"""
Unified News Scraper & Sentiment Analysis Application
Combines scraping, processing, and visualization in one interface
Modified for Hugging Face Spaces - uses /tmp directory
"""

import streamlit as st
import pandas as pd
import matplotlib.pyplot as plt
import plotly.express as px
import plotly.graph_objects as go
from matplotlib.ticker import MaxNLocator
import subprocess
import sys
import os
from pathlib import Path
import time
from datetime import datetime
import warnings
import json
import requests
import spacy

warnings.filterwarnings('ignore')

# Constants
INDIA_GEOJSON_URL = 'https://raw.githubusercontent.com/geohacker/india/master/state/india_state.geojson'
BASE_DIR = Path('/tmp/news_scraper')

# Page config
st.set_page_config(
    page_title="News Scraper & Analysis Platform",
    page_icon="📰",
    layout="wide",
    initial_sidebar_state="expanded"
)

# Custom CSS
def load_css():
    st.markdown("""
    <style>
        .main-header {
            font-size: 2.8rem !important;
            font-weight: 700 !important;
            text-align: center !important;
            padding: 1.5rem 0 !important;
            background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
            -webkit-background-clip: text;
            -webkit-text-fill-color: transparent;
        }
        .sub-header {
            font-size: 1.8rem !important;
            font-weight: 600 !important;
            padding: 1rem 0 !important;
            color: #2c3e50;
        }
        .feature-card {
            background: white;
            border-radius: 15px;
            padding: 25px;
            box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
            margin: 10px 0;
            transition: transform 0.3s;
        }
        .feature-card:hover {
            transform: translateY(-5px);
            box-shadow: 0 8px 12px rgba(0, 0, 0, 0.15);
        }
        .metric-card {
            background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
            border-radius: 10px;
            padding: 20px;
            color: white;
            text-align: center;
        }
        .status-running {
            background-color: #fff3cd;
            border-left: 4px solid #ffc107;
            padding: 15px;
            border-radius: 5px;
        }
        .status-success {
            background-color: #d4edda;
            border-left: 4px solid #28a745;
            padding: 15px;
            border-radius: 5px;
        }
        .status-error {
            background-color: #f8d7da;
            border-left: 4px solid #dc3545;
            padding: 15px;
            border-radius: 5px;
        }
    </style>
    """, unsafe_allow_html=True)

# Initialize session state
def init_session_state():
    defaults = {
        'scraped_data': {},
        'scraping_active': False,
        'processing_status': {},
        'selected_dataset': None,
        'base_dir': BASE_DIR
    }
    for key, value in defaults.items():
        if key not in st.session_state:
            st.session_state[key] = value

# Setup directories
def setup_directories():
    """Create necessary directories in /tmp"""
    try:
        for dir_name in ['output', 'data', 'temp']:
            dir_path = BASE_DIR / dir_name
            dir_path.mkdir(parents=True, exist_ok=True)
        return True
    except Exception as e:
        st.error(f"❌ Directory setup error: {e}")
        return False

# Load India GeoJSON
@st.cache_data
def load_india_geojson():
    """Load India GeoJSON data for mapping"""
    try:
        response = requests.get(INDIA_GEOJSON_URL, timeout=10)
        return json.loads(response.text)
    except Exception as e:
        st.warning(f"Could not load India map: {e}")
        return None

# Load spaCy model
@st.cache_resource
def load_spacy_model():
    try:
        return spacy.load("en_core_web_sm")
    except OSError:
        st.info("Downloading spaCy model...")
        subprocess.call([sys.executable, "-m", "spacy", "download", "en_core_web_sm"])
        return spacy.load("en_core_web_sm")

