From e07ac16bc4cc97b8290e0e4500357a9573943bba Mon Sep 17 00:00:00 2001
From: Olivier 'reivilibre <olivier@librepush.net>
Date: Fri, 31 Mar 2023 22:59:23 +0100
Subject: [PATCH] Skip raking of weeded URLs

May be useful for retroactively clearing out URLs
---
 quickpeep_raker/src/raking/task.rs | 25 +++++++++++++++++++++++++
 quickpeep_raker/src/storage.rs     |  9 +++++++++
 2 files changed, 34 insertions(+)

diff --git a/quickpeep_raker/src/raking/task.rs b/quickpeep_raker/src/raking/task.rs
index 9bf1ec8..8c9dfdc 100644
--- a/quickpeep_raker/src/raking/task.rs
+++ b/quickpeep_raker/src/raking/task.rs
@@ -94,6 +94,17 @@ impl TaskContext {
         let mut current_robot_rules: Option<Cylon> = None;
         let mut wait_until: Option<Instant> = None;
 
+        let domain_record = {
+            let txn = self.store.ro_txn()?;
+            let dr = txn.get_domain_record(&domain)?;
+            match dr {
+                None => {
+                    return Ok(());
+                }
+                Some(dr) => dr,
+            }
+        };
+
         while !self.graceful_stop.load(Ordering::Relaxed) {
             // Get a URL to process
             let url = {
@@ -138,6 +149,20 @@ impl TaskContext {
             let url = Url::parse(&url_str)
                 .with_context(|| format!("failed to parse as URL: {url_str:?}"))?;
 
+            if !domain_record.is_url_rakeable(&url).unwrap_or(false) {
+                // This is now a weed: skip.
+                let domain = domain.clone();
+                let url = url.clone();
+                self.store
+                    .async_rw_txn(move |txn| {
+                        txn.dequeue_url(&domain, url.as_str())?;
+                        txn.commit()?;
+                        Ok(())
+                    })
+                    .await?;
+                continue;
+            }
+
             // Check our robot rules are valid for that URL.
             let robot_url = robots_txt_url_for(&url)
                 .with_context(|| format!("failed to get robots.txt URL for {url_str:?}"))?;
diff --git a/quickpeep_raker/src/storage.rs b/quickpeep_raker/src/storage.rs
index 71d0693..f9f6fdd 100644
--- a/quickpeep_raker/src/storage.rs
+++ b/quickpeep_raker/src/storage.rs
@@ -373,6 +373,15 @@ impl<'a> RakerTxn<'a, RW> {
         Ok(())
     }
 
+    /// Marks a URL as visited and takes it out of the queue.
+    pub fn dequeue_url(&self, domain: &str, url_str: &str) -> anyhow::Result<()> {
+        let queue_urls = &self.mdbx.borrow_dbs().queue_urls;
+        let queue_key = format!("{}\n{}", domain, url_str);
+        self.mdbx_txn.del(&queue_urls, queue_key.as_bytes(), None)?;
+
+        Ok(())
+    }
+
     pub fn start_backing_off(
         &self,
         domain: &str,