Mercurial > lbo > hg > scrapeprice
view src/driver.rs @ 19:b9af1d5065b4
Make driver traits async
author | Lewin Bormann <lbo@spheniscida.de> |
---|---|
date | Tue, 22 Sep 2020 19:01:57 +0200 |
parents | 1c464fb19d9f |
children | b16039ffcb17 |
line wrap: on
line source
#![allow(unused)] //! Drive the scraping process. use std::iter; use crate::err; use crate::extract; use crate::http; use hyper::Uri; use log::{info,warn,error}; /// Store fetched results, which come as key/value pairs, somewhere. #[async_trait::async_trait] pub trait Storage<T: Send> { async fn store(&mut self, d: Box<dyn Iterator<Item=T> + Send>) ->Result<(), err::HTTPError>; } /// Return Uris to explore, both as initial set and for every fetched page. #[async_trait::async_trait] pub trait Explorer { /// Return pages to fetch in any case, e.g. time-based. Called on every iteration of the /// driver. All returned Uris are appended to the queue. async fn idle(&mut self) -> Vec<Uri>; /// Return pages to fetch based on a fetched document. async fn next(&mut self, uri: &Uri, doc: &extract::Document) -> Vec<Uri>; } /// An Extractor retrieves information from a Document. pub trait Extractor<T: Send> { fn extract(&mut self, uri: &Uri, doc: &extract::Document) -> Vec<T> { vec![] } } /// DriverLogic holds the driven implementation. The members tell the driver what to fetch, and /// what and how to store it. pub struct DriverLogic<T> { pub explore: Box<dyn Explorer>, pub store: Box<dyn Storage<T>>, pub extract: Box<dyn Extractor<T>>, } pub struct Driver<T> { https: http::HTTPS, logic: DriverLogic<T>, // This could be made into a more elaborate scheduler. queue: Vec<Uri>, } impl<T: 'static + Send> Driver<T> { /// Create a new Driver instance. pub fn new(logic: DriverLogic<T>, https: Option<http::HTTPS>) -> Driver<T> { Driver { https: https.unwrap_or(http::HTTPS::new()), logic: logic, queue: Vec::with_capacity(64) } } /// Run Driver a single step, i.e. first explore, then process one page. Returns true if a page /// was processed. pub async fn drive(&mut self) -> Result<bool, err::HTTPError> { let new = self.logic.explore.idle().await; info!("Appended URIs to queue: {:?}", new); self.queue.extend(new.into_iter()); if let Some(uri) = self.queue.pop() { info!("Starting fetch of {}", uri); let resp = self.https.get(&uri).await?; let doc = extract::parse_response(resp)?; let extracted = self.logic.extract.extract(&uri, &doc); self.logic.store.store(Box::new(extracted.into_iter())); let next = self.logic.explore.next(&uri, &doc).await; info!("Appended URIs after fetch: {:?}", next); self.queue.extend(next); return Ok(true); } else { Ok(false) } } }