# State mapping
def get_state_mapping():
    return {
        'andhra pradesh': 'Andhra Pradesh', 'arunachal pradesh': 'Arunachal Pradesh',
        'assam': 'Assam', 'bihar': 'Bihar', 'chhattisgarh': 'Chhattisgarh',
        'goa': 'Goa', 'gujarat': 'Gujarat', 'haryana': 'Haryana',
        'himachal pradesh': 'Himachal Pradesh', 'jharkhand': 'Jharkhand',
        'karnataka': 'Karnataka', 'kerala': 'Kerala', 'madhya pradesh': 'Madhya Pradesh',
        'maharashtra': 'Maharashtra', 'manipur': 'Manipur', 'meghalaya': 'Meghalaya',
        'mizoram': 'Mizoram', 'nagaland': 'Nagaland', 'odisha': 'Odisha',
        'punjab': 'Punjab', 'rajasthan': 'Rajasthan', 'sikkim': 'Sikkim',
        'tamil nadu': 'Tamil Nadu', 'telangana': 'Telangana', 'tripura': 'Tripura',
        'uttar pradesh': 'Uttar Pradesh', 'uttarakhand': 'Uttarakhand',
        'west bengal': 'West Bengal', 'delhi': 'Delhi', 'new delhi': 'Delhi',
        'jammu and kashmir': 'Jammu and Kashmir', 'j&k': 'Jammu and Kashmir',
        'ladakh': 'Ladakh', 'chandigarh': 'Chandigarh', 'puducherry': 'Puducherry',
        'mumbai': 'Maharashtra', 'kolkata': 'West Bengal', 'chennai': 'Tamil Nadu',
        'bangalore': 'Karnataka', 'bengaluru': 'Karnataka', 'hyderabad': 'Telangana',
        'ahmedabad': 'Gujarat', 'pune': 'Maharashtra', 'jaipur': 'Rajasthan',
    }

# Extract locations from text
@st.cache_data
def extract_locations_from_descriptions(df, description_column='desc'):
    """Extract state names from description using spaCy"""
    nlp = load_spacy_model()
    state_mapping = get_state_mapping()
    
    locations = []
    progress_bar = st.progress(0)
    
    for idx, (_, row) in enumerate(df.iterrows()):
        if idx % 100 == 0:
            progress_bar.progress(min(idx / len(df), 1.0))
        
        if pd.isna(row.get(description_column, None)):
            locations.append(None)
            continue
        
        description = str(row[description_column]).lower()
        doc = nlp(description)
        
        found_locations = []
        for ent in doc.ents:
            if ent.label_ in ["GPE", "LOC"]:
                loc_name = ent.text.lower()
                if loc_name in state_mapping:
                    found_locations.append(state_mapping[loc_name])
        
        for state_var, standard_name in state_mapping.items():
            if state_var in description and standard_name not in found_locations:
                found_locations.append(standard_name)
        
        locations.append(found_locations[0] if found_locations else None)
    
    progress_bar.progress(1.0)
    df = df.copy()
    df['extracted_location'] = locations
    
    return df

# Analyze sentiment by state
def analyze_sentiment_by_state(df, sentiment_column='sentiment_score'):
    """Analyze sentiment by state"""
    df_with_locations = df.dropna(subset=['extracted_location', sentiment_column])
    
    if len(df_with_locations) == 0:
        return None
    
    sentiment_by_state = df_with_locations.groupby('extracted_location')[sentiment_column].agg([
        ('avg_sentiment', 'mean'),
        ('count', 'count')
    ]).reset_index()
    
    return sentiment_by_state

# Create India sentiment map
def create_india_sentiment_map(sentiment_data, geojson_data, title):
    """Create choropleth map of India showing sentiment by state"""
    if sentiment_data is None or geojson_data is None:
        return None
    
    min_sentiment = sentiment_data['avg_sentiment'].min()
    max_sentiment = sentiment_data['avg_sentiment'].max()
    
    if min_sentiment < 0 and max_sentiment > 0:
        abs_max = max(abs(min_sentiment), abs(max_sentiment))
        color_range = [-abs_max, abs_max]
    else:
        color_range = [min_sentiment - 0.1, max_sentiment + 0.1]
    
    fig = px.choropleth_mapbox(
        sentiment_data,
        geojson=geojson_data,
        locations='extracted_location',
        featureidkey="properties.NAME_1",
        color='avg_sentiment',
        color_continuous_scale="RdBu",
        range_color=color_range,
        mapbox_style="carto-positron",
        zoom=3.5,
        center={"lat": 20.5937, "lon": 78.9629},
        opacity=0.7,
        hover_data=['count'],
        labels={
            'avg_sentiment': 'Avg Sentiment',
            'extracted_location': 'State',
            'count': 'Articles'
        }
    )
    
    fig.update_layout(
        title=dict(text=title, font=dict(size=20), x=0.5),
        height=600,
        margin={"r":0,"t":50,"l":0,"b":0}
    )
    
