(url) Fix urlencoding issues with certain symbols

Problems primarily cropped up with sideloaded wikipedia articles, though the search engine has been returning inconsistently URLEncoded search results for a while, though browsers and servers have seemingly magically fixed the issues in many scenarios.

This addresses Issue #195 and Issue #131.

build.gradle

@@ -5,7 +5,7 @@ plugins {
 
     // This is a workaround for a bug in the Jib plugin that causes it to stall randomly
     // https://github.com/GoogleContainerTools/jib/issues/3347
-    id 'com.google.cloud.tools.jib' version '3.4.4' apply(false)
+    id 'com.google.cloud.tools.jib' version '3.4.5' apply(false)
 }
 
 group 'marginalia'
@@ -47,7 +47,7 @@ ext {
     dockerImageBase='container-registry.oracle.com/graalvm/jdk:24'
     dockerImageTag='latest'
     dockerImageRegistry='marginalia'
-    jibVersion = '3.4.4'
+    jibVersion = '3.4.5'
 }
 
 idea {

code/common/model/java/nu/marginalia/model/EdgeUrl.java

@@ -1,16 +1,14 @@
 package nu.marginalia.model;
 
 import nu.marginalia.util.QueryParams;
+import org.apache.commons.lang3.StringUtils;
 
 import javax.annotation.Nullable;
 import java.io.Serializable;
-import java.net.MalformedURLException;
+import java.net.*;
-import java.net.URI;
+import java.nio.charset.StandardCharsets;
-import java.net.URISyntaxException;
-import java.net.URL;
 import java.util.Objects;
 import java.util.Optional;
-import java.util.regex.Pattern;
 
 public class EdgeUrl implements Serializable {
     public final String proto;
@@ -33,7 +31,7 @@ public class EdgeUrl implements Serializable {
 
     private static URI parseURI(String url) throws URISyntaxException {
         try {
-            return new URI(urlencodeFixer(url));
+            return EdgeUriFactory.uriFromString(url);
         } catch (URISyntaxException ex) {
             throw new URISyntaxException("Failed to parse URI '" + url + "'", ex.getMessage());
         }
@@ -51,58 +49,6 @@ public class EdgeUrl implements Serializable {
         }
     }
 
-    private static Pattern badCharPattern = Pattern.compile("[ \t\n\"<>\\[\\]()',|]");
-
-    /* Java's URI parser is a bit too strict in throwing exceptions when there's an error.
-
-       Here on the Internet, standards are like the picture on the box of the frozen pizza,
-       and what you get is more like what's on the inside, we try to patch things instead,
-       just give it a best-effort attempt att cleaning out broken or unnecessary constructions
-       like bad or missing URLEncoding
-     */
-    public static String urlencodeFixer(String url) throws URISyntaxException {
-        var s = new StringBuilder();
-        String goodChars = "&.?:/-;+$#";
-        String hexChars = "0123456789abcdefABCDEF";
-
-        int pathIdx = findPathIdx(url);
-        if (pathIdx < 0) { // url looks like http://marginalia.nu
-            return url + "/";
-        }
-        s.append(url, 0, pathIdx);
-
-        // We don't want the fragment, and multiple fragments breaks the Java URIParser for some reason
-        int end = url.indexOf("#");
-        if (end < 0) end = url.length();
-
-        for (int i = pathIdx; i < end; i++) {
-            int c = url.charAt(i);
-
-            if (goodChars.indexOf(c) >= 0 || (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || (c >= '0' && c <= '9')) {
-                s.appendCodePoint(c);
-            } else if (c == '%' && i + 2 < end) {
-                int cn = url.charAt(i + 1);
-                int cnn = url.charAt(i + 2);
-                if (hexChars.indexOf(cn) >= 0 && hexChars.indexOf(cnn) >= 0) {
-                    s.appendCodePoint(c);
-                } else {
-                    s.append("%25");
-                }
-            } else {
-                s.append(String.format("%%%02X", c));
-            }
-        }
-
-        return s.toString();
-    }
-
-    private static int findPathIdx(String url) throws URISyntaxException {
-        int colonIdx = url.indexOf(':');
-        if (colonIdx < 0 || colonIdx + 2 >= url.length()) {
-            throw new URISyntaxException(url, "Lacking protocol");
-        }
-        return url.indexOf('/', colonIdx + 2);
-    }
 
     public EdgeUrl(URI URI) {
         try {
@@ -247,3 +193,123 @@ public class EdgeUrl implements Serializable {
     }
 
