morrigan/crabbysearch
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

add code to evade ip blocking, improve pagination code and fix documentation

Browse source
This commit is contained in:
neon_arch 2023-05-02 11:58:21 +03:00
parent f8c3c8dcbe
commit c170de8194
14 changed files with 264 additions and 61 deletions
Download patch file Download diff file Expand all files Collapse all files

44
src/engines/duckduckgo.rs
View file

@ -2,9 +2,10 @@
//! by querying the upstream duckduckgo search engine with user provided query and with a page
//! number if provided.
use std::collections::HashMap;
use std::{collections::HashMap, time::Duration};
use reqwest::header::USER_AGENT;
use rand::Rng;
use reqwest::header::{HeaderMap, CONTENT_TYPE, REFERER, USER_AGENT};
use scraper::{Html, Selector};
use crate::search_results_handler::aggregation_models::RawSearchResult;
@ -17,7 +18,7 @@ use crate::search_results_handler::aggregation_models::RawSearchResult;
/// # Arguments
///
/// * `query` - Takes the user provided query to query to the upstream search engine with.
/// * `page` - Takes an Option<u32> as argument which can be either None or a valid page number.
/// * `page` - Takes an u32 as an argument.
/// * `user_agent` - Takes a random user agent string as an argument.
///
/// # Errors
@ -27,32 +28,41 @@ use crate::search_results_handler::aggregation_models::RawSearchResult;
/// selector fails to initialize"
pub async fn results(
query: &str,
page: Option<u32>,
page: u32,
user_agent: &str,
) -> Result<HashMap<String, RawSearchResult>, Box<dyn std::error::Error>> {
// Page number can be missing or empty string and so appropriate handling is required
// so that upstream server recieves valid page number.
let url: String = match page {
Some(page_number) => {
if page_number <= 1 {
format!("https://html.duckduckgo.com/html/?q={query}&s=&dc=&v=1&o=json&api=/d.js")
} else {
format!(
"https://duckduckgo.com/html/?q={}&s={}&dc={}&v=1&o=json&api=/d.js",
query,
page_number / 2 * 30,
page_number / 2 * 30 + 1
)
}
1 => {
format!("https://html.duckduckgo.com/html/?q={query}&s=&dc=&v=1&o=json&api=/d.js")
}
_ => {
format!(
"https://duckduckgo.com/html/?q={}&s={}&dc={}&v=1&o=json&api=/d.js",
query,
(page / 2 + (page % 2)) * 30,
(page / 2 + (page % 2)) * 30 + 1
)
}
None => format!("https://html.duckduckgo.com/html/?q={query}&s=&dc=&v=1&o=json&api=/d.js"),
};
// Add a random delay before making the request.
let mut rng = rand::thread_rng();
let delay_secs = rng.gen_range(1, 10);
std::thread::sleep(Duration::from_secs(delay_secs));
// initializing HeaderMap and adding appropriate headers.
let mut header_map = HeaderMap::new();
header_map.insert(USER_AGENT, user_agent.parse()?);
header_map.insert(REFERER, "https://google.com/".parse()?);
header_map.insert(CONTENT_TYPE, "text/html; charset=UTF-8".parse()?);
// fetch the html from upstream duckduckgo engine
// TODO: Write better error handling code to handle no results case.
let results: String = reqwest::Client::new()
.get(url)
.header(USER_AGENT, user_agent)
.headers(header_map) // add spoofed headers to emulate human behaviour
.send()
.await?
.text()