    return fig

# Top locations chart
def create_top_locations_chart(df, title):
    """Create bar chart of top mentioned locations"""
    if 'extracted_location' not in df.columns or df['extracted_location'].isna().all():
        return None
    
    location_counts = df['extracted_location'].value_counts().head(15).reset_index()
    location_counts.columns = ['Location', 'Count']
    
    fig = px.bar(
        location_counts,
        y='Location',
        x='Count',
        title=title,
        orientation='h',
        color='Count',
        color_continuous_scale='Viridis'
    )
    
    fig.update_layout(height=500, yaxis={'categoryorder':'total ascending'})
    return fig

# Discover datasets
@st.cache_data
def discover_datasets():
    """Discover datasets in /tmp directory"""
    datasets = {}
    
    search_paths = [
        BASE_DIR / 'data',
        BASE_DIR / 'output',
        BASE_DIR,
    ]
    
    for directory in search_paths:
        if directory.exists():
            try:
                for csv_file in directory.glob('*.csv'):
                    name = csv_file.stem.replace('_articles', '').replace('_', ' ').title()
                    if name not in datasets:
                        datasets[name] = str(csv_file)
            except Exception:
                continue
    
    return datasets

# Load data
@st.cache_data
def load_data(file_path):
    try:
        df = pd.read_csv(file_path)
        
        date_cols = [col for col in df.columns if 'date' in col.lower()]
        if date_cols:
            df['date'] = pd.to_datetime(df[date_cols[0]], errors='coerce')
        
        sentiment_cols = [col for col in df.columns if 'sentiment' in col.lower()]
        if sentiment_cols and 'sentiment_value' not in df.columns:
            df['sentiment_value'] = df[sentiment_cols[0]]
        
        if 'sentiment_score' not in df.columns and 'sentiment_value' in df.columns:
            sentiment_map = {'positive': 1.0, 'negative': -1.0, 'neutral': 0.0}
            df['sentiment_score'] = df['sentiment_value'].str.lower().map(sentiment_map).fillna(0)
        
        return df
    except Exception as e:
        st.error(f"Error loading data: {str(e)}")
        return None

# Plotting functions
def plot_sentiment_trends(df, title):
    if 'date' not in df.columns or 'sentiment_value' not in df.columns:
        return None
    
    plt.style.use('seaborn-v0_8-whitegrid')
    df['year'] = df['date'].dt.year
    
    valid_sentiments = {"positive", "negative", "neutral"}
    df['sentiment'] = df['sentiment_value'].apply(
        lambda x: x.lower() if isinstance(x, str) and x.lower() in valid_sentiments else "neutral"
    )
    
    sentiment_counts = df.groupby(['year', 'sentiment']).size().reset_index(name='count')
    year_totals = sentiment_counts.groupby('year')['count'].sum().reset_index(name='total')
    sentiment_counts = sentiment_counts.merge(year_totals, on='year')
    sentiment_counts['percentage'] = sentiment_counts['count'] / sentiment_counts['total'] * 100
    
    sentiment_pivot = sentiment_counts.pivot(
        index='year', columns='sentiment', values='percentage'
    ).fillna(0)
    
    for sentiment in ['negative', 'neutral', 'positive']:
        if sentiment not in sentiment_pivot.columns:
            sentiment_pivot[sentiment] = 0
    
    fig, ax = plt.subplots(figsize=(12, 6))
    
    colors = {
        'negative': '#e74c3c',
        'neutral': '#95a5a6',
        'positive': '#2ecc71'
    }
    
    for sentiment in ['negative', 'neutral', 'positive']:
        ax.plot(
            sentiment_pivot.index,
            sentiment_pivot[sentiment],
            marker='o',
            linewidth=2.5,
            label=sentiment.capitalize(),
            color=colors[sentiment],
            markersize=7
        )
    
    ax.set_ylabel('Percentage (%)', fontweight='bold')
    ax.set_xlabel('Year', fontweight='bold')
    ax.set_title(title, fontweight='bold', pad=15)
    ax.legend(loc='best', frameon=True)
    ax.grid(axis='y', linestyle='--', alpha=0.3)
    
    plt.tight_layout()
    return fig

def create_sentiment_pie(df, title):
    if 'sentiment_value' not in df.columns:
        return None
    
    sentiment_counts = df['sentiment_value'].str.lower().value_counts()
    
    fig = px.pie(
        values=sentiment_counts.values,
        names=[s.title() for s in sentiment_counts.index],
        title=title,
        color_discrete_map={
            'Positive': '#2ecc71',
            'Neutral': '#95a5a6',
            'Negative': '#e74c3c'
        }
    )
    fig.update_traces(textposition='inside', textinfo='percent+label')
    return fig

