From 347facfebf03f6ce404c372179f6d974bbcff19e Mon Sep 17 00:00:00 2001
From: Bryan Newbold <bnewbold@archive.org>
Date: Tue, 25 May 2021 17:11:57 -0700
Subject: download: fix wayback URL rewriting, and add a minimal test

---
 fatcat-cli/src/download.rs | 30 +++++++++++++++++++++++++++---
 1 file changed, 27 insertions(+), 3 deletions(-)

diff --git a/fatcat-cli/src/download.rs b/fatcat-cli/src/download.rs
index 4cb3143..44a11b7 100644
--- a/fatcat-cli/src/download.rs
+++ b/fatcat-cli/src/download.rs
@@ -84,16 +84,21 @@ impl<W: std::io::Write> std::io::Write for Sha1WriteWrapper<W> {
 }
 
 // eg, https://web.archive.org/web/20140802044207/http://www.geo.coop:80/sites/default/files/labs_of_oligarchy.pdf
-fn rewrite_wayback_url(url: Url) -> Result<Url> {
-    // TODO: make this function correct, and add tests
+fn rewrite_wayback_url(mut url: Url) -> Result<Url> {
     let mut segments: Vec<String> = url
         .path_segments()
         .unwrap()
         .map(|x| x.to_string())
         .collect();
-    if segments[0] == "web" && segments[1].len() == 14 {
+    if segments.len() < 3 || url.host_str() != Some("web.archive.org") {
+        return Err(anyhow!("not a valid wayback URL: {:?}", url));
+    }
+    // NOTE: the "12" digit timestamp here is to hack around a fraction of bad metadata in fatcat
+    // catalog circa 2019. should be cleaned up eventually then this can be deprecated.
+    if segments[0] == "web" && (segments[1].len() == 14 || segments[1].len() == 12) {
         segments[1] = format!("{}id_", segments[1]);
     }
+    url.set_path(&segments.join("/"));
     Ok(url)
 }
 
@@ -440,3 +445,22 @@ pub fn download_batch(
     done_receiver.recv()?;
     Ok(count)
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_rewrite_wayback_url() -> () {
+        assert_eq!(rewrite_wayback_url(Url::parse("https://web.archive.org/web/20140802044207/http://www.geo.coop:80/sites/default/files/labs_of_oligarchy.pdf").unwrap()).unwrap(),
+        Url::parse("https://web.archive.org/web/20140802044207id_/http://www.geo.coop:80/sites/default/files/labs_of_oligarchy.pdf").unwrap());
+
+        // not a wayback URL
+        assert!(rewrite_wayback_url(Url::parse("https://fatcat.wiki/blah.pdf").unwrap()).is_err());
+
+        // too short
+        assert!(
+            rewrite_wayback_url(Url::parse("https://web.archive.org/file.pdf").unwrap()).is_err()
+        );
+    }
+}
-- 
cgit v1.2.3