From 3ef28b5dd76249ae7e9f9fdc028e94edef566d0f Mon Sep 17 00:00:00 2001
From: Sebastian Nagel <sebastian@commoncrawl.org>
Date: Tue, 11 Nov 2025 15:45:52 +0100
Subject: [PATCH 1/2] BasicURLCanonicalizer: more efficient normalization of
 dots in host name

---
 .../archive/url/BasicURLCanonicalizer.java    | 35 +++++++++++++++++--
 1 file changed, 33 insertions(+), 2 deletions(-)

diff --git a/src/main/java/org/archive/url/BasicURLCanonicalizer.java b/src/main/java/org/archive/url/BasicURLCanonicalizer.java
index 37b448c1..fe2e0d42 100644
--- a/src/main/java/org/archive/url/BasicURLCanonicalizer.java
+++ b/src/main/java/org/archive/url/BasicURLCanonicalizer.java
@@ -34,7 +34,9 @@ public class BasicURLCanonicalizer implements URLCanonicalizer {
 			.compile("^(0[0-7]*)(\\.[0-7]+)?(\\.[0-7]+)?(\\.[0-7]+)?$");
 	Pattern DECIMAL_IP = Pattern
 			.compile("^([1-9][0-9]*)(\\.[0-9]+)?(\\.[0-9]+)?(\\.[0-9]+)?$");
+	Pattern MULTIDOT = Pattern.compile("\\.{2,}");
 
+	@Override
 	public void canonicalize(HandyURL url) {
 		url.setHash(null);
 		url.setAuthUser(minimalEscape(url.getAuthUser()));
@@ -55,8 +57,7 @@ public void canonicalize(HandyURL url) {
 				host = hostE;
 
 			}
-			host = host.replaceAll("^\\.+", "").replaceAll("\\.\\.+", ".")
-					.replaceAll("\\.$", "");
+			host = normalizeDots(host);
 		}
 
 		String ip = null;
@@ -74,6 +75,36 @@ public void canonicalize(HandyURL url) {
 		url.setPath(escapeOnce(normalizePath(path)));
 	}
 
+	/**
+	 * Normalize dots in the host name.
+	 *
+	 * @param host
+	 * @return host name with all sequences of dots replaced with a single dot,
+	 *         and all leading and trailing dots removed
+	 */
+	private String normalizeDots(String host) {
+		if (host.indexOf('.') == -1) {
+			return host;
+		}
+		int start = 0, end = host.length();
+		boolean changed = false;
+		while (host.charAt(start) == '.') {
+			start++;
+			changed = true;
+		}
+		while (host.charAt(end - 1) == '.') {
+			end--;
+			changed = true;
+		}
+		if (changed) {
+			host = host.substring(start, end);
+		}
+		if (host.contains("..")) {
+			host = MULTIDOT.matcher(host).replaceAll(".");
+		}
+		return host;
+	}
+
 	private static final Pattern SINGLE_FORWARDSLASH_PATTERN = Pattern
 			.compile("/");
 

From 44ec22772a24d5dc916b0b0730625988e20cc865 Mon Sep 17 00:00:00 2001
From: Sebastian Nagel <sebastian@commoncrawl.org>
Date: Thu, 13 Nov 2025 21:13:58 +0100
Subject: [PATCH 2/2] BasicURLCanonicalizer: more efficient normalization of
 dots in host name

Add unit test and prevent from StringIndexOutOfBoundsException.
---
 .../java/org/archive/url/BasicURLCanonicalizer.java    |  4 ++--
 .../org/archive/url/BasicURLCanonicalizerTest.java     | 10 ++++++++++
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/src/main/java/org/archive/url/BasicURLCanonicalizer.java b/src/main/java/org/archive/url/BasicURLCanonicalizer.java
index fe2e0d42..f2cee60f 100644
--- a/src/main/java/org/archive/url/BasicURLCanonicalizer.java
+++ b/src/main/java/org/archive/url/BasicURLCanonicalizer.java
@@ -88,11 +88,11 @@ private String normalizeDots(String host) {
 		}
 		int start = 0, end = host.length();
 		boolean changed = false;
-		while (host.charAt(start) == '.') {
+		while (start < end && host.charAt(start) == '.') {
 			start++;
 			changed = true;
 		}
-		while (host.charAt(end - 1) == '.') {
+		while (end > start && host.charAt(end - 1) == '.') {
 			end--;
 			changed = true;
 		}
diff --git a/src/test/java/org/archive/url/BasicURLCanonicalizerTest.java b/src/test/java/org/archive/url/BasicURLCanonicalizerTest.java
index 19b1984f..0ab1e3b9 100644
--- a/src/test/java/org/archive/url/BasicURLCanonicalizerTest.java
+++ b/src/test/java/org/archive/url/BasicURLCanonicalizerTest.java
@@ -286,6 +286,16 @@ public void testUnicodeEscaping() throws URISyntaxException {
 		checkCanonicalization("http://example.org/%F0%9F%82%A1", "http://example.org/%F0%9F%82%A1");
 	}
 	
+	@Test
+	public void testHostDots() throws URISyntaxException {
+		checkCanonicalization("https://foobar.org./", "https://foobar.org/");
+		checkCanonicalization("https://.foobar.org/", "https://foobar.org/");
+		checkCanonicalization("https://foo...bar.org/", "https://foo.bar.org/");
+		checkCanonicalization("https://...foo...bar.org.../", "https://foo.bar.org/");
+		checkCanonicalization("https://localhost/path/file.txt", "https://localhost/path/file.txt");
+		checkCanonicalization("https://....../path/file.txt", "https:///path/file.txt");
+	}
+
 	private void checkCanonicalization(String in, String want) throws URISyntaxException {
 		HandyURL h = URLParser.parse(in);
 		guc.canonicalize(h);