# Run scraper function
def run_scraper(source, topic, max_workers=4, max_articles=100):
    """
    Run the appropriate scraper based on source
    Returns dict with success status, file path, and article count
    """
    try:
        source_map = {
            'toi': ('scrapers.toi_scraper', 'TOIArticleScraper'),
            'ndtv': ('scrapers.ndtv_scraper', 'NDTVArticleScraper'),
            'wion': ('scrapers.wion_scraper', 'WIONArticleScraper'),
            'scroll': ('scrapers.scroll_scraper', 'ScrollArticleScraper')
        }
        
        if source not in source_map:
            return {'success': False, 'error': f'Unknown source: {source}'}
        
        module_name, class_name = source_map[source]
        
        try:
            module = __import__(module_name, fromlist=[class_name])
            ScrapeClass = getattr(module, class_name)
        except (ImportError, AttributeError) as e:
            return {
                'success': False, 
                'error': f'Scraper module not found. Please ensure scraper files are available.'
            }
        
        scraper = ScrapeClass(max_workers=max_workers)
        
        output_dir = BASE_DIR / 'output'
        output_dir.mkdir(parents=True, exist_ok=True)
        
        original_dir = os.getcwd()
        os.chdir(output_dir)
        
        try:
            if source == 'toi':
                topic_url = f"{scraper.base_url}/topic/{topic}/news"
                final_csv = scraper.scrape_topic(topic_url, topic)
            elif source == 'ndtv':
                final_csv = scraper.scrape_topic(topic)
            elif source in ['wion', 'scroll']:
                final_csv = scraper.scrape_topic(topic.lower(), topic)
            
            articles = getattr(scraper, 'articles', None) or getattr(scraper, 'scraped_articles', [])
            article_count = len(articles)
            
            if article_count > max_articles:
                df = pd.read_csv(final_csv)
                df = df.head(max_articles)
                df.to_csv(final_csv, index=False)
                article_count = max_articles
            
            return {
                'success': True,
                'file': str(output_dir / final_csv),
                'count': article_count
            }
            
        finally:
            os.chdir(original_dir)
            
    except Exception as e:
        return {
            'success': False,
            'error': str(e)
        }

# MAIN APP PAGES

def show_home_page():
    st.markdown('<h1 class="main-header">📰 News Scraper & Analysis Platform</h1>', 
                unsafe_allow_html=True)
    
    st.markdown("""
    <div style="text-align: center; padding: 20px; background-color: #f8f9fa; 
                border-radius: 10px; margin: 20px 0;">
        <h3>Complete Pipeline for News Collection and Sentiment Analysis</h3>
        <p>Scrape or upload articles from major Indian news sources and analyze sentiment trends</p>
    </div>
    """, unsafe_allow_html=True)
    
    st.info(f"💾 **Storage Location:** `{BASE_DIR}` (temporary - cleared on restart)")
    
    col1, col2, col3, col4 = st.columns(4)
    
    with col1:
        st.markdown('<div class="feature-card">', unsafe_allow_html=True)
        st.markdown("### 🔍 Scrape")
        st.write("Collect articles from major Indian news sources")
        st.markdown('</div>', unsafe_allow_html=True)
    
    with col2:
        st.markdown('<div class="feature-card">', unsafe_allow_html=True)
        st.markdown("### 📤 Upload")
        st.write("Upload pre-scraped CSV files for analysis")
        st.markdown('</div>', unsafe_allow_html=True)
    
    with col3:
        st.markdown('<div class="feature-card">', unsafe_allow_html=True)
        st.markdown("### 📊 Analyze")
        st.write("Automatic sentiment classification and trend analysis")
        st.markdown('</div>', unsafe_allow_html=True)
    
