2024-10-04 19:52:34 +00:00
|
|
|
extern crate markup5ever_rcdom as rcdom;
|
|
|
|
extern crate html5ever;
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-31 21:09:48 +00:00
|
|
|
use std::rc::Rc;
|
2024-10-07 17:14:56 +00:00
|
|
|
use db::connect;
|
2024-10-04 19:52:34 +00:00
|
|
|
use html5ever::{parse_document, tendril::TendrilSink, tree_builder::TreeBuilderOpts, ParseOpts};
|
2024-10-07 17:14:56 +00:00
|
|
|
use rcdom::{Node, RcDom};
|
|
|
|
use surrealdb::{engine::remote::ws::Client, Surreal};
|
2024-10-31 21:32:37 +00:00
|
|
|
use tracing::{debug, error, info, warn};
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-07 17:14:56 +00:00
|
|
|
mod db;
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-04 19:52:34 +00:00
|
|
|
#[tokio::main]
|
|
|
|
async fn main() {
|
|
|
|
tracing_subscriber::fmt::init();
|
|
|
|
debug!("Starting...");
|
|
|
|
|
|
|
|
let url = "https://oliveratkinson.net";
|
|
|
|
|
2024-10-07 17:14:56 +00:00
|
|
|
let db = connect().await.expect("Failed to connect to db, aborting.");
|
2024-10-31 21:09:48 +00:00
|
|
|
let dom = get(url).await;
|
2024-10-07 17:14:56 +00:00
|
|
|
|
|
|
|
walk(&dom, &db, url).await;
|
2024-10-31 21:09:48 +00:00
|
|
|
|
|
|
|
info!("Done");
|
2024-10-07 17:14:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
async fn get(url: &str) -> Rc<Node> {
|
2024-10-04 19:52:34 +00:00
|
|
|
let response = reqwest::get(url).await.unwrap();
|
|
|
|
let data = response.text().await.unwrap();
|
|
|
|
|
|
|
|
let opts = ParseOpts {
|
|
|
|
tree_builder: TreeBuilderOpts {
|
|
|
|
drop_doctype: true,
|
|
|
|
..Default::default()
|
|
|
|
},
|
|
|
|
..Default::default()
|
2024-08-25 21:50:59 +00:00
|
|
|
};
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-04 19:52:34 +00:00
|
|
|
let dom = parse_document(RcDom::default(), opts)
|
|
|
|
.from_utf8()
|
|
|
|
.read_from(&mut data.as_bytes())
|
|
|
|
.unwrap();
|
2024-10-07 17:14:56 +00:00
|
|
|
dom.document
|
|
|
|
}
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-07 17:14:56 +00:00
|
|
|
async fn walk(node: &rcdom::Handle, db: &Surreal<Client> , site_name: &str) {
|
2024-10-31 21:09:48 +00:00
|
|
|
// Insert Or Update
|
2024-10-31 21:32:37 +00:00
|
|
|
let _: Option<Vec<db::Record>> = match db.upsert(("website", site_name)).content(db::Website { href: String::from("/"), crawled: true, site: site_name.to_string() } ).await {
|
2024-10-31 21:09:48 +00:00
|
|
|
Ok(e) => {
|
|
|
|
// Return this for type coercion
|
|
|
|
e
|
|
|
|
},
|
2024-10-07 17:14:56 +00:00
|
|
|
Err(e) => {
|
2024-10-31 21:32:37 +00:00
|
|
|
// error!("{}", e);
|
|
|
|
None
|
2024-10-07 17:14:56 +00:00
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2024-10-04 19:52:34 +00:00
|
|
|
match &node.data {
|
|
|
|
rcdom::NodeData::Element { name, attrs, template_contents, mathml_annotation_xml_integration_point } => {
|
2024-10-31 20:10:14 +00:00
|
|
|
for attr in attrs.borrow().clone() {
|
2024-10-04 19:52:34 +00:00
|
|
|
let name = name.local.to_string();
|
2024-10-07 17:14:56 +00:00
|
|
|
if name == "a" {
|
2024-10-31 20:10:14 +00:00
|
|
|
if attr.value.starts_with("mailto") {
|
2024-10-07 17:14:56 +00:00
|
|
|
// mailto link, lol
|
2024-10-31 21:09:48 +00:00
|
|
|
let created: Option<db::Record> = db.create("email").content(db::Email {
|
|
|
|
email: attr.value.to_string()
|
|
|
|
}).await.unwrap();
|
|
|
|
warn!("{:?}", created)
|
2024-10-07 17:14:56 +00:00
|
|
|
} else {
|
2024-10-31 21:32:37 +00:00
|
|
|
// FIXME this isn't actually creating records...?
|
|
|
|
let _: Option<db::Record> = match db.create("website").content(db::Website {
|
|
|
|
href: attr.value.to_string(),
|
2024-10-31 21:09:48 +00:00
|
|
|
crawled: false,
|
|
|
|
site: site_name.to_string()
|
2024-10-31 21:32:37 +00:00
|
|
|
}).await {
|
|
|
|
Ok(e) => {
|
|
|
|
if let Some(a) = &e {
|
|
|
|
debug!("{:?}", a);
|
|
|
|
}
|
|
|
|
e
|
|
|
|
},
|
|
|
|
Err(_) => todo!(),
|
|
|
|
};
|
2024-10-07 17:14:56 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
};
|
2024-10-04 19:52:34 +00:00
|
|
|
},
|
2024-10-31 21:09:48 +00:00
|
|
|
_ => {},
|
2024-10-04 19:52:34 +00:00
|
|
|
};
|
2024-08-23 11:22:49 +00:00
|
|
|
|
2024-10-31 20:10:14 +00:00
|
|
|
for child in node.children.borrow().iter() {
|
|
|
|
Box::pin(walk(child, db, site_name)).await;
|
2024-10-07 17:14:56 +00:00
|
|
|
}
|
2024-08-23 11:22:49 +00:00
|
|
|
}
|