"""
================================================================================
Text Classification Dataset - Model Training Script
================================================================================
Project: Text Classification Dataset
Category: Text Data / NLP

Author: Molla Samser
Designer & Tester: Rima Khatun
Website: https://rskworld.in
Email: help@rskworld.in | support@rskworld.in
Phone: +91 93305 39277

Copyright (c) 2026 RSK World - All Rights Reserved
Content used for educational purposes only.

Created: December 2026
================================================================================
"""

import os
import json
import argparse
from datetime import datetime
from typing import Dict, List, Tuple, Optional

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.linear_model import LogisticRegression
from sklearn.svm import LinearSVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import (
    accuracy_score, precision_score, recall_score, f1_score,
    classification_report, confusion_matrix
)
from sklearn.pipeline import Pipeline
import joblib

# Project information
__author__ = "Molla Samser"
__website__ = "https://rskworld.in"
__email__ = "help@rskworld.in"


class TextClassifier:
    """
    A flexible text classification system supporting multiple algorithms.
    
    Supported algorithms:
    - Naive Bayes (MultinomialNB)
    - Logistic Regression
    - Support Vector Machine (LinearSVC)
    - Random Forest
    
    Author: Molla Samser | RSK World (https://rskworld.in)
    """
    
    ALGORITHMS = {
        'naive_bayes': MultinomialNB,
        'logistic_regression': LogisticRegression,
        'svm': LinearSVC,
        'random_forest': RandomForestClassifier
    }
    
    CATEGORIES = {
        0: 'Technology',
        1: 'Sports',
        2: 'Politics',
        3: 'Entertainment',
        4: 'Business',
        5: 'Science'
    }
    
    def __init__(
        self,
        algorithm: str = 'logistic_regression',
        max_features: int = 10000,
        ngram_range: Tuple[int, int] = (1, 2),
        random_state: int = 42
    ):
        """
        Initialize the TextClassifier.
        
        Args:
            algorithm: Classification algorithm to use
            max_features: Maximum number of TF-IDF features
            ngram_range: Range of n-grams for feature extraction
            random_state: Random seed for reproducibility
        """
        self.algorithm = algorithm
        self.max_features = max_features
        self.ngram_range = ngram_range
        self.random_state = random_state
        
        # Validate algorithm
        if algorithm not in self.ALGORITHMS:
            raise ValueError(
                f"Unknown algorithm: {algorithm}. "
                f"Available: {list(self.ALGORITHMS.keys())}"
            )
        
        # Initialize vectorizer
        self.vectorizer = TfidfVectorizer(
            max_features=max_features,
            ngram_range=ngram_range,
            stop_words='english',
            sublinear_tf=True
        )
        
        # Initialize classifier
        if algorithm == 'logistic_regression':
            self.classifier = LogisticRegression(
                max_iter=1000,
                random_state=random_state,
                class_weight='balanced'
            )
        elif algorithm == 'svm':
            self.classifier = LinearSVC(
                max_iter=1000,
                random_state=random_state,
                class_weight='balanced'
            )
        elif algorithm == 'random_forest':
            self.classifier = RandomForestClassifier(
                n_estimators=100,
                random_state=random_state,
                class_weight='balanced',
                n_jobs=-1
            )
        else:
            self.classifier = MultinomialNB()
        
        # Create pipeline
        self.pipeline = Pipeline([
            ('vectorizer', self.vectorizer),
            ('classifier', self.classifier)
        ])
        
        self.is_trained = False
        self.training_history = {}
    
    def train(
        self,
        X_train: List[str],
        y_train: List[int],
        X_val: Optional[List[str]] = None,
        y_val: Optional[List[int]] = None
    ) -> Dict:
        """
        Train the classification model.
        
