import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import os

# Download data using the terminal (e.g., curl or wget)
# Replace 'your-api-key' with an actual API key from a data provider
# Here, we simulate this with a placeholder for clarity
api_key = "your-api-key"
symbol = "EURUSD"
output_csv = "EURUSD_3_years.csv"

# Command to download the data from Alpha Vantage or any similar service
# Example using Alpha Vantage (Daily FX data): https://www.alphavantage.co
command = f"curl -o {output_csv} 'https://www.alphavantage.co/query?function=FX_DAILY&from_symbol=EUR&to_symbol=USD&outputsize=full&apikey={api_key}&datatype=csv'"
os.system(command)

# Read the downloaded CSV file
data = pd.read_csv(output_csv)

# Ensure the CSV is structured correctly for further processing
# Rename columns if necessary to match yfinance format
data.rename(columns={"close": "Close"}, inplace=True)

# Print the first few rows to confirm
print(data.head())

# Extract the 'Close' prices from the data
prices = data['Close'].values

# Normalize the prices for generating synthetic data
min_price = prices.min()
max_price = prices.max()
normalized_prices = (prices - min_price) / (max_price - min_price)

# Example: Generating some mock data
def generate_real_data(samples=100):
    # Real data following a sine wave pattern
    time = np.linspace(0, 4 * np.pi, samples)
    data = np.sin(time) + np.random.normal(0, 0.1, samples)  # Add some noise
    return time, data

def generate_fake_data(generator, samples=100):
    # Fake data generated by the GAN
    noise = np.random.normal(0, 1, (samples, 1))
    generated_data = generator.predict(noise).flatten()
    return generated_data

# Mock generator function (replace with actual GAN generator model)
class MockGenerator:
    def predict(self, noise):
        # Simulate GAN output with a cosine pattern (for illustration)
        return np.cos(np.linspace(0, 4 * np.pi, len(noise))).reshape(-1, 1)

# Instantiate a mock generator for demonstration
generator = MockGenerator()

# Generate synthetic data: Let's use a simple random walk model as a basic example
# (this is a placeholder for a more sophisticated method, like using GANs)
np.random.seed(42)  # Set seed for reproducibility
synthetic_prices_normalized = normalized_prices[0] + np.cumsum(np.random.normal(0, 0.01, len(prices)))

# Denormalize the synthetic prices back to the original scale
synthetic_prices = synthetic_prices_normalized * (max_price - min_price) + min_price

# Configure font sizes
plt.rcParams.update({
    'font.size': 12,       # General font size
    'axes.titlesize': 16,  # Title font size
    'axes.labelsize': 14,  # Axis labels font size
    'legend.fontsize': 12, # Legend font size
    'xtick.labelsize': 10, # X-axis tick labels font size
    'ytick.labelsize': 10  # Y-axis tick labels font size
})

# Plot both historical and synthetic data on the same graph
plt.figure(figsize=(14, 7))
plt.plot(prices, label="Historical EURUSD", color='blue')
plt.plot(synthetic_prices, label="Synthetic EURUSD", linestyle="--", color='red')
plt.xlabel("Time Steps", fontsize=14)  # Adjust fontsize directly if needed
plt.ylabel("Price", fontsize=14)       # Adjust fontsize directly if needed
plt.title("Comparison of Historical and Synthetic EURUSD Data", fontsize=16)
plt.legend()
plt.show()