 }
+
+/* Java's URI parser is a bit too strict in throwing exceptions when there's an error.
+
+   Here on the Internet, standards are like the picture on the box of the frozen pizza,
+   and what you get is more like what's on the inside, we try to patch things instead,
+   just give it a best-effort attempt att cleaning out broken or unnecessary constructions
+   like bad or missing URLEncoding
+ */
+class EdgeUriFactory {
+    public static URI uriFromString(String url) throws URISyntaxException {
+        var s = new StringBuilder();
+
+        int pathIdx = findPathIdx(url);
+        if (pathIdx < 0) { // url looks like http://marginalia.nu
+            return new URI(url + "/");
+        }
+        s.append(url, 0, pathIdx);
+
+        // We don't want the fragment, and multiple fragments breaks the Java URIParser for some reason
+        int end = url.indexOf("#");
+        if (end < 0) end = url.length();
+
+        int queryIdx = url.indexOf('?');
+        if (queryIdx < 0) queryIdx = end;
+
+        recombinePaths(s, url.substring(pathIdx, queryIdx));
+        if (queryIdx < end) {
+            recombineQueryString(s, url.substring(queryIdx + 1, end));
+        }
+        return new URI(s.toString());
+    }
+
+    private static void recombinePaths(StringBuilder sb, String path) {
+        if (path == null || path.isEmpty()) {
+            return;
+        }
+
+        String[] pathParts = StringUtils.split(path, '/');
+        if (pathParts.length == 0) {
+            sb.append('/');
+            return;
+        }
+
+        for (String pathPart : pathParts) {
+            if (pathPart.isEmpty()) continue;
+
+            if (needsUrlEncode(pathPart)) {
+                sb.append('/');
+                sb.append(URLEncoder.encode(pathPart, StandardCharsets.UTF_8));
+            } else {
+                sb.append('/');
+                sb.append(pathPart);
+            }
+        }
+
+    }
+
+    private static void recombineQueryString(StringBuilder sb, String param) {
+        if (param == null || param.isEmpty()) {
+            return;
+        }
+
+        sb.append('?');
+        String[] pathParts = StringUtils.split(param, '&');
+        boolean first = true;
+        for (String pathPart : pathParts) {
+            if (pathPart.isEmpty()) continue;
+
+            if (first) {
+                first = false;
+            } else {
+                sb.append('&');
+            }
+            if (needsUrlEncode(pathPart)) {
+                sb.append(URLEncoder.encode(pathPart, StandardCharsets.UTF_8));
+            } else {
+                sb.append(pathPart);
+            }
+        }
+    }
+
+
+    /** Test if the url element needs URL encoding.
+     * <p></p>
+     * Note we may have been given an already encoded path element,
+     * so we include % and + in the list of good characters
+     */
+    private static boolean needsUrlEncode(String urlElement) {
+        for (int i = 0; i < urlElement.length(); i++) {
+            char c = urlElement.charAt(i);
+
+            if (c >= 'a' && c <= 'z') continue;
+            if (c >= 'A' && c <= 'Z') continue;
+            if (c >= '0' && c <= '9') continue;
+            if ("-_.~+?=&".indexOf(c) >= 0) continue;
+            if (c == '%' && i + 2 < urlElement.length()) {
+                char c1 = urlElement.charAt(i + 1);
+                char c2 = urlElement.charAt(i + 2);
+                if (c1 >= '0' && c1 <= '9' && c2 >= '0' && c2 <= '9') {
+                    i += 2;
+                    continue;
+                }
+            }
+
+            return true;
+        }
+
+        return false;
+    }
+
+    private static int findPathIdx(String url) throws URISyntaxException {
+        int colonIdx = url.indexOf(':');
+        if (colonIdx < 0 || colonIdx + 3 >= url.length()) {
+            throw new URISyntaxException(url, "Lacking protocol");
+        }
+        return url.indexOf('/', colonIdx + 3);
+    }
+
+
+}

code/common/model/test/nu/marginalia/model/EdgeUrlTest.java

@@ -1,6 +1,6 @@
 package nu.marginalia.model;
 