        Args:
            X_train: Training texts
            y_train: Training labels
            X_val: Validation texts (optional)
            y_val: Validation labels (optional)
            
        Returns:
            Dictionary containing training metrics
        """
        print(f"\n{'='*60}")
        print(f"Training Text Classifier - RSK World")
        print(f"Algorithm: {self.algorithm}")
        print(f"Author: {__author__} | Website: {__website__}")
        print(f"{'='*60}\n")
        
        # Train the pipeline
        print(f"Training on {len(X_train)} samples...")
        self.pipeline.fit(X_train, y_train)
        self.is_trained = True
        
        # Calculate training metrics
        train_preds = self.pipeline.predict(X_train)
        train_accuracy = accuracy_score(y_train, train_preds)
        
        results = {
            'algorithm': self.algorithm,
            'train_samples': len(X_train),
            'train_accuracy': train_accuracy,
            'timestamp': datetime.now().isoformat()
        }
        
        print(f"Training Accuracy: {train_accuracy:.4f}")
        
        # Validation metrics if provided
        if X_val is not None and y_val is not None:
            val_preds = self.pipeline.predict(X_val)
            val_accuracy = accuracy_score(y_val, val_preds)
            val_f1 = f1_score(y_val, val_preds, average='weighted')
            
            results['val_samples'] = len(X_val)
            results['val_accuracy'] = val_accuracy
            results['val_f1'] = val_f1
            
            print(f"Validation Accuracy: {val_accuracy:.4f}")
            print(f"Validation F1 Score: {val_f1:.4f}")
        
        self.training_history = results
        return results
    
    def evaluate(self, X_test: List[str], y_test: List[int]) -> Dict:
        """
        Evaluate the model on test data.
        
        Args:
            X_test: Test texts
            y_test: Test labels
            
        Returns:
            Dictionary containing evaluation metrics
        """
        if not self.is_trained:
            raise RuntimeError("Model must be trained before evaluation")
        
        print(f"\n{'='*60}")
        print("Model Evaluation")
        print(f"{'='*60}\n")
        
        # Generate predictions
        predictions = self.pipeline.predict(X_test)
        
        # Calculate metrics
        accuracy = accuracy_score(y_test, predictions)
        precision = precision_score(y_test, predictions, average='weighted')
        recall = recall_score(y_test, predictions, average='weighted')
        f1 = f1_score(y_test, predictions, average='weighted')
        
        # Classification report
        report = classification_report(
            y_test, predictions,
            target_names=list(self.CATEGORIES.values()),
            output_dict=True
        )
        
        # Confusion matrix
        conf_matrix = confusion_matrix(y_test, predictions)
        
        results = {
            'accuracy': accuracy,
            'precision': precision,
            'recall': recall,
            'f1_score': f1,
            'classification_report': report,
            'confusion_matrix': conf_matrix.tolist()
        }
        
        # Print results
        print(f"Test Accuracy:  {accuracy:.4f}")
        print(f"Test Precision: {precision:.4f}")
        print(f"Test Recall:    {recall:.4f}")
        print(f"Test F1 Score:  {f1:.4f}")
        print(f"\nClassification Report:")
        print(classification_report(
            y_test, predictions,
            target_names=list(self.CATEGORIES.values())
        ))
        
        return results
    
    def predict(self, texts: List[str]) -> List[int]:
        """
        Predict categories for new texts.
        
        Args:
            texts: List of texts to classify
            
        Returns:
            List of predicted labels
        """
        if not self.is_trained:
            raise RuntimeError("Model must be trained before prediction")
        
        return self.pipeline.predict(texts)
    
    def predict_with_labels(self, texts: List[str]) -> List[Dict]:
        """
        Predict categories with human-readable labels.
        