    with col4:
        st.markdown('<div class="feature-card">', unsafe_allow_html=True)
        st.markdown("### 📈 Visualize")
        st.write("Interactive charts and geographic sentiment mapping")
        st.markdown('</div>', unsafe_allow_html=True)
    
    datasets = discover_datasets()
    if datasets:
        st.markdown("---")
        st.markdown("### 📊 Available Datasets")
        
        cols = st.columns(min(len(datasets), 4))
        for idx, (name, path) in enumerate(list(datasets.items())[:4]):
            with cols[idx]:
                df = load_data(path)
                if df is not None:
                    st.markdown('<div class="metric-card">', unsafe_allow_html=True)
                    st.metric(name, f"{len(df):,} articles")
                    st.markdown('</div>', unsafe_allow_html=True)

def show_scraper_page():
    st.markdown('<h2 class="sub-header">📤 Data Upload & Scraper</h2>', unsafe_allow_html=True)
    
    st.warning("""
    ⚠️ **Hugging Face Spaces Notice:** 
    - Data is stored in `/tmp` and will be cleared on app restart
    - Download your processed data regularly
    - Scraping may timeout on HF Spaces - consider running locally for large scrapes
    """)
    
    tab1, tab2 = st.tabs(["📤 Upload Data", "🔍 Scrape Articles"])
    
    with tab1:
        st.markdown("### 📤 Upload Your Data")
        uploaded_file = st.file_uploader(
            "Upload CSV file with scraped articles",
            type=['csv'],
            help="Upload a CSV file with columns: title, date, desc, sentiment_value, etc."
        )
        
        if uploaded_file:
            try:
                data_dir = BASE_DIR / 'data'
                data_dir.mkdir(parents=True, exist_ok=True)
                
                save_path = data_dir / uploaded_file.name
                with open(save_path, 'wb') as f:
                    f.write(uploaded_file.getbuffer())
                
                st.success(f"✅ File uploaded successfully! Saved to `{save_path}`")
                st.info("Go to the Analysis page to view your data.")
                
                discover_datasets.clear()
                
                with st.expander("📋 Preview Data"):
                    df = pd.read_csv(save_path)
                    st.dataframe(df.head(10), use_container_width=True)
                    st.caption(f"Total rows: {len(df):,}")
                
            except Exception as e:
                st.error(f"❌ Error saving file: {e}")
    
    with tab2:
        st.markdown("### 🔍 Scrape News Articles")
        
        st.info("""
        **Note:** This feature requires the scraper modules to be available. 
        On HF Spaces, scraping may timeout for large datasets. For production use, 
        consider running scrapers locally and uploading the results.
        """)
        
        col1, col2 = st.columns(2)
        
        with col1:
            news_source = st.selectbox(
                "News Source",
                ["TOI (Times of India)", "NDTV", "WION", "Scroll.in"],
                help="Select which news website to scrape from"
            )
        
        with col2:
            topic = st.text_input(
                "Topic to Search",
                placeholder="e.g., Climate Change, Politics, Education",
                help="Enter the topic you want to scrape articles about"
            )
        
        col3, col4 = st.columns(2)
        
        with col3:
            max_workers = st.slider(
                "Worker Threads",
                min_value=1,
                max_value=8,
                value=4,
                help="More workers = faster scraping, but may overload the server"
            )
        
        with col4:
            max_articles = st.number_input(
                "Max Articles",
                min_value=10,
                max_value=500,
                value=100,
                step=10,
                help="Maximum number of articles to scrape (set lower for HF Spaces)"
            )
        
        if st.button("🚀 Start Scraping", type="primary", disabled=not topic):
            if not topic:
                st.error("Please enter a topic to search for!")
            else:
                source_map = {
                    "TOI (Times of India)": "toi",
                    "NDTV": "ndtv",
                    "WION": "wion",
                    "Scroll.in": "scroll"
                }
                source_code = source_map[news_source]
                
                try:
                    with st.spinner(f"🔄 Scraping {news_source} for '{topic}'... This may take a few minutes."):
                        result = run_scraper(source_code, topic, max_workers, max_articles)
                        
                        if result['success']:
                            st.success(f"✅ Successfully scraped {result['count']} articles!")
                            st.info(f"📁 Saved to: `{result['file']}`")
                            
                            discover_datasets.clear()
                            