-import nu.marginalia.model.EdgeUrl;
+import org.junit.jupiter.api.Assertions;
 import org.junit.jupiter.api.Test;
 
 import java.net.URISyntaxException;
@@ -21,25 +21,25 @@ class EdgeUrlTest {
                 new EdgeUrl("https://memex.marginalia.nu/#here")
         );
     }
+
     @Test
-    public void testParam() throws URISyntaxException {
+    void testUriFromString() throws URISyntaxException {
-        System.out.println(new EdgeUrl("https://memex.marginalia.nu/index.php?id=1").toString());
+        Assertions.assertEquals("https://www.example.com/", EdgeUriFactory.uriFromString("https://www.example.com/#heredoc").toString());
-        System.out.println(new EdgeUrl("https://memex.marginalia.nu/showthread.php?id=1&count=5&tracking=123").toString());
+        Assertions.assertEquals("https://www.example.com/%25-sign", EdgeUriFactory.uriFromString("https://www.example.com/%-sign").toString());
-    }
+        Assertions.assertEquals("https://www.example.com/%22-sign", EdgeUriFactory.uriFromString("https://www.example.com/%22-sign").toString());
-    @Test
+        Assertions.assertEquals("https://www.example.com/%0A+%22huh%22", EdgeUriFactory.uriFromString("https://www.example.com/\n \"huh\"").toString());
-    void urlencodeFixer() throws URISyntaxException {
+        Assertions.assertEquals("https://en.wikipedia.org/wiki/S%C3%A1mi", EdgeUriFactory.uriFromString("https://en.wikipedia.org/wiki/Sámi").toString());
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/#heredoc"));
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/%-sign"));
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/%22-sign"));
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/\n \"huh\""));
     }
 
     @Test
     void testParms() throws URISyntaxException {
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?id=123"));
+        Assertions.assertEquals("id=123", new EdgeUrl("https://search.marginalia.nu/?id=123").param);
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?t=123"));
+        Assertions.assertEquals("t=123", new EdgeUrl("https://search.marginalia.nu/?t=123").param);
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?v=123"));
+        Assertions.assertEquals("v=123", new EdgeUrl("https://search.marginalia.nu/?v=123").param);
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?m=123"));
+        Assertions.assertEquals("id=1", new EdgeUrl("https://memex.marginalia.nu/showthread.php?id=1&count=5&tracking=123").param);
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?follow=123"));
+        Assertions.assertEquals("id=1&t=5", new EdgeUrl("https://memex.marginalia.nu/shöwthrëad.php?id=1&t=5&tracking=123").param);
+        Assertions.assertEquals("id=1&t=5", new EdgeUrl("https://memex.marginalia.nu/shöwthrëad.php?trëaking=123&id=1&t=5&").param);
+        Assertions.assertNull(new EdgeUrl("https://search.marginalia.nu/?m=123").param);
+        Assertions.assertNull(new EdgeUrl("https://search.marginalia.nu/?follow=123").param);
     }
 }

code/common/service/java/nu/marginalia/process/control/ProcessAdHocTaskHeartbeatImpl.java

@@ -59,16 +59,13 @@ public class ProcessAdHocTaskHeartbeatImpl implements AutoCloseable, ProcessAdHo
      */
     @Override
     public void progress(String step, int stepProgress, int stepCount) {
+        int lastProgress = this.progress;
         this.step = step;
-
-
-        // off by one since we calculate the progress based on the number of steps,
-        // and Enum.ordinal() is zero-based (so the 5th step in a 5 step task is 4, not 5; resulting in the
-        // final progress being 80% and not 100%)
-
         this.progress = (int) Math.round(100. * stepProgress / (double) stepCount);
 
-        logger.info("ProcessTask {} progress: {}%", taskBase, progress);
+        if (this.progress / 10 != lastProgress / 10) {
+            logger.info("ProcessTask {} progress: {}%", taskBase, progress);
+        }
     }
 
     /** Wrap a collection to provide heartbeat progress updates as it's iterated through */

code/common/service/java/nu/marginalia/service/control/ServiceAdHocTaskHeartbeatImpl.java

@@ -57,16 +57,13 @@ public class ServiceAdHocTaskHeartbeatImpl implements AutoCloseable, ServiceAdHo
      */
     @Override
     public void progress(String step, int stepProgress, int stepCount) {
+        int lastProgress = this.progress;
         this.step = step;
-
-
-        // off by one since we calculate the progress based on the number of steps,
-        // and Enum.ordinal() is zero-based (so the 5th step in a 5 step task is 4, not 5; resulting in the
-        // final progress being 80% and not 100%)
-
         this.progress = (int) Math.round(100. * stepProgress / (double) stepCount);
 
