Support canonical URLs and pass down is_cf

quickpeep/src/raking.rs

@@ -153,11 +153,11 @@ impl Raker {
     ) -> anyhow::Result<RakeOutcome> {
         let response = client.get(url.clone()).send().await?;
 
-        if let Some(remote_addr) = response.remote_addr() {
-            eprintln!("rA {:?}", remote_addr);
-            let is_cf = self.antifeature_ip_set.contains(remote_addr.ip());
-            eprintln!("CF? {:?}", is_cf);
-        }
+        let is_cf = if let Some(remote_addr) = response.remote_addr() {
+            self.antifeature_ip_set.contains(remote_addr.ip())
+        } else {
+            false
+        };
 
         let http_code = response.status().as_u16();
 
@@ -213,9 +213,9 @@ impl Raker {
 
         if content_type == "text/html" && (intent == RakeIntent::Any || intent == RakeIntent::Page)
         {
-            match self.rake_html_page(&content, url) {
+            match self.rake_html_page(&content, url, is_cf) {
                 Ok(page_rake) => {
-                    return Ok(RakeOutcome::RakedPage(page_rake));
+                    return Ok(page_rake);
                 }
                 Err(error) => {
                     debug!("Failed to rake HTML page: {:?}", error);
@@ -254,13 +254,33 @@ impl Raker {
         }));
     }
 
-    pub fn rake_html_page(&self, content: &[u8], url: &Url) -> anyhow::Result<RakedPage> {
+    pub fn rake_html_page(&self, content: &[u8], url: &Url, is_cf: bool) -> anyhow::Result<RakeOutcome> {
         let content_str = std::str::from_utf8(content)?;
 
         let root_node: NodeRef = kuchiki::parse_html().one(content_str);
 
+        // See whether this page is at the canonical URL for the page.
+        // If it's not, then we redirect the raker to the canonical URL.
+        if let Ok(canonical_link_node) = root_node.select_first("head link[rel=canonical]") {
+            if let Some(canonical_href) = canonical_link_node.attributes.borrow().get("href") {
+                let canonical_url = url.join(canonical_href)
+                    .context("Failed to resolve or parse canonical URL")?;
+
+                if &canonical_url != url {
+                    return Ok(RakeOutcome::Redirect {
+                        reason: RedirectReason::NotCanonical,
+                        new_url: canonical_url
+                    });
+                }
+            }
+        }
+
         let mut antifeature_flags = AnalysisAntifeatures::empty();
 
+        if is_cf {
+            antifeature_flags |= AnalysisAntifeatures::CLOUDFLARE;
+        }
+
         for (engine_antifeature_flag, adblock_engine) in &self.adblock_engines {
             match analyse_with_ad_block_cosmetic_filter(
                 &root_node,
@@ -297,9 +317,9 @@ impl Raker {
         let bare_size = serde_bare::to_vec(&dense_doc)?.len();
         eprintln!("CS {:?} → {:?}", content.len(), bare_size);
 
-        Ok(RakedPage {
+        Ok(RakeOutcome::RakedPage(RakedPage {
             // TODO
-        })
+        }))
     }
 }