2024-10-04 19:52:34 +00:00
|
|
|
extern crate html5ever;
|
2024-11-13 06:08:09 +00:00
|
|
|
|
|
|
|
use std::time::Instant;
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-11-09 18:30:32 +00:00
|
|
|
use db::{connect, Website};
|
2024-11-13 04:03:58 +00:00
|
|
|
use s3::S3;
|
2024-11-13 06:08:09 +00:00
|
|
|
use surrealdb::{engine::remote::ws::Client, Surreal};
|
2024-11-13 01:40:10 +00:00
|
|
|
use tracing::{debug, info, instrument, trace, trace_span};
|
2024-11-09 18:30:32 +00:00
|
|
|
use tracing_subscriber::EnvFilter;
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-07 17:14:56 +00:00
|
|
|
mod db;
|
2024-11-13 04:03:58 +00:00
|
|
|
mod s3;
|
2024-11-13 06:08:09 +00:00
|
|
|
mod parser;
|
2024-11-13 04:03:58 +00:00
|
|
|
|
|
|
|
struct Config<'a> {
|
|
|
|
surreal_ns: &'a str,
|
|
|
|
surreal_db: &'a str,
|
|
|
|
surreal_url: &'a str,
|
|
|
|
surreal_username: &'a str,
|
|
|
|
surreal_password: &'a str,
|
|
|
|
|
|
|
|
s3_url: &'a str,
|
|
|
|
s3_bucket: &'a str,
|
|
|
|
s3_access_key: &'a str,
|
|
|
|
s3_secret_key: &'a str,
|
|
|
|
}
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-04 19:52:34 +00:00
|
|
|
#[tokio::main]
|
|
|
|
async fn main() {
|
2024-11-09 18:30:32 +00:00
|
|
|
tracing_subscriber::fmt()
|
|
|
|
.with_env_filter(EnvFilter::from_default_env())
|
|
|
|
.with_line_number(true)
|
|
|
|
.without_time()
|
|
|
|
.init();
|
2024-10-04 19:52:34 +00:00
|
|
|
debug!("Starting...");
|
2024-11-13 04:03:58 +00:00
|
|
|
|
|
|
|
let config = Config {
|
|
|
|
surreal_ns: "test",
|
2024-11-13 06:08:09 +00:00
|
|
|
surreal_db: "v1.7",
|
2024-11-13 04:03:58 +00:00
|
|
|
surreal_url: "localhost:8000",
|
|
|
|
surreal_username: "root",
|
|
|
|
surreal_password: "root",
|
|
|
|
s3_url: "http://localhost:9000",
|
2024-11-13 06:08:09 +00:00
|
|
|
s3_bucket: "v1.7",
|
2024-11-13 04:03:58 +00:00
|
|
|
s3_access_key: "8tUJn7e1paMFZQr0PKIT",
|
|
|
|
s3_secret_key: "uSMvYxNOeCejCUgXVqgTfYlUEcmiZY0xcZ91M9E0",
|
|
|
|
};
|
2024-10-04 19:52:34 +00:00
|
|
|
|
2024-11-09 18:30:32 +00:00
|
|
|
// Would probably take these in as parameters from a cli
|
2024-11-13 04:03:58 +00:00
|
|
|
let starting_url = "https://oliveratkinson.net/";
|
2024-11-13 06:08:09 +00:00
|
|
|
let budget = 15;
|
2024-11-09 18:30:32 +00:00
|
|
|
let mut crawled = 0;
|
|
|
|
|
2024-11-13 04:03:58 +00:00
|
|
|
let s3 = S3::connect(&config).await.expect("Failed to connect to minio, aborting.");
|
|
|
|
let db = connect(&config).await.expect("Failed to connect to surreal, aborting.");
|
|
|
|
|
|
|
|
let reqwest = reqwest::Client::builder()
|
2024-11-11 03:37:00 +00:00
|
|
|
// .use_rustls_tls()
|
|
|
|
.build()
|
|
|
|
.unwrap();
|
|
|
|
|
2024-11-09 18:30:32 +00:00
|
|
|
// Kick off the whole machine - This Website object doesn't matter, it's just to allow for
|
|
|
|
// get() to work.
