internet_mapper/src/main.rs

96 lines
3.3 KiB
Rust
Raw Normal View History

extern crate markup5ever_rcdom as rcdom;
extern crate html5ever;
2024-08-23 11:22:49 +00:00
2024-10-07 17:14:56 +00:00
use std::{env, rc::Rc};
use db::connect;
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};
use tracing::{debug, error, info, trace, 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
#[tokio::main]
async fn main() {
tracing_subscriber::fmt::init();
debug!("Starting...");
let url = "https://oliveratkinson.net";
let budget = "10";
2024-10-07 17:14:56 +00:00
let dom = get(url).await;
let db = connect().await.expect("Failed to connect to db, aborting.");
warn!("Walking...");
walk(&dom, &db, url).await;
}
async fn get(url: &str) -> Rc<Node> {
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
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) {
// Should protentailly just check for the record first.
2024-10-31 20:10:14 +00:00
let created: Option<db::Record> = match db.create("website").content(db::Website { href: String::from("/"), crawled: true, site: site_name.to_string() } ).await {
2024-10-07 17:14:56 +00:00
Ok(e) => e,
Err(e) => {
match e {
surrealdb::Error::Db(_) => todo!(),
surrealdb::Error::Api(api) => {
match api {
surrealdb::error::Api::Query(query) => {
error!(query);
None
},
_ => todo!(),
}
},
}
}
};
2024-10-31 20:10:14 +00:00
info!{"{:?}", created};
2024-10-07 17:14:56 +00:00
match &node.data {
rcdom::NodeData::Document => (),
2024-10-31 20:10:14 +00:00
rcdom::NodeData::ProcessingInstruction { target, contents } => debug!("ProcessingInstruction"),
rcdom::NodeData::Doctype { name, public_id, system_id } => debug!("doctype"),
rcdom::NodeData::Text { contents } => {},
rcdom::NodeData::Comment { contents } => debug!("comment"),
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() {
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 20:10:14 +00:00
let created: Option<db::Record> = db.create("email").content(db::Email { email: attr.value.to_string() }).await.unwrap();
2024-10-07 17:14:56 +00:00
info!("{:?}", created)
} else {
2024-10-31 20:10:14 +00:00
// Every not-mailto link
let created: Option<db::Record> = db.create("website").content(db::Website { href: attr.value.to_string(), crawled: false, site: site_name.to_string() } ).await.unwrap();
info!{"{:?}", created};
2024-10-07 17:14:56 +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
}