import gradio as gr
import pandas as pd
import requests
from io import BytesIO

def convert_hf_dataset(file_url: str):
    file_url = file_url.strip()
    # Check that the URL is from Hugging Face
    if "huggingface.co" not in file_url:
        raise ValueError("Please provide a URL from Hugging Face datasets.")
    
    # Ensure the URL has a scheme; if not, add "https://"
    if not file_url.lower().startswith(("http://", "https://")):
        file_url = "https://" + file_url

    # Download the content from the URL
    response = requests.get(file_url)
    response.raise_for_status()
    content = response.content

    # Determine file type from URL extension and convert accordingly
    if file_url.lower().endswith(".csv"):
        # If it's a CSV, read it and convert to Parquet
        df = pd.read_csv(BytesIO(content))
        output_file = "output.parquet"
        df.to_parquet(output_file, index=False)
        converted_format = "Parquet"
    elif file_url.lower().endswith(".parquet"):
        # If it's a Parquet file, read it and convert to CSV
        df = pd.read_parquet(BytesIO(content))
        output_file = "output.csv"
        df.to_csv(output_file, index=False)
        converted_format = "CSV"
    else:
        raise ValueError("The URL must point to a .csv or .parquet file.")
    
    # Create a preview of the top 10 rows
    preview = df.head(10).to_string(index=False)
    info_message = (
        f"Input file: {file_url.split('/')[-1]}\n"
        f"Converted file format: {converted_format}\n\n"
        f"Preview (Top 10 Rows):\n{preview}"
    )
    
    return output_file, info_message

demo = gr.Interface(
    fn=convert_hf_dataset,
    inputs=gr.Textbox(
        label="Hugging Face Dataset URL", 
        placeholder="e.g., huggingface.co/datasets/username/dataset/filename.csv"
    ),
    outputs=[
        gr.File(label="Converted File"),
        gr.Textbox(label="Preview (Top 10 Rows)", lines=15)
    ],
    title="Hugging Face CSV <-> Parquet Converter",
    description=(
        "Enter the URL of a Hugging Face dataset file (must end with .csv or .parquet). "
        "The app will automatically detect the file type, convert it to the opposite format, "
        "and display a preview of the top 10 rows."
    )
)

if __name__ == "__main__":
    demo.launch()