|
2024-11-09 22:28:10 +00:00
|
|
|
let span = trace_span!("Pre-Loop");
|
|
|
|
let pre_loop_span = span.enter();
|
2024-11-13 00:50:06 +00:00
|
|
|
// Download the site
|
2024-11-13 04:03:58 +00:00
|
|
|
let mut site = Website::new(&starting_url, false);
|
|
|
|
get(&mut site, &db, &reqwest, &s3, &mut crawled).await;
|
2024-11-13 00:50:06 +00:00
|
|
|
|
2024-11-09 22:28:10 +00:00
|
|
|
drop(pre_loop_span);
|
2024-11-09 18:30:32 +00:00
|
|
|
|
2024-11-09 22:28:10 +00:00
|
|
|
let span = trace_span!("Loop");
|
|
|
|
let span = span.enter();
|
2024-11-09 18:30:32 +00:00
|
|
|
while crawled < budget {
|
2024-11-13 06:08:09 +00:00
|
|
|
let get_num = if budget - crawled < 100 { budget - crawled } else { 100 };
|
2024-11-10 06:30:57 +00:00
|
|
|
|
|
|
|
let uncrawled = get_uncrawled_links(&db, get_num).await;
|
2024-11-13 00:50:06 +00:00
|
|
|
if uncrawled.len() == 0 {
|
|
|
|
info!("Had more budget but finished crawling everything.");
|
|
|
|
return;
|
|
|
|
}
|
2024-11-09 18:30:32 +00:00
|
|
|
debug!("Crawling {} pages...", uncrawled.len());
|
|
|
|
|
2024-11-09 22:28:10 +00:00
|
|
|
let span = trace_span!("Crawling");
|
|
|
|
let _ = span.enter();
|
|
|
|
|
2024-11-09 18:30:32 +00:00
|
|
|
for mut site in uncrawled {
|
2024-11-13 04:03:58 +00:00
|
|
|
get(&mut site, &db, &reqwest, &s3, &mut crawled).await;
|
2024-11-13 06:08:09 +00:00
|
|
|
|
2024-11-13 01:40:10 +00:00
|
|
|
let percent = format!("{:.2}%", (crawled as f32 / budget as f32) * 100f32);
|
|
|
|
info!("Crawled {crawled} out of {budget} pages. ({percent})");
|
2024-11-09 18:30:32 +00:00
|
|
|
}
|
|
|
|
}
|
2024-11-09 22:28:10 +00:00
|
|
|
drop(span);
|
2024-10-31 21:09:48 +00:00
|
|
|
|
|
|
|
info!("Done");
|
2024-10-07 17:14:56 +00:00
|
|
|
}
|
|
|
|
|
2024-11-09 18:30:32 +00:00
|
|
|
#[instrument(skip_all)]
|
2024-11-13 06:08:09 +00:00
|
|
|
/// Downloads and crawls and stores a webpage.
|
2024-11-13 01:40:10 +00:00
|
|
|
async fn get(
|
|
|
|
site: &mut Website,
|
|
|
|
db: &Surreal<Client>,
|
2024-11-13 04:03:58 +00:00
|
|
|
reqwest: &reqwest::Client,
|
|
|
|
s3: &S3,
|
2024-11-13 01:40:10 +00:00
|
|
|
count: &mut usize,
|
|
|
|
) {
|
2024-11-09 22:28:10 +00:00
|
|
|
trace!("Get: {}", site.to_string());
|
2024-11-11 03:24:04 +00:00
|
|
|
let timer = Timer::start("Got page");
|
2024-11-13 04:03:58 +00:00
|
|
|
if let Ok(response) = reqwest.get(site.to_string()).send().await {
|
2024-11-13 01:40:10 +00:00
|
|
|
timer.stop();
|
2024-11-10 06:30:57 +00:00
|
|
|
|
2024-11-13 01:40:10 +00:00
|
|
|
// Get body
|
2024-11-09 18:30:32 +00:00
|
|
|
let data = response.text().await.unwrap();
|
2024-11-13 06:08:09 +00:00
|
|
|
// Store document
|
2024-11-13 04:03:58 +00:00
|
|
|
s3.store(&data, &site.site).await;
|
2024-11-13 06:08:09 +00:00
|
|
|
// Parse document and store relationships
|
|
|
|
parser::parse(db, site, data).await;
|
2024-11-13 01:40:10 +00:00
|
|
|
*count += 1;
|
2024-11-09 18:30:32 +00:00
|
|
|
}
|
2024-11-09 22:28:10 +00:00
|
|
|
trace!("Failed to get: {}", site.to_string());
|
2024-10-07 17:14:56 +00:00
|
|
|
}
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-11-09 22:28:10 +00:00
|
|
|
/// Returns uncrawled links
|
|
|
|
async fn get_uncrawled_links(db: &Surreal<Client>, mut count: usize) -> Vec<Website> {
|
|
|
|
if count > 100 {
|
|
|
|
count = 100
|
|
|
|
}
|
|
|
|
|
|
|
|
let mut response = db
|
|
|
|
.query("SELECT * FROM website WHERE crawled = false LIMIT $count")
|
|
|
|
.bind(("count", count))
|
|
|
|
.await
|
|
|
|
.expect("Hard-coded query failed..?");
|
2024-11-13 01:40:10 +00:00
|
|
|
response
|
|
|
|
.take(0)
|
|
|
|
.expect("Returned websites couldn't be parsed")
|
2024-11-09 18:30:32 +00:00
|
|
|
}
|
|
|
|
|
2024-11-11 03:24:04 +00:00
|
|
|
pub struct Timer<'a> {
|
|
|
|
start: Instant,
|
|
|
|
msg: &'a str,
|
|
|
|
}
|
|
|
|
|
|
|
|
impl<'a> Timer<'a> {
|
|
|
|
#[inline]
|
|
|
|
pub fn start(msg: &'a str) -> Self {
|
2024-11-13 01:40:10 +00:00
|
|
|
Self {
|
|
|
|
start: Instant::now(),
|
|
|
|
msg,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
pub fn stop(&self) -> f64 {
|
|
|
|
let dif = self.start.elapsed().as_micros();
|
|
|
|
let ms = dif as f64 / 1000.;
|
|
|
|
debug!("{}", format!("{} in {:.3}ms", self.msg, ms));
|
|
|
|
ms
|
2024-11-11 03:24:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2024-11-13 01:40:10 +00:00
|
|
|
impl Drop for Timer<'_> {
|
2024-11-11 03:24:04 +00:00
|
|
|
fn drop(&mut self) {
|
2024-11-13 01:40:10 +00:00
|
|
|
self.stop();
|
2024-11-11 03:24:04 +00:00
|
|
|
}
|
2024-11-13 00:50:06 +00:00
|
|
|
}
|