151 lines
6.1 KiB
Rust
151 lines
6.1 KiB
Rust
use std::default::Default;
|
|
|
|
use html5ever::tokenizer::{BufferQueue, TokenizerResult};
|
|
use html5ever::tokenizer::{StartTag, TagToken};
|
|
use html5ever::tokenizer::{Token, TokenSink, TokenSinkResult, Tokenizer, TokenizerOpts};
|
|
use html5ever::{local_name, tendril::*};
|
|
use tracing::{error, instrument, trace, warn};
|
|
use url::Url;
|
|
|
|
use crate::db::Website;
|
|
|
|
impl TokenSink for Website {
|
|
type Handle = Vec<Website>;
|
|
|
|
fn process_token(&self, token: Token, _line_number: u64) -> TokenSinkResult<Self::Handle> {
|
|
match token {
|
|
TagToken(tag) => {
|
|
if tag.kind == StartTag {
|
|
match tag.name {
|
|
// this should be all the html elements that have links
|
|
local_name!("a")
|
|
| local_name!("audio")
|
|
| local_name!("area")
|
|
| local_name!("img")
|
|
| local_name!("link")
|
|
| local_name!("object")
|
|
| local_name!("source")
|
|
| local_name!("base")
|
|
| local_name!("video") => {
|
|
let mut links = Vec::new();
|
|
for attr in &tag.attrs {
|
|
let attr_name = attr.name.local.to_string();
|
|
if attr_name == "src" || attr_name == "href" || attr_name == "data"
|
|
{
|
|
trace!(url = self.site.as_str(),"Found `{}` in html `{}` tag", &attr.value, tag.name);
|
|
let url = try_get_url(&self.site, &attr.value);
|
|
|
|
if let Some(mut parsed) = url {
|
|
parsed.set_query(None);
|
|
parsed.set_fragment(None);
|
|
trace!(url = self.site.as_str(), "Final cleaned URL: `{}`", parsed.to_string());
|
|
let web = Website::new(&parsed.to_string(), false);
|
|
links.push(web);
|
|
}
|
|
}
|
|
}
|
|
return TokenSinkResult::Script(links);
|
|
}
|
|
local_name!("button") | local_name!("meta") | local_name!("iframe") => {
|
|
// dbg!(attrs);
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
}
|
|
_ => {}
|
|
}
|
|
TokenSinkResult::Continue
|
|
}
|
|
}
|
|
|
|
#[instrument(skip(data))]
|
|
/// Parses the passed site and returns all the sites it links to.
|
|
pub async fn parse(site: &Website, data: &[u8]) -> Vec<Website> {
|
|
trace!(url = site.site.as_str(), "Parsing {}", site.site.to_string());
|
|
// prep work
|
|
let mut other_sites: Vec<Website> = Vec::new();
|
|
|
|
// change data into something that can be tokenized
|
|
let s: Result<Tendril<fmt::UTF8>, ()> = Tendril::try_from_byte_slice(data);
|
|
if let Ok(chunk) = s {
|
|
// create buffer of tokens and push our input into it
|
|
let token_buffer = BufferQueue::default();
|
|
token_buffer.push_back(
|
|
chunk
|
|
.try_reinterpret::<fmt::UTF8>()
|
|
.expect("Failed to reinterpret chunk!"),
|
|
);
|
|
// create the tokenizer
|
|
let tokenizer = Tokenizer::new(site.clone(), TokenizerOpts::default());
|
|
|
|
// go thru buffer
|
|
while let TokenizerResult::Script(mut sites) = tokenizer.feed(&token_buffer) {
|
|
other_sites.append(&mut sites);
|
|
// other_sites.push(sites);
|
|
}
|
|
assert!(token_buffer.is_empty());
|
|
tokenizer.end();
|
|
} else {
|
|
warn!(url = site.site.as_str(), "Tendril failed to parse on: {}", site.site.to_string());
|
|
}
|
|
|
|
other_sites
|
|
}
|
|
|
|
#[instrument]
|
|
fn try_get_url(parent: &Url, link: &str) -> Option<Url> {
|
|
match Url::parse(link) {
|
|
Ok(ok) => Some(ok),
|
|
Err(e) => {
|
|
if link.starts_with('#') {
|
|
trace!(url = parent.as_str(), "Rejecting # url");
|
|
None
|
|
} else if link.starts_with("//") {
|
|
// if a url starts with "//" is assumed that it will adopt
|
|
// the same scheme as it's parent
|
|
// https://stackoverflow.com/questions/9646407/two-forward-slashes-in-a-url-src-href-attribute
|
|
let scheme = parent.scheme();
|
|
|
|
match Url::parse(&format!("{scheme}://{link}")) {
|
|
Ok(url) => Some(url),
|
|
Err(err) => {
|
|
error!("Failed parsing relative scheme url: {}", err);
|
|
None
|
|
}
|
|
}
|
|
} else {
|
|
// # This is some sort of relative url, gonna try patching it up into an absolute
|
|
// url
|
|
match e {
|
|
url::ParseError::RelativeUrlWithoutBase => {
|
|
// Is: scheme://host:port
|
|
let mut origin = parent.origin().ascii_serialization();
|
|
if !origin.ends_with('/') && !link.starts_with('/') {
|
|
origin += "/";
|
|
}
|
|
let url = origin.clone() + link;
|
|
|
|
if let Ok(url) = Url::parse(&url) {
|
|
trace!(url = parent.as_str(), "Built `{url}` from `{origin} + `{}`", link.to_string());
|
|
Some(url)
|
|
} else {
|
|
error!(
|
|
"Failed to reconstruct a url from relative url: `{}` on site: `{}`. Failed url was: {}",
|
|
link,
|
|
parent.to_string(),
|
|
url
|
|
);
|
|
None
|
|
}
|
|
}
|
|
_ => {
|
|
error!("MISC error: {:?} {:?}", e, link);
|
|
None
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|