                            with st.expander("📋 Preview Scraped Data"):
                                df = pd.read_csv(result['file'])
                                st.dataframe(df.head(10), use_container_width=True)
                                st.caption(f"Total rows: {len(df):,}")
                            
                            st.balloons()
                        else:
                            st.error(f"❌ Scraping failed: {result['error']}")
                            
                except Exception as e:
                    st.error(f"❌ Error during scraping: {str(e)}")
                    st.info("Scraper modules may not be available on this platform. Consider uploading pre-scraped data instead.")
    
    st.markdown("---")
    
    datasets = discover_datasets()
    if datasets:
        st.markdown("### 📁 Available Datasets")
        for name, path in datasets.items():
            col_a, col_b, col_c = st.columns([3, 1, 1])
            with col_a:
                st.text(f"📄 {name}")
            with col_b:
                df = load_data(path)
                if df is not None:
                    st.text(f"{len(df):,} articles")
            with col_c:
                if Path(path).exists():
                    with open(path, 'rb') as f:
                        st.download_button(
                            "⬇️",
                            f,
                            file_name=f"{name}.csv",
                            mime='text/csv',
                            key=f"dl_{name}"
                        )
    else:
        st.info("No datasets found. Upload a CSV file or scrape articles to get started!")
    
    st.markdown("---")
    
    st.markdown("### 💾 Storage Information")
    try:
        if BASE_DIR.exists():
            file_count = sum(1 for _ in BASE_DIR.rglob('*.csv'))
            st.metric("CSV Files", file_count)
            st.caption(f"Location: `{BASE_DIR}`")
        else:
            st.info("No data directory created yet. Upload a file to get started.")
    except Exception as e:
        st.error(f"Could not access storage: {e}")

def show_analysis_page():
    st.markdown('<h2 class="sub-header">📊 Sentiment Analysis Dashboard</h2>', 
                unsafe_allow_html=True)
    
    datasets = discover_datasets()
    
    if not datasets:
        st.warning("⚠️ No datasets available. Please upload a CSV file or scrape articles first!")
        st.info("👉 Go to the 'Upload & Scrape' page to get started.")
        return
    
    selected = st.selectbox("Select Dataset", options=list(datasets.keys()))
    
    if selected:
        df = load_data(datasets[selected])
        
        if df is not None:
            col1, col2, col3, col4 = st.columns(4)
            
            with col1:
                st.metric("📄 Total Articles", f"{len(df):,}")
            
            with col2:
                if 'date' in df.columns:
                    years = f"{df['date'].dt.year.min()}-{df['date'].dt.year.max()}"
                    st.metric("📅 Years", years)
            
            with col3:
                if 'sentiment_value' in df.columns:
                    pos_pct = (df['sentiment_value'].str.lower() == 'positive').mean() * 100
                    st.metric("😊 Positive", f"{pos_pct:.1f}%")
            
            with col4:
                if 'sentiment_value' in df.columns:
                    neg_pct = (df['sentiment_value'].str.lower() == 'negative').mean() * 100
                    st.metric("😞 Negative", f"{neg_pct:.1f}%")
            
            st.markdown("---")
            
            tab1, tab2, tab3, tab4 = st.tabs(["📈 Trends", "🥧 Distribution", "🗺️ Geographic", "📝 Articles"])
            
            with tab1:
                fig = plot_sentiment_trends(df, f"{selected} - Sentiment Trends")
                if fig:
                    st.pyplot(fig)
                else:
                    st.info("Insufficient data for trend analysis")
            
            with tab2:
                col_a, col_b = st.columns([2, 1])
                with col_a:
                    pie_fig = create_sentiment_pie(df, "Sentiment Distribution")
                    if pie_fig:
                        st.plotly_chart(pie_fig, use_container_width=True)
                
                with col_b:
                    if 'sentiment_value' in df.columns:
                        st.markdown("### Breakdown")
                        counts = df['sentiment_value'].value_counts()
                        for sentiment, count in counts.items():
                            st.metric(sentiment.title(), f"{count:,}")
            
            with tab3:
                st.markdown("### 🗺️ Geographic Sentiment Analysis")
                
                if 'extracted_location' not in df.columns:
                    if 'desc' in df.columns or 'description' in df.columns:
                        if st.button("🔍 Extract Locations from Articles"):
                            with st.spinner("Extracting locations... This may take a few minutes."):
                                desc_col = 'desc' if 'desc' in df.columns else 'description'
                                df = extract_locations_from_descriptions(df, desc_col)
                                df.to_csv(datasets[selected], index=False)
                                st.success("✅ Locations extracted successfully!")
                                st.rerun()
                    else:
                        st.info("No description column found. Cannot extract locations.")
                else:
                    col_left, col_right = st.columns([3, 2])
                    
                    with col_left:
                        st.markdown("#### Sentiment by State")
                        
