use crate::{
command,
sql::Db,
tg_bot::Tg,
};
use std::{
borrow::Cow,
collections::{
BTreeMap,
HashSet,
},
sync::Arc,
};
use async_compat::Compat;
use chrono::{
DateTime,
Local,
};
use lazy_static::lazy_static;
use regex::Regex;
use reqwest::header::LAST_MODIFIED;
use smol::{
Timer,
lock::Mutex,
};
use tgbot::{
handler::UpdateHandler,
types::{
ChatPeerId,
Command,
ParseMode,
Update,
UpdateType,
UserPeerId,
},
};
use stacked_errors::{
Result,
StackableErr,
anyhow,
bail,
};
lazy_static!{
pub static ref RE_SPECIAL: Regex = Regex::new(r"([\-_*\[\]()~`>#+|{}\.!])").unwrap();
}
/// Escape characters that are special in Telegram MarkdownV2 by prefixing them with a backslash.
///
/// This ensures the returned string can be used as MarkdownV2-formatted Telegram message content
/// without special characters being interpreted as MarkdownV2 markup.
pub fn encode (text: &str) -> Cow<'_, str> {
RE_SPECIAL.replace_all(text, "\\$1")
}
// This one does nothing except making sure only one token exists for each id
pub struct Token {
running: Arc<Mutex<HashSet<i32>>>,
my_id: i32,
}
impl Token {
/// Attempts to acquire a per-id token by inserting `my_id` into the shared `running` set.
///
/// If the id was not already present, the function inserts it and returns `Some(Token)`.
/// When the returned `Token` is dropped, the id will be removed from the `running` set,
/// allowing subsequent acquisitions for the same id.
///
/// # Parameters
///
/// - `running`: Shared set tracking active ids.
/// - `my_id`: Identifier to acquire a token for.
///
/// # Returns
///
/// `Ok(Token)` if the id was successfully acquired, `Error` if a token for the id is already active.
async fn new (running: &Arc<Mutex<HashSet<i32>>>, my_id: i32) -> Result<Token> {
let running = running.clone();
let mut set = running.lock_arc().await;
if set.contains(&my_id) {
bail!("Token already taken");
} else {
set.insert(my_id);
Ok(Token {
running,
my_id,
})
}
}
}
impl Drop for Token {
/// Releases this token's claim on the shared running-set when the token is dropped.
///
/// The token's identifier is removed from the shared `running` set so that future
/// operations for the same id may proceed.
///
/// TODO: is using block_on inside block_on safe? Currently tested and working fine.
fn drop (&mut self) {
smol::block_on(async {
let mut set = self.running.lock_arc().await;
set.remove(&self.my_id);
})
}
}
#[derive(Clone)]
pub struct Core {
pub tg: Tg,
pub db: Db,
running: Arc<Mutex<HashSet<i32>>>,
http_client: reqwest::Client,
}
pub struct Post {
uri: String,
_title: String,
_authors: String,
_summary: String,
}
impl Core {
/// Create a Core instance from configuration and start its background autofetch loop.
///
/// The provided `settings` must include:
/// - `owner` (integer): default chat id to use as the owner/destination,
/// - `api_key` (string): Telegram bot API key,
/// - `api_gateway` (string): Telegram API gateway host,
/// - `pg` (string): PostgreSQL connection string,
/// - optional `proxy` (string): proxy URL for the HTTP client.
///
/// On success returns an initialized `Core` with Telegram and HTTP clients, database connection,
/// an empty running set for per-id tokens, and a spawned background task that periodically runs
/// `autofetch`. If any required setting is missing or initialization fails, an error is returned.
pub async fn new(settings: config::Config) -> Result<Core> {
let mut client = reqwest::Client::builder();
if let Ok(proxy) = settings.get_string("proxy") {
let proxy = reqwest::Proxy::all(proxy).stack()?;
client = client.proxy(proxy);
}
let core = Core {
tg: Tg::new(&settings).await.stack()?,
db: Db::new(&settings.get_string("pg").stack()?)?,
running: Arc::new(Mutex::new(HashSet::new())),
http_client: client.build().stack()?,
};
let clone = core.clone();
smol::spawn(Compat::new(async move {
loop {
let delay = match &clone.autofetch().await {
Err(err) => {
if let Err(err) = clone.tg.send(format!("🛑 {err}"), None, None).await {
eprintln!("Autofetch error: {err:?}");
};
std::time::Duration::from_secs(60)
},
Ok(time) => *time,
};
Timer::after(delay).await;
}
})).detach();
Ok(core)
}
/// Fetches the feed for a source, sends any newly discovered posts to the appropriate chat, and records them in the database.