-        logger.info("ServiceTask {} progress: {}%", taskBase, progress);
+        if (this.progress / 10 != lastProgress / 10) {
+            logger.info("ProcessTask {} progress: {}%", taskBase, progress);
+        }
     }
 
     public void shutDown() {

code/execution/api/java/nu/marginalia/executor/client/ExecutorExportClient.java

@@ -48,12 +48,13 @@ public class ExecutorExportClient {
         return msgId;
     }
 
-    public void exportSampleData(int node, FileStorageId fid, int size, String name) {
+    public void exportSampleData(int node, FileStorageId fid, int size, String ctFilter, String name) {
         channelPool.call(ExecutorExportApiBlockingStub::exportSampleData)
                 .forNode(node)
                 .run(RpcExportSampleData.newBuilder()
                         .setFileStorageId(fid.id())
                         .setSize(size)
+                        .setCtFilter(ctFilter)
                         .setName(name)
                         .build());
     }

code/execution/api/src/main/protobuf/executor-api.proto

@@ -100,6 +100,7 @@ message RpcExportSampleData {
   int64 fileStorageId = 1;
   int32 size = 2;
   string name = 3;
+  string ctFilter = 4;
 }
 message RpcDownloadSampleData {
   string sampleSet = 1;

code/execution/java/nu/marginalia/actor/task/DownloadSampleActor.java

@@ -8,6 +8,7 @@ import nu.marginalia.actor.state.ActorResumeBehavior;
 import nu.marginalia.actor.state.ActorStep;
 import nu.marginalia.actor.state.Resume;
 import nu.marginalia.service.control.ServiceEventLog;
+import nu.marginalia.service.control.ServiceHeartbeat;
 import nu.marginalia.storage.FileStorageService;
 import nu.marginalia.storage.model.FileStorage;
 import nu.marginalia.storage.model.FileStorageId;
@@ -19,6 +20,7 @@ import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 import java.io.*;
+import java.net.HttpURLConnection;
 import java.net.MalformedURLException;
 import java.net.URI;
 import java.net.URL;
@@ -32,6 +34,7 @@ public class DownloadSampleActor extends RecordActorPrototype {
 
     private final FileStorageService storageService;
     private final ServiceEventLog eventLog;
+    private final ServiceHeartbeat heartbeat;
     private final Logger logger = LoggerFactory.getLogger(getClass());
 
     @Resume(behavior = ActorResumeBehavior.ERROR)
@@ -66,15 +69,39 @@ public class DownloadSampleActor extends RecordActorPrototype {
 
                 Files.deleteIfExists(Path.of(tarFileName));
 
-                try (var is = new BufferedInputStream(new URI(downloadURI).toURL().openStream());
+                HttpURLConnection urlConnection = (HttpURLConnection) new URI(downloadURI).toURL().openConnection();
-                     var os = new BufferedOutputStream(Files.newOutputStream(Path.of(tarFileName), StandardOpenOption.CREATE))) {
+
-                    is.transferTo(os);
+                try (var hb = heartbeat.createServiceAdHocTaskHeartbeat("Downloading sample")) {
+                    long size = urlConnection.getContentLengthLong();
+                    byte[] buffer = new byte[8192];
+
+                    try (var is = new BufferedInputStream(urlConnection.getInputStream());
+                         var os = new BufferedOutputStream(Files.newOutputStream(Path.of(tarFileName), StandardOpenOption.CREATE))) {
+                        long copiedSize = 0;
+
+                        while (copiedSize < size) {
+                            int read = is.read(buffer);
+
+                            if (read < 0) // We've been promised a file of length 'size'
+                                throw new IOException("Unexpected end of stream");
+
+                            os.write(buffer, 0, read);
+                            copiedSize += read;
+
+                            // Update progress bar
+                            hb.progress(String.format("%d MB", copiedSize / 1024 / 1024), (int) (copiedSize / 1024), (int) (size / 1024));
+                        }
+                    }
+
                 }
                 catch (Exception ex) {
                     eventLog.logEvent(DownloadSampleActor.class, "Error downloading sample");
                     logger.error("Error downloading sample", ex);
                     yield new Error();
                 }
+                finally {
+                    urlConnection.disconnect();
+                }
 
                 eventLog.logEvent(DownloadSampleActor.class, "Download complete");
                 yield new Extract(fileStorageId, tarFileName);
@@ -170,11 +197,12 @@ public class DownloadSampleActor extends RecordActorPrototype {
     @Inject
     public DownloadSampleActor(Gson gson,
                                FileStorageService storageService,
-                               ServiceEventLog eventLog)
+                               ServiceEventLog eventLog, ServiceHeartbeat heartbeat)
     {
         super(gson);
         this.storageService = storageService;
         this.eventLog = eventLog;
+        this.heartbeat = heartbeat;
     }
 
