fix quadratic time when removing doctype declarations

commafeed-server/src/main/java/com/commafeed/backend/feed/parser/XMLCleaner.java

@@ -8,12 +8,13 @@ import org.apache.commons.lang3.StringUtils;
 import org.jdom2.Verifier;
 
 import java.util.Collection;
+import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 
 @Singleton
 public class XMLCleaner {
 
-    private static final Pattern DOCTYPE_PATTERN = Pattern.compile("<!DOCTYPE[^>]*>", Pattern.CASE_INSENSITIVE);
+    private static final Pattern DOCTYPE_START_PATTERN = Pattern.compile("<!DOCTYPE", Pattern.CASE_INSENSITIVE);
 
     private final Trie trie = Trie.builder().ignoreOverlaps().addKeywords(HtmlEntities.HTML_ENTITIES).build();
 
@@ -76,6 +77,22 @@ public class XMLCleaner {
             return null;
         }
 
-        return DOCTYPE_PATTERN.matcher(xml).replaceAll("");
+        // Equivalent to removing all matches of "<!DOCTYPE[^>]*>", without the quadratic cost of that regex when the
+        // input contains many "<!DOCTYPE" and no closing bracket after them.
+        Matcher matcher = DOCTYPE_START_PATTERN.matcher(xml);
+        StringBuilder sb = new StringBuilder(xml.length());
+
+        int pos = 0;
+        while (matcher.find(pos)) {
+            int end = xml.indexOf('>', matcher.end());
+            if (end < 0) {
+                break;
+            }
+
+            sb.append(xml, pos, matcher.start());
+            pos = end + 1;
+        }
+
+        return sb.append(xml, pos, xml.length()).toString();
     }
 }

commafeed-server/src/test/java/com/commafeed/backend/feed/parser/XMLCleanerTest.java

@@ -4,6 +4,8 @@ import org.junit.jupiter.api.Assertions;
 import org.junit.jupiter.api.Nested;
 import org.junit.jupiter.api.Test;
 
+import java.time.Duration;
+
 class XMLCleanerTest {
 
     XMLCleaner xmlCleaner = new XMLCleaner();
@@ -279,5 +281,39 @@ class XMLCleanerTest {
             String source = "<!DOCTYPE   html   ><html></html>";
             Assertions.assertEquals("<html></html>", xmlCleaner.removeDoctypeDeclarations(source));
         }
+
+        @Test
+        void preservesUnterminatedDoctype() {
+            String source = "<html></html><!DOCTYPE html";
+            Assertions.assertEquals(source, xmlCleaner.removeDoctypeDeclarations(source));
+        }
+
+        @Test
+        void removesTerminatedDoctypeAndPreservesUnterminatedDoctype() {
+            String source = "<!DOCTYPE html><html></html><!DOCTYPE html";
+            Assertions.assertEquals("<html></html><!DOCTYPE html", xmlCleaner.removeDoctypeDeclarations(source));
+        }
+
+        @Test
+        void removesNestedDoctypeUpToFirstClosingBracket() {
+            String source = "<!DOCTYPE<!DOCTYPE html><html></html>";
+            Assertions.assertEquals("<html></html>", xmlCleaner.removeDoctypeDeclarations(source));
+        }
+
+        @Test
+        void handlesManyUnterminatedDoctypesInLinearTime() {
+            String source = "<!DOCTYPE".repeat(64000);
+            Assertions.assertTimeout(
+                    Duration.ofSeconds(2),
+                    () -> Assertions.assertEquals(source, xmlCleaner.removeDoctypeDeclarations(source)));
+        }
+
+        @Test
+        void handlesManyUnterminatedDoctypesAfterClosingBracketInLinearTime() {
+            String source = "<a>" + "<!DOCTYPE".repeat(64000);
+            Assertions.assertTimeout(
+                    Duration.ofSeconds(2),
+                    () -> Assertions.assertEquals(source, xmlCleaner.removeDoctypeDeclarations(source)));
+        }
     }
 }