                        india_geojson = load_india_geojson()
                        
                        if india_geojson:
                            sentiment_by_state = analyze_sentiment_by_state(df)
                            
                            if sentiment_by_state is not None and not sentiment_by_state.empty:
                                map_fig = create_india_sentiment_map(
                                    sentiment_by_state, 
                                    india_geojson,
                                    f"{selected} - Sentiment by Indian States"
                                )
                                if map_fig:
                                    st.plotly_chart(map_fig, use_container_width=True)
                                
                                with st.expander("📊 State-wise Statistics"):
                                    sentiment_by_state_display = sentiment_by_state.sort_values('count', ascending=False)
                                    st.dataframe(
                                        sentiment_by_state_display,
                                        use_container_width=True,
                                        hide_index=True
                                    )
                            else:
                                st.warning("No location data with valid sentiment found.")
                        else:
                            st.error("Could not load India map data.")
                    
                    with col_right:
                        st.markdown("#### Top Mentioned Locations")
                        top_loc_fig = create_top_locations_chart(df, "Top 15 Locations")
                        if top_loc_fig:
                            st.plotly_chart(top_loc_fig, use_container_width=True)
                        
                        total_articles = len(df)
                        articles_with_location = df['extracted_location'].notna().sum()
                        coverage = (articles_with_location / total_articles) * 100
                        
                        st.metric("Location Coverage", f"{coverage:.1f}%")
                        st.caption(f"{articles_with_location:,} out of {total_articles:,} articles have location data")
            
            with tab4:
                col_a, col_b, col_c = st.columns(3)
                
                with col_a:
                    sentiment_filter = st.selectbox(
                        "Sentiment",
                        ["All"] + sorted(df['sentiment_value'].unique().tolist())
                    )
                
                with col_b:
                    if 'date' in df.columns:
                        years = sorted(df['date'].dt.year.dropna().unique())
                        year_filter = st.selectbox("Year", ["All"] + years)
                    else:
                        year_filter = "All"
                
                with col_c:
                    num_articles = st.slider("Display", 5, 50, 10)
                
                filtered_df = df.copy()
                if sentiment_filter != "All":
                    filtered_df = filtered_df[filtered_df['sentiment_value'] == sentiment_filter]
                if year_filter != "All" and 'date' in df.columns:
                    filtered_df = filtered_df[filtered_df['date'].dt.year == year_filter]
                
                st.write(f"Showing {min(num_articles, len(filtered_df))} of {len(filtered_df)} articles")
                
                for idx, row in filtered_df.head(num_articles).iterrows():
                    with st.expander(f"📰 {row.get('title', 'Untitled')}"):
                        col_x, col_y = st.columns([3, 1])
                        
                        with col_x:
                            st.write(row.get('desc', row.get('description', 'No description')))
                            if 'link' in row:
                                st.markdown(f"[Read more →]({row['link']})")
                        
                        with col_y:
                            sentiment = row.get('sentiment_value', 'Unknown')
                            sentiment_emoji = {
                                'positive': '😊',
                                'negative': '😞',
                                'neutral': '😐'
                            }.get(sentiment.lower(), '❓')
                            
                            st.metric("Sentiment", f"{sentiment_emoji} {sentiment.title()}")
                            if 'date' in row:
                                st.caption(f"📅 {row['date'].strftime('%d %b %Y')}")

def show_about_page():
    st.markdown('<h2 class="sub-header">ℹ️ About This Platform</h2>', 
                unsafe_allow_html=True)
    
    st.markdown("""
    ## 🎯 Overview
    
    This platform provides sentiment analysis and visualization for news articles,
    specifically designed for Indian news sources. It includes both upload functionality
    and integrated web scraping capabilities.
    