        Args:
            texts: List of texts to classify
            
        Returns:
            List of dictionaries with predictions
        """
        predictions = self.predict(texts)
        results = []
        
        for text, pred in zip(texts, predictions):
            results.append({
                'text': text[:100] + '...' if len(text) > 100 else text,
                'predicted_label': int(pred),
                'predicted_category': self.CATEGORIES[pred]
            })
        
        return results
    
    def save_model(self, filepath: str):
        """Save the trained model to disk."""
        if not self.is_trained:
            raise RuntimeError("Model must be trained before saving")
        
        model_data = {
            'pipeline': self.pipeline,
            'algorithm': self.algorithm,
            'max_features': self.max_features,
            'ngram_range': self.ngram_range,
            'training_history': self.training_history,
            'metadata': {
                'author': __author__,
                'website': __website__,
                'email': __email__,
                'saved_at': datetime.now().isoformat()
            }
        }
        
        joblib.dump(model_data, filepath)
        print(f"Model saved to: {filepath}")
    
    @classmethod
    def load_model(cls, filepath: str) -> 'TextClassifier':
        """Load a trained model from disk."""
        model_data = joblib.load(filepath)
        
        classifier = cls(
            algorithm=model_data['algorithm'],
            max_features=model_data['max_features'],
            ngram_range=model_data['ngram_range']
        )
        
        classifier.pipeline = model_data['pipeline']
        classifier.is_trained = True
        classifier.training_history = model_data.get('training_history', {})
        
        print(f"Model loaded from: {filepath}")
        return classifier


def load_dataset(data_dir: str) -> Tuple[pd.DataFrame, pd.DataFrame, pd.DataFrame]:
    """
    Load train, validation, and test datasets.
    
    Args:
        data_dir: Path to the data directory
        
    Returns:
        Tuple of (train_df, val_df, test_df)
    """
    train_path = os.path.join(data_dir, 'csv', 'train.csv')
    val_path = os.path.join(data_dir, 'csv', 'validation.csv')
    test_path = os.path.join(data_dir, 'csv', 'test.csv')
    
    train_df = pd.read_csv(train_path, comment='#')
    val_df = pd.read_csv(val_path, comment='#')
    test_df = pd.read_csv(test_path, comment='#')
    
    return train_df, val_df, test_df


def main():
    """Main training script."""
    parser = argparse.ArgumentParser(
        description='Train Text Classification Model - RSK World'
    )
    parser.add_argument(
        '--data-dir', type=str, default='../data',
        help='Path to data directory'
    )
    parser.add_argument(
        '--algorithm', type=str, default='logistic_regression',
        choices=['naive_bayes', 'logistic_regression', 'svm', 'random_forest'],
        help='Classification algorithm'
    )
    parser.add_argument(
        '--max-features', type=int, default=10000,
        help='Maximum TF-IDF features'
    )
    parser.add_argument(
        '--output', type=str, default='model.joblib',
        help='Output model path'
    )
    
    args = parser.parse_args()
    
    print(f"\n{'='*60}")
    print("Text Classification Training - RSK World")
    print(f"Author: {__author__}")
    print(f"Website: {__website__}")
    print(f"Email: {__email__}")
    print(f"{'='*60}\n")
    
    # Load data
    print("Loading dataset...")
    train_df, val_df, test_df = load_dataset(args.data_dir)
    
    print(f"Train samples: {len(train_df)}")
    print(f"Validation samples: {len(val_df)}")
    print(f"Test samples: {len(test_df)}")
    
    # Initialize classifier
    classifier = TextClassifier(
        algorithm=args.algorithm,
        max_features=args.max_features
    )
    
    # Train model
    classifier.train(
        X_train=train_df['text'].tolist(),
        y_train=train_df['label'].tolist(),
        X_val=val_df['text'].tolist(),
        y_val=val_df['label'].tolist()
    )
    
    # Evaluate on test set
    classifier.evaluate(
        X_test=test_df['text'].tolist(),
        y_test=test_df['label'].tolist()
    )
    
    # Save model
    classifier.save_model(args.output)
    
    # Demo predictions
    print(f"\n{'='*60}")
    print("Demo Predictions")
    print(f"{'='*60}\n")
    
    demo_texts = [
        "New smartphone features revolutionary camera technology.",
        "Team wins championship in overtime thriller.",
        "Government passes new legislation on climate change."
    ]
    
    predictions = classifier.predict_with_labels(demo_texts)
    for pred in predictions:
        print(f"Text: {pred['text']}")
        print(f"Predicted: {pred['predicted_category']}")
        print()


if __name__ == "__main__":
    main()