Merge pull request 'stream_response' (#6) from stream_response into main

Reviewed-on: #6

src/db.rs

@@ -43,7 +43,6 @@ impl Website {
     }
 
     pub fn set_crawled(&mut self) {
-        trace!("Set crawled to true");
         self.crawled = true
     }
 

src/filesystem.rs

@@ -1,30 +1,37 @@
-use std::{ffi::OsStr, io::ErrorKind, path::PathBuf};
+use std::{io::ErrorKind, path::PathBuf};
 
+use reqwest::header::HeaderValue;
 use tokio::fs;
-use tracing::{error, trace};
+use tracing::{error, trace, warn};
 use url::Url;
 
-pub fn as_path(url: &Url) -> PathBuf {
+pub fn as_path(url: &Url, content_type: &HeaderValue) -> PathBuf {
     // extract data from url to save it accurately
-    let url_path = PathBuf::from("./downloaded/".to_string() + url.domain().unwrap_or("UnknownDomain") + url.path());
+    let mut url_path = PathBuf::from("./downloaded/".to_string() + url.domain().unwrap_or("UnknownDomain") + url.path());
 
-    // if it's a file
-    let (basepath, filename) = if url_path.extension().filter(valid_file_extension).is_some() {
-        // get everything up till the file
-        let basepath = url_path.ancestors().skip(1).take(1).collect::<PathBuf>();
-        // get the file name
-        let filename = url_path.file_name().expect("This should exist").to_string_lossy();
-        trace!("Save path: {:?} and base path: {:?}", &url_path, &basepath);
-        (basepath, filename.to_string())
+    if let Ok(header) = content_type.to_str() {
+        // text/html; charset=UTF-8; option=value
+        let ttype = if let Some((t, _)) = header.split_once(';') {
+            t
+        } else {
+            header
+        };
+
+        if let Some((ttype, subtype)) = ttype.split_once('/') {
+            trace!("Found Content-Type to be: {ttype}/{subtype} for {}", url.to_string());
+            // If the Content-Type header is "*/html" (most likely "text/html") and the path's
+            // extension is anything but html:
+            if subtype=="html" && !url_path.extension().is_some_and(|f| f=="html" || f=="htm" ) {
+                // time to slap a index.html to the end of that path there!
+                url_path = url_path.join("index.html");
+            }
+        }
     } else {
-        (url_path.clone(), "index.html".into())
-    };
+        warn!("Header: {:?} couldn't be parsed into a string!", content_type);
+    }
+    trace!("Final path for {} is: {:?}", url, url_path);
 
-    let mut path = PathBuf::new();
-    path = path.join(basepath);
-    path = path.join(filename);
-
-    path
+    url_path
 }
 
 pub async fn init(filename: &PathBuf) -> Option<fs::File> {
@@ -50,29 +57,10 @@ pub async fn init(filename: &PathBuf) -> Option<fs::File> {
                     error!("Couldn't get file's parents: {:?}", &filename);
                 }
             } else {
-                error!("File creation: {err} {:?}", filename);
+                error!("File open error: {err} {:?}", filename);
             }
             // we don't care about other errors, we can't/shouldn't fix them
             None
         }
     }
 }
-
-fn valid_file_extension(take: &&OsStr) -> bool {
-    let los = take.to_string_lossy();
-    let all = los.split('.');
-    match all.last() {
-        Some(s) => {
-            // FIXME it's worth noting that the dumb tlds like .zip are in here,
-            // which could cause problems
-            let all_domains = include_str!("tlds-alpha-by-domain.txt");
-
-            // check if it is a domain
-            match all_domains.lines().map(str::to_lowercase).find(|x| x==s.to_lowercase().as_str()) {
-                Some(_) => false,
-                None => true
-            }
-        },
-        None => false,
-    }
-}

