"""Skyulf EDA Example: Time Series & Geospatial Analysis
===================================================

This script demonstrates how to use Skyulf's EDA module for:
- Time Series Analysis (Trend, Seasonality)
- Geospatial Analysis (Lat/Lon distribution)
- Manual column specification for analysis

Requirements:
    pip install skyulf-core[viz] polars numpy
"""

from datetime import datetime, timedelta

import numpy as np
import polars as pl
from rich.console import Console
from rich.panel import Panel

from skyulf import EDAAnalyzer, EDAVisualizer

console = Console()


def generate_synthetic_data(n_rows=1000):
    """Generates a synthetic dataset with time series and geospatial patterns."""
    np.random.seed(42)

    # Time Series: 1 year of data
    start_date = datetime(2023, 1, 1)
    dates = [start_date + timedelta(days=np.random.randint(0, 365)) for _ in range(n_rows)]
    dates.sort()

    # Geospatial: Clusters around NYC and LA
    # NYC: 40.7128° N, 74.0060° W
    # LA: 34.0522° N, 118.2437° W
    lats = []
    lons = []
    cities = []

    for _ in range(n_rows):
        if np.random.random() > 0.5:
            # NYC
            lats.append(40.7128 + np.random.normal(0, 0.1))
            lons.append(-74.0060 + np.random.normal(0, 0.1))
            cities.append("NYC")
        else:
            # LA
            lats.append(34.0522 + np.random.normal(0, 0.1))
            lons.append(-118.2437 + np.random.normal(0, 0.1))
            cities.append("LA")

    # Numeric Value with Trend + Seasonality
    # Value increases over time + sine wave
    values = []
    for d in dates:
        day_of_year = d.timetuple().tm_yday
        trend = day_of_year * 0.1
        seasonality = 10 * np.sin(2 * np.pi * day_of_year / 365)
        noise = np.random.normal(0, 5)
        values.append(trend + seasonality + noise)

    return pl.DataFrame(
        {
            "timestamp": dates,
            "latitude": lats,
            "longitude": lons,
            "city": cities,
            "sales": values,
            "category": np.random.choice(["A", "B", "C"], n_rows),
        }
    )


def main():
    console.print(Panel.fit("Skyulf Advanced EDA: Time Series & Geospatial", style="bold purple"))

    # 1. Generate Data
    with console.status("[bold green]Generating synthetic data..."):
        df = generate_synthetic_data(2000)

    console.print(f"Generated dataset: {df.shape[0]} rows, {df.shape[1]} columns")
    console.print(df.head())

    # 2. Run Analysis
    # We explicitly specify the columns to ensure they are picked up correctly,
    # although Skyulf can often auto-detect them.
    with console.status("[bold green]Running analysis..."):
        analyzer = EDAAnalyzer(df)
        profile = analyzer.analyze(
            target_col="sales", date_col="timestamp", lat_col="latitude", lon_col="longitude"
        )

    # 3. Visualize Results
    viz = EDAVisualizer(profile, df)

    # Print rich terminal dashboard
    viz.summary()

    # Show plots (Time Series Trend, Geospatial Map, etc.)
    viz.plot()


if __name__ == "__main__":
    main()