    ### ✨ Key Features
    
    - **Web Scraping**: Built-in scrapers for TOI, NDTV, WION, and Scroll.in
    - **Data Upload**: Upload pre-scraped CSV files for analysis
    - **Automatic Analysis**: Sentiment classification and scoring
    - **Interactive Visualizations**: Trends, distributions, and comparisons
    - **Geographic Analysis**: State-wise sentiment mapping for India
    - **Data Export**: Download processed datasets
    
    ### 🔧 Technical Stack
    
    - **Frontend**: Streamlit
    - **Data Processing**: Pandas, NumPy
    - **Visualization**: Plotly, Matplotlib
    - **NLP**: spaCy for location extraction
    - **Web Scraping**: BeautifulSoup4, Selenium (if available)
    - **Storage**: `/tmp` directory (HF Spaces compatible)
    
    ### 📖 How to Use
    
    1. **Scrape or Upload**: 
       - Use the built-in scraper to collect articles from news sources
       - Or upload your own CSV file with pre-scraped data
    2. **Analyze**: Go to Analysis page and select your dataset
    3. **Explore**: View trends, distributions, and geographic sentiment
    4. **Extract Locations**: Use the Geographic tab to extract state information
    5. **Export**: Download processed data for further use
    
    ### 🔍 Supported News Sources
    
    - **Times of India (TOI)**: India's largest English daily
    - **NDTV**: Leading Indian news channel
    - **WION**: International news perspective
    - **Scroll.in**: Independent digital news
    
    ### 📋 CSV File Format
    
    Your uploaded CSV should contain these columns:
    - `title`: Article headline
    - `date`: Publication date
    - `desc` or `description`: Article content/summary
    - `sentiment_value`: Sentiment label (positive/negative/neutral)
    - `link` (optional): URL to original article
    
    ### ⚠️ Hugging Face Spaces Limitations
    
    - Data stored in `/tmp` is temporary and cleared on restart
    - Scraping may timeout for large datasets (>500 articles)
    - Download your processed data regularly
    - For heavy scraping, consider running locally
    
    ### 💡 Tips for Best Results
    
    - Start with smaller scrapes (50-100 articles) to test
    - Use specific topics for better quality results
    - Lower worker threads (2-4) for stability on HF Spaces
    - Download results immediately after scraping
    
    ### 🤝 Support
    
    For issues or questions, please refer to the documentation or contact support.
    
    ---
    
    **Version**: 2.0.0 (HF Spaces Edition with Scraping)  
    **Last Updated**: October 2025  
    **Storage**: `/tmp/news_scraper`
    """)

# MAIN APP
def main():
    try:
        load_css()
        init_session_state()
        
        if not setup_directories():
            st.error("Failed to setup directories. Some features may not work.")
        
        with st.sidebar:
            st.markdown("### 📰 News Analysis")
            st.markdown("---")
            
            page = st.radio(
                "Navigation",
                ["🏠 Home", "📤 Upload & Scrape", "📊 Analysis", "ℹ️ About"],
                label_visibility="collapsed"
            )
            
            st.markdown("---")
            
            st.markdown("### 💾 Storage")
            st.caption(f"Location: `/tmp/news_scraper`")
            st.caption("⚠️ Temporary storage")
            
            st.markdown("---")
            
            try:
                datasets = discover_datasets()
                if datasets:
                    st.markdown("### 📊 Quick Stats")
                    total_articles = 0
                    for path in datasets.values():
                        try:
                            df = load_data(path)
                            if df is not None:
                                total_articles += len(df)
                        except:
                            continue
                    
                    if total_articles > 0:
                        st.metric("Total Articles", f"{total_articles:,}")
                        st.metric("Datasets", len(datasets))
                else:
                    st.info("No data yet. Upload a CSV to start!")
            except Exception:
                pass
        
        if page == "🏠 Home":
            show_home_page()
        elif page == "📤 Upload & Scrape":
            show_scraper_page()
        elif page == "📊 Analysis":
            show_analysis_page()
        else:
            show_about_page()
            
    except Exception as e:
        st.error(f"❌ Application error: {str(e)}")
        st.info("Try refreshing the page. If the problem persists, the app may need to restart.")
        
        with st.expander("🔍 Error Details"):
            import traceback
            st.code(traceback.format_exc())

if __name__ == "__main__":
    main()