use std::collections::HashSet;
use std::fs::OpenOptions;
use std::io::Write;
use std::sync::{Arc, Mutex};
use std::time::{Duration, Instant};
use tokio::sync::Semaphore;
use reqwest::Client;
use scraper::{Html, Selector};
use url::Url;

const MAX_CONCURRENT: usize = 500;
const OUTPUT_FILE: &str = "/workspace/crawl.jsonl";

#[tokio::main]
async fn main() {
    let client = Client::builder()
        .timeout(Duration::from_secs(10))
        .user_agent("Mozilla/5.0 (compatible; Googlebot/2.1)")
        .build()
        .unwrap();

    let visited: Arc<Mutex<HashSet<String>>> = Arc::new(Mutex::new(HashSet::new()));
    let queue: Arc<Mutex<Vec<String>>> = Arc::new(Mutex::new(Vec::new()));
    let file = Arc::new(Mutex::new(
        OpenOptions::new().create(true).append(true).open(OUTPUT_FILE).unwrap()
    ));
    let semaphore = Arc::new(Semaphore::new(MAX_CONCURRENT));
    
    // Seed URLs
    let seeds = vec![
        "https://en.wikipedia.org", "https://www.bbc.com", "https://arxiv.org",
        "https://www.nature.com", "https://www.theguardian.com", "https://reuters.com",
        "https://www.npr.org", "https://www.sciencedaily.com", "https://arstechnica.com",
    ];
    
    {
        let mut q = queue.lock().unwrap();
        for seed in seeds {
            q.push(seed.to_string());
        }
    }

    let start = Instant::now();
    let pages = Arc::new(Mutex::new(0u64));
    let bytes = Arc::new(Mutex::new(0u64));

    loop {
        let url = {
            let mut q = queue.lock().unwrap();
            q.pop()
        };

        let url = match url {
            Some(u) => u,
            None => {
                tokio::time::sleep(Duration::from_millis(100)).await;
                continue;
            }
        };

        {
            let mut v = visited.lock().unwrap();
            if v.contains(&url) { continue; }
            v.insert(url.clone());
        }

        let permit = semaphore.clone().acquire_owned().await.unwrap();
        let client = client.clone();
        let queue = queue.clone();
        let file = file.clone();
        let pages = pages.clone();
        let bytes = bytes.clone();
        let visited = visited.clone();
        let start = start.clone();

        tokio::spawn(async move {
            let _permit = permit;
            
            if let Ok(resp) = client.get(&url).send().await {
                if let Ok(text) = resp.text().await {
                    let text_len = text.len();
                    
                    // Write to file
                    let json = serde_json::json!({
                        "url": url,
                        "text": text,
                        "tokens_est": text_len / 4
                    });
                    
                    {
                        let mut f = file.lock().unwrap();
                        writeln!(f, "{}", json).ok();
                    }

                    // Extract links
                    let doc = Html::parse_document(&text);
                    let selector = Selector::parse("a[href]").unwrap();
                    let base = Url::parse(&url).ok();
                    
                    let mut new_urls = Vec::new();
                    for elem in doc.select(&selector) {
                        if let Some(href) = elem.value().attr("href") {
                            if let Some(base) = &base {
                                if let Ok(abs) = base.join(href) {
                                    let abs_str = abs.to_string();
                                    if abs_str.starts_with("http") {
                                        new_urls.push(abs_str);
                                    }
                                }
                            }
                        }
                    }

                    {
                        let v = visited.lock().unwrap();
                        let mut q = queue.lock().unwrap();
                        for u in new_urls {
                            if !v.contains(&u) && q.len() < 5_000_000 {
                                q.push(u);
                            }
                        }
                    }

                    // Update stats
                    {
                        let mut p = pages.lock().unwrap();
                        let mut b = bytes.lock().unwrap();
                        *p += 1;
                        *b += text_len as u64;
                        
                        if *p % 100 == 0 {
                            let elapsed = start.elapsed().as_secs();
                            let mb = *b / 1_048_576;
                            let qlen = queue.lock().unwrap().len();
                            let vlen = visited.lock().unwrap().len();
                            eprintln!("📊 {}s | {} pages | {} MB | queue: {} | visited: {}", 
                                elapsed, p, mb, qlen, vlen);
                        }
                    }
                }
            }
        });
    }
}