///
/// This acquires a per-source guard to prevent concurrent checks for the same `id`. If a check is already running for
/// the given `id`, the function returns an error. If `last_scrape` is provided, it is sent as the `If-Modified-Since`
/// header to the feed request. The function parses RSS or Atom feeds, sends unseen post URLs to either the source's
/// channel (when `real` is true) or the source owner (when `real` is false), and persists posted entries so they are
/// not reposted later.
///
/// Parameters:
/// - `id`: Identifier of the source to check.
/// - `real`: When `true`, send posts to the source's channel; when `false`, send to the source owner.
/// - `last_scrape`: Optional timestamp used to set the `If-Modified-Since` header for the HTTP request.
///
/// # Returns
///
/// `Posted: N` where `N` is the number of posts processed and sent.
pub async fn check (&self, id: i32, real: bool, last_scrape: Option<DateTime<Local>>) -> Result<String> {
let mut posted: i32 = 0;
let mut conn = self.db.begin().await.stack()?;
let _token = Token::new(&self.running, id).await.stack()?;
let source = conn.get_source(id, self.tg.owner).await.stack()?;
conn.set_scrape(id).await.stack()?;
let destination = ChatPeerId::from(match real {
true => source.channel_id,
false => source.owner,
});
let mut this_fetch: Option<DateTime<chrono::FixedOffset>> = None;
let mut posts: BTreeMap<DateTime<chrono::FixedOffset>, Post> = BTreeMap::new();
let mut builder = self.http_client.get(&source.url);
if let Some(last_scrape) = last_scrape {
builder = builder.header(LAST_MODIFIED, last_scrape.to_rfc2822());
};
let response = builder.send().await.stack()?;
#[cfg(debug_assertions)]
{
use reqwest::header::{
CACHE_CONTROL,
EXPIRES,
};
let headers = response.headers();
let expires = headers.get(EXPIRES);
let cache = headers.get(CACHE_CONTROL);
if expires.is_some() || cache.is_some() {
println!("{} {} {:?} {:?} {:?}", Local::now().to_rfc2822(), &source.url, last_scrape, expires, cache);
}
}
let status = response.status();
let content = response.bytes().await.stack()?;
match rss::Channel::read_from(&content[..]) {
Ok(feed) => {
for item in feed.items() {
if let Some(link) = item.link() {
let date = match item.pub_date() {
Some(feed_date) => DateTime::parse_from_rfc2822(feed_date),
None => DateTime::parse_from_rfc3339(match item.dublin_core_ext() {
Some(ext) => {
let dates = ext.dates();
if dates.is_empty() {
bail!("Feed item has Dublin Core extension but no dates.")
} else {
&dates[0]
}
},
None => bail!("Feed item misses posting date."),
}),
}.stack()?;
posts.insert(date, Post{
uri: link.to_string(),
_title: item.title().unwrap_or("").to_string(),
_authors: item.author().unwrap_or("").to_string(),
_summary: item.content().unwrap_or("").to_string(),
});
}
};
},
Err(err) => match err {
rss::Error::InvalidStartTag => {
match atom_syndication::Feed::read_from(&content[..]) {
Ok(feed) => {
for item in feed.entries() {
let date = item.published()
.stack_err("Feed item missing publishing date.")?;
let uri = {
let links = item.links();
if links.is_empty() {
bail!("Feed item missing post links.");
} else {
links[0].href().to_string()
}
};
let _authors = item.authors().iter().map(|x| format!("{} <{:?}>", x.name(), x.email())).collect::<Vec<String>>().join(", ");
let _summary = if let Some(sum) = item.summary() { sum.value.clone() } else { String::new() };
posts.insert(*date, Post{
uri,
_title: item.title().to_string(),
_authors,
_summary,
});
};
},
Err(err) => {
bail!("Unsupported or mangled content:\n{:?}\n{err}\n{status:#?}\n", &source.url)
},
}
},
rss::Error::Eof => (),
_ => bail!("Unsupported or mangled content:\n{:?}\n{err}\n{status:#?}\n", &source.url)
}
};
for (date, post) in posts.iter() {
let post_url: Cow<str> = match source.url_re {
Some(ref x) => sedregex::ReplaceCommand::new(x).stack()?.execute(&post.uri),
None => post.uri.clone().into(),
};
if ! conn.exists(&post_url, id).await.stack()? {
if this_fetch.is_none() || *date > this_fetch.unwrap() {
this_fetch = Some(*date);
};
self.tg.send( match &source.iv_hash {
Some(hash) => format!("<a href=\"https://t.me/iv?url={post_url}&rhash={hash}\"> </a>{post_url}"),
None => format!("{post_url}"),
}, Some(destination), Some(ParseMode::Html)).await.stack()?;
conn.add_post(id, date, &post_url).await.stack()?;
posted += 1;
};
};
posts.clear();
Ok(format!("Posted: {posted}"))
}
/// Determine the delay until the next scheduled fetch and spawn background checks for any overdue sources.
