lemmy/crates/apub_lib/src/object_id.rs

202 lines
6 KiB
Rust
Raw Normal View History

2021-11-05 00:24:10 +00:00
use crate::{traits::ApubObject, APUB_JSON_CONTENT_TYPE};
use activitystreams::chrono::{Duration as ChronoDuration, NaiveDateTime, Utc};
use anyhow::anyhow;
use diesel::NotFound;
use lemmy_utils::{
request::{build_user_agent, retry},
settings::structs::Settings,
LemmyError,
};
use log::info;
use reqwest::{Client, StatusCode};
use serde::{Deserialize, Serialize};
use std::{
fmt::{Debug, Display, Formatter},
marker::PhantomData,
time::Duration,
};
use url::Url;
/// Maximum number of HTTP requests allowed to handle a single incoming activity (or a single object
/// fetch through the search). This should be configurable.
static REQUEST_LIMIT: i32 = 25;
lazy_static! {
static ref CLIENT: Client = Client::builder()
.user_agent(build_user_agent(&Settings::get()))
.build()
.unwrap();
}
#[derive(Clone, PartialEq, Serialize, Deserialize, Debug)]
#[serde(transparent)]
pub struct ObjectId<Kind>(Url, #[serde(skip)] PhantomData<Kind>)
where
Kind: ApubObject + Send + 'static,
for<'de2> <Kind as ApubObject>::ApubType: serde::Deserialize<'de2>;
impl<Kind> ObjectId<Kind>
where
Kind: ApubObject + Send + 'static,
for<'de2> <Kind as ApubObject>::ApubType: serde::Deserialize<'de2>,
{
pub fn new<T>(url: T) -> Self
where
T: Into<Url>,
{
ObjectId(url.into(), PhantomData::<Kind>)
}
pub fn inner(&self) -> &Url {
&self.0
}
/// Fetches an activitypub object, either from local database (if possible), or over http.
pub async fn dereference(
&self,
data: &<Kind as ApubObject>::DataType,
request_counter: &mut i32,
) -> Result<Kind, LemmyError> {
let db_object = self.dereference_from_db(data).await?;
// if its a local object, only fetch it from the database and not over http
if self.0.domain() == Some(&Settings::get().get_hostname_without_port()?) {
return match db_object {
None => Err(NotFound {}.into()),
Some(o) => Ok(o),
};
}
// object found in database
if let Some(object) = db_object {
// object is old and should be refetched
if let Some(last_refreshed_at) = object.last_refreshed_at() {
if should_refetch_object(last_refreshed_at) {
return self
.dereference_from_http(data, request_counter, Some(object))
.await;
}
}
Ok(object)
}
// object not found, need to fetch over http
else {
self
.dereference_from_http(data, request_counter, None)
.await
}
}
/// Fetch an object from the local db. Instead of falling back to http, this throws an error if
/// the object is not found in the database.
pub async fn dereference_local(
&self,
data: &<Kind as ApubObject>::DataType,
) -> Result<Kind, LemmyError> {
let object = self.dereference_from_db(data).await?;
object.ok_or_else(|| anyhow!("object not found in database {}", self).into())
}
/// returning none means the object was not found in local db
async fn dereference_from_db(
&self,
data: &<Kind as ApubObject>::DataType,
) -> Result<Option<Kind>, LemmyError> {
let id = self.0.clone();
ApubObject::read_from_apub_id(id, data).await
}
async fn dereference_from_http(
&self,
data: &<Kind as ApubObject>::DataType,
request_counter: &mut i32,
db_object: Option<Kind>,
) -> Result<Kind, LemmyError> {
// dont fetch local objects this way
debug_assert!(self.0.domain() != Some(&Settings::get().hostname));
info!("Fetching remote object {}", self.to_string());
*request_counter += 1;
if *request_counter > REQUEST_LIMIT {
return Err(LemmyError::from(anyhow!("Request limit reached")));
}
let res = retry(|| {
CLIENT
.get(self.0.as_str())
.header("Accept", APUB_JSON_CONTENT_TYPE)
.timeout(Duration::from_secs(60))
.send()
})
.await?;
if res.status() == StatusCode::GONE {
if let Some(db_object) = db_object {
db_object.delete(data).await?;
}
return Err(anyhow!("Fetched remote object {} which was deleted", self).into());
}
let res2: Kind::ApubType = res.json().await?;
Ok(Kind::from_apub(res2, data, self.inner(), request_counter).await?)
}
}
2021-11-05 00:24:10 +00:00
static ACTOR_REFETCH_INTERVAL_SECONDS: i64 = 24 * 60 * 60;
static ACTOR_REFETCH_INTERVAL_SECONDS_DEBUG: i64 = 10;
/// Determines when a remote actor should be refetched from its instance. In release builds, this is
/// `ACTOR_REFETCH_INTERVAL_SECONDS` after the last refetch, in debug builds
/// `ACTOR_REFETCH_INTERVAL_SECONDS_DEBUG`.
///
/// TODO it won't pick up new avatars, summaries etc until a day after.
/// Actors need an "update" activity pushed to other servers to fix this.
fn should_refetch_object(last_refreshed: NaiveDateTime) -> bool {
let update_interval = if cfg!(debug_assertions) {
// avoid infinite loop when fetching community outbox
ChronoDuration::seconds(ACTOR_REFETCH_INTERVAL_SECONDS_DEBUG)
} else {
ChronoDuration::seconds(ACTOR_REFETCH_INTERVAL_SECONDS)
};
let refresh_limit = Utc::now().naive_utc() - update_interval;
last_refreshed.lt(&refresh_limit)
}
impl<Kind> Display for ObjectId<Kind>
where
Kind: ApubObject + Send + 'static,
for<'de2> <Kind as ApubObject>::ApubType: serde::Deserialize<'de2>,
{
2021-11-05 00:24:10 +00:00
#[allow(clippy::to_string_in_display)]
fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
2021-11-05 00:24:10 +00:00
// Use to_string here because Url.display is not useful for us
write!(f, "{}", self.0.to_string())
}
}
impl<Kind> From<ObjectId<Kind>> for Url
where
Kind: ApubObject + Send + 'static,
for<'de2> <Kind as ApubObject>::ApubType: serde::Deserialize<'de2>,
{
fn from(id: ObjectId<Kind>) -> Self {
id.0
}
}
2021-11-05 00:24:10 +00:00
#[cfg(test)]
mod tests {
use super::*;
use crate::object_id::should_refetch_object;
#[test]
fn test_should_refetch_object() {
let one_second_ago = Utc::now().naive_utc() - ChronoDuration::seconds(1);
assert!(!should_refetch_object(one_second_ago));
let two_days_ago = Utc::now().naive_utc() - ChronoDuration::days(2);
assert!(should_refetch_object(two_days_ago));
}
}