_path)
        for idx in range(len(df_metadata)):
            row = df_metadata.iloc[idx]
            img_path = f"{img_dir}/{row["session"]}/{row["filename"]}"
            img_pillow = Image.open(img_path)

            yield idx, {
                    "session": row["session"],
                    "filename": row["filename"],
                    "sampling_type" : row["sampling_type"],
                    "image": img_pillow,
                    "site": row["site"],
                    "day": row["day"],
                    "plant": row["plant"],
                    "level": row["level"],
                    "view": row["view"],
                    "healthy": row["healthy"],
                    "rust" : row["rust"],
                    "leaf_miner" : row["leaf_miner"],
                    "other_insect" : row["other_insect"], 
                    "mechanical_damage" : row["mechanical_damage"],
                    "other_remarks" : row["other_remarks"],
                    "expert_healthy" : row["expert_healthy"],
                    "expert_rust" : row["expert_rust"],
                    "expert_leaf_miner" : row["expert_leaf_miner"],
                    "expert_other_insect" : row["expert_other_insect"],
                    "expert_mechanical_damage" : row["expert_mechanical_damage"],
                    "expert_confidence" : row["expert_confidence"],
                    "expert_other_remarks" : row["expert_other_remarks"],
                    "expert_notes" : row["expert_notes"]
                }