src/main.rs

@@ -19,8 +19,8 @@ use metrics::{counter, gauge};
 use metrics_exporter_prometheus::PrometheusBuilder;
 use serde::Deserialize;
 use surrealdb::{engine::remote::ws::Client, Surreal};
-use tokio::{io::AsyncWriteExt, task::JoinSet};
-use tracing::{debug, error, info, instrument, level_filters::LevelFilter, trace, trace_span};
+use tokio::{io::{AsyncWriteExt, BufWriter}, task::JoinSet};
+use tracing::{debug, error, info, instrument, level_filters::LevelFilter, trace, trace_span, warn};
 use tracing_subscriber::{fmt, layer::SubscriberExt, EnvFilter, Layer, Registry};
 
 mod db;
@@ -174,21 +174,41 @@ async fn process(mut site: Website, db: Surreal<Client>, reqwest: reqwest::Clien
     // Send the http request (get)
     if let Ok(response) = request_builder.send().await {
         // Get body from response
+        let headers = response.headers();
 
-        let path = filesystem::as_path(&site.site);
+        #[allow(non_snake_case)]
+        let CT = headers.get("Content-Type");
+        let ct = headers.get("content-type");
+
+        if CT.is_none() && ct.is_none() {
+        }
+        let ct = match (CT,ct) {
+            (None, None) => {
+                warn!("Server did not respond with Content-Type header. Url: {} Headers: ({:?})", site.site.to_string(), headers);
+                return
+            },
+            (None, Some(a)) => a,
+            (Some(a), None) => a,
+            (Some(a), Some(_)) => a,
+        };
+
+        let path = filesystem::as_path(&site.site, ct);
 
         // make sure that the file is good to go
-        if let Some(mut file) = filesystem::init(&path).await {
+        if let Some(file) = filesystem::init(&path).await {
             let should_parse = path.to_string_lossy().ends_with(".html");
             let mut buf: Vec<u8> = Vec::new();
 
+            let mut writer = BufWriter::new(file);
+
             // stream the response onto the disk
             let mut stream = response.bytes_stream();
+
+            info!("Writing at: {:?}", path);
             while let Some(data) = stream.next().await {
                 match data {
                     Ok(data) => {
-                        debug!("Writing at: {:?}", path);
-                        let _ = file.write_all(&data).await;
+                        let _ = writer.write_all(&data).await;
                         // If we are going to parse this file later, we will save it
                         // into memory as well as the disk.
                         if should_parse {
@@ -200,6 +220,7 @@ async fn process(mut site: Website, db: Surreal<Client>, reqwest: reqwest::Clien
                     },
                 }
             }
+            let _ = writer.flush();
 
             if should_parse {
                 // Parse document and get relationships

src/parser.rs

@@ -39,7 +39,7 @@ impl TokenSink for Website {
                                     if let Some(mut parsed) = url {
                                         parsed.set_query(None);
                                         parsed.set_fragment(None);
-                                        debug!("Final cleaned URL: `{}`", parsed.to_string());
+                                        trace!("Final cleaned URL: `{}`", parsed.to_string());
                                         let web = Website::new(&parsed.to_string(), false);
                                         links.push(web);
                                     }
@@ -127,16 +127,15 @@ fn try_get_url(parent: &Url, link: &str) -> Option<Url> {
                         }
                         let url = origin.clone() + link;
 
-                        trace!("Built `{url}` from `{origin} + `{}`", link.to_string());
-
                         if let Ok(url) = Url::parse(&url) {
-                            trace!("Saved relative url `{}` AS: `{}`", link, url);
+                            trace!("Built `{url}` from `{origin} + `{}`", link.to_string());
                             Some(url)
                         } else {
                             error!(
-                                "Failed to reconstruct a url from relative url: `{}` on site: `{}`",
+                                "Failed to reconstruct a url from relative url: `{}` on site: `{}`. Failed url was: {}",
                                 link,
-                                parent.to_string()
+                                parent.to_string(),
+                                url
                             );
                             None
                         }