///
/// This scans the database queue, spawns background tasks to run checks for sources whose `next_fetch`
/// is in the past (each task uses a Core clone with the appropriate owner), and computes the shortest
/// duration until the next `next_fetch`.
async fn autofetch(&self) -> Result<std::time::Duration> {
let mut delay = chrono::Duration::minutes(1);
let now = chrono::Local::now();
let queue = {
let mut conn = self.db.begin().await.stack()?;
conn.get_queue().await.stack()?
};
for row in queue {
if let Some(next_fetch) = row.next_fetch {
if next_fetch < now {
if let (Some(owner), Some(source_id), last_scrape) = (row.owner, row.source_id, row.last_scrape) {
let clone = Core {
tg: self.tg.with_owner(owner),
..self.clone()
};
let source = {
let mut conn = self.db.begin().await.stack()?;
match conn.get_one(owner, source_id).await {
Ok(Some(source)) => source.to_string(),
Ok(None) => "Source not found in database?".to_string(),
Err(err) => format!("Failed to fetch source data:\n{err}"),
}
};
smol::spawn(Compat::new(async move {
if let Err(err) = clone.check(source_id, true, Some(last_scrape)).await
&& let Err(err) = clone.tg.send(&format!("🛑 {source}\n{}", encode(&err.to_string())), None, Some(ParseMode::MarkdownV2)).await
{
eprintln!("Check error: {err}");
};
})).detach();
}
} else if next_fetch - now < delay {
delay = next_fetch - now;
}
}
};
delay.to_std().stack()
}
pub async fn list (&self, owner: UserPeerId) -> Result<String> {
let mut reply: Vec<String> = vec![];
reply.push("Channels:".into());
let mut conn = self.db.begin().await.stack()?;
for row in conn.get_list(owner).await.stack()? {
reply.push(row.to_string());
};
Ok(reply.join("\n\n"))
}
}
impl UpdateHandler for Core {
/// Dispatches an incoming Telegram update to a matching command handler and reports handler errors to the originating chat.
///
/// This method inspects the update; if it contains a message that can be parsed as a bot command,
/// it executes the corresponding command handler. If the handler returns an error, the error text
/// is sent back to the message's chat using MarkdownV2 formatting. Unknown commands produce an error
/// which is also reported to the chat.
async fn handle (&self, update: Update) {
if let UpdateType::Message(msg) = update.update_type
&& let Ok(cmd) = Command::try_from(msg)
{
let msg = cmd.get_message();
let words = cmd.get_args();
let command = cmd.get_name();
let res = match command {
"/check" | "/clean" | "/enable" | "/delete" | "/disable" => command::command(self, command, msg, words).await,
"/start" => command::start(self, msg).await,
"/list" => command::list(self, msg).await,
"/add" | "/update" => command::update(self, command, msg, words).await,
any => Err(anyhow!("Unknown command: {any}")),
};
if let Err(err) = res
&& let Err(err2) = self.tg.send(format!("\\#error\n```\n{err}\n```"),
Some(msg.chat.get_id()),
Some(ParseMode::MarkdownV2)
).await
{
dbg!(err2);
}
} // TODO: debug log for skipped updates?;
}
}