wip: try to implement parallel crawl #8

Closed
nutomic wants to merge 2 commits from parallel-crawl into main
Showing only changes of commit 1bd510fbb8 - Show all commits

View file

@ -3,52 +3,74 @@ use crate::node_info::NodeInfo;
use crate::REQUEST_TIMEOUT;
use anyhow::anyhow;
use anyhow::Error;
use futures::try_join;
use futures::future::join_all;
use futures::stream::FuturesUnordered;
use futures::{stream, try_join, StreamExt, TryStreamExt};
use reqwest::Client;
use serde::Serialize;
use std::collections::VecDeque;
use std::future::Future;
pub async fn crawl(
start_instances: Vec<String>,
exclude: Vec<String>,
max_depth: i32,
) -> Result<(Vec<InstanceDetails>, i32), Error> {
let mut pending_instances: VecDeque<CrawlInstance> = start_instances
let mut pending_instances: VecDeque<CrawlInstanceTask> = start_instances
.iter()
.map(|s| CrawlInstance::new(s.to_string(), 0))
.map(|s| CrawlInstanceTask::new(s.to_string(), 0))
.collect();
let mut crawled_instances = vec![];
let mut instance_details = vec![];
let mut failed_instances = 0;
while let Some(current_instance) = pending_instances.pop_back() {
crawled_instances.push(current_instance.domain.clone());
if current_instance.depth > max_depth || exclude.contains(&current_instance.domain) {
continue;
}
match fetch_instance_details(&current_instance.domain).await {
Ok(details) => {
instance_details.push(details.to_owned());
for i in details.linked_instances {
let is_in_crawled = crawled_instances.contains(&i);
let is_in_pending = pending_instances.iter().any(|p| p.domain == i);
if !is_in_crawled && !is_in_pending {
let ci = CrawlInstance::new(i, current_instance.depth + 1);
pending_instances.push_back(ci);
}
}
}
Err(e) => {
failed_instances += 1;
eprintln!("Failed to crawl {}: {}", current_instance.domain, e)
}
}
}
//let mut failed_instances = 0;
let mut futures = stream::iter(pending_instances)
.then(|instance: CrawlInstanceTask| async {
crawled_instances.push(instance.domain.clone());
crawl_instance(instance, exclude.clone(), max_depth).await?
})
.flat_map(|(instance_details, depth)| {
let futures = instance_details
.linked_instances
.iter()
.map(|i| {
crawled_instances.push(i.clone());
crawl_instance(
CrawlInstanceTask::new(i.clone(), depth),
exclude.clone(),
max_depth,
)
})
.collect();
stream::iter(futures)
})
.collect::<FuturesUnordered<dyn Future<Output = Result<(InstanceDetails, i32), Error>>>>()
.await;
todo!()
/*
let mut crawl_result: Vec<InstanceDetails> = todo!();
// Sort by active monthly users descending
instance_details.sort_by_key(|i| i.users_active_month);
instance_details.reverse();
crawl_result.sort_by_key(|i| i.users_active_month);
crawl_result.reverse();
Ok((instance_details, failed_instances))
Ok((crawl_result, failed_instances))
*/
}
async fn crawl_instance(
current_instance: CrawlInstanceTask,
exclude: Vec<String>,
max_depth: i32,
) -> Result<(InstanceDetails, i32), Error> {
if current_instance.depth > max_depth || exclude.contains(&current_instance.domain) {
return Err(anyhow!("max depth reached"));
}
Ok((
fetch_instance_details(&current_instance.domain).await?,
current_instance.depth + 1,
))
}
#[derive(Serialize, Clone)]
@ -70,14 +92,14 @@ pub struct InstanceDetails {
pub linked_instances: Vec<String>,
}
struct CrawlInstance {
struct CrawlInstanceTask {
domain: String,
depth: i32,
}
impl CrawlInstance {
pub fn new(domain: String, depth: i32) -> CrawlInstance {
CrawlInstance { domain, depth }
impl CrawlInstanceTask {
pub fn new(domain: String, depth: i32) -> CrawlInstanceTask {
CrawlInstanceTask { domain, depth }
}
}