 }

code/execution/java/nu/marginalia/actor/task/ExportSampleDataActor.java

@@ -26,32 +26,32 @@ public class ExportSampleDataActor extends RecordActorPrototype {
     private final MqOutbox exportTasksOutbox;
     private final Logger logger = LoggerFactory.getLogger(getClass());
 
-    public record Export(FileStorageId crawlId, int size, String name) implements ActorStep {}
+    public record Export(FileStorageId crawlId, int size, String ctFilter, String name) implements ActorStep {}
-    public record Run(FileStorageId crawlId, FileStorageId destId, int size, String name, long msgId) implements ActorStep {
+    public record Run(FileStorageId crawlId, FileStorageId destId, int size, String ctFilter, String name, long msgId) implements ActorStep {
-        public Run(FileStorageId crawlId, FileStorageId destId, int size, String name) {
+        public Run(FileStorageId crawlId, FileStorageId destId, int size, String name, String ctFilter) {
-            this(crawlId, destId, size, name, -1);
+            this(crawlId, destId, size, name, ctFilter,-1);
         }
     }
 
     @Override
     public ActorStep transition(ActorStep self) throws Exception {
         return switch(self) {
-            case Export(FileStorageId crawlId, int size, String name) -> {
+            case Export(FileStorageId crawlId, int size, String ctFilter, String name) -> {
                 var storage = storageService.allocateStorage(FileStorageType.EXPORT,
                         "crawl-sample-export",
                         "Crawl Data Sample " + name + "/" + size + " " + LocalDateTime.now()
                 );
 
                 if (storage == null) yield new Error("Bad storage id");
-                yield new Run(crawlId, storage.id(), size, name);
+                yield new Run(crawlId, storage.id(), size, ctFilter, name);
             }
-            case Run(FileStorageId crawlId, FileStorageId destId, int size, String name, long msgId) when msgId < 0 -> {
+            case Run(FileStorageId crawlId, FileStorageId destId, int size, String ctFilter, String name, long msgId) when msgId < 0 -> {
                 storageService.setFileStorageState(destId, FileStorageState.NEW);
 
-                long newMsgId = exportTasksOutbox.sendAsync(ExportTaskRequest.sampleData(crawlId, destId, size, name));
+                long newMsgId = exportTasksOutbox.sendAsync(ExportTaskRequest.sampleData(crawlId, destId, ctFilter, size, name));
-                yield new Run(crawlId, destId, size, name, newMsgId);
+                yield new Run(crawlId, destId, size, ctFilter, name, newMsgId);
             }
-            case Run(_, FileStorageId destId, _, _, long msgId) -> {
+            case Run(_, FileStorageId destId, _, _, _, long msgId) -> {
                 var rsp = processWatcher.waitResponse(exportTasksOutbox, ProcessService.ProcessId.EXPORT_TASKS, msgId);
 
                 if (rsp.state() != MqMessageState.OK) {
@@ -70,7 +70,7 @@ public class ExportSampleDataActor extends RecordActorPrototype {
 
     @Override
     public String describe() {
-        return "Export RSS/Atom feeds from crawl data";
+        return "Export sample crawl data";
     }
 
     @Inject

code/execution/java/nu/marginalia/execution/ExecutorExportGrpcService.java

@@ -49,6 +49,7 @@ public class ExecutorExportGrpcService
                     new ExportSampleDataActor.Export(
                             FileStorageId.of(request.getFileStorageId()),
                             request.getSize(),
+                            request.getCtFilter(),
                             request.getName()
                     )
             );

code/functions/live-capture/java/nu/marginalia/rss/svc/FeedFetcherService.java

@@ -229,13 +229,15 @@ public class FeedFetcherService {
                     .timeout(Duration.ofSeconds(15))
                     ;
 
-            if (ifModifiedSinceDate != null) {
+            // Set the If-Modified-Since or If-None-Match headers if we have them
+            // though since there are certain idiosyncrasies in server implementations,
+            // we avoid setting both at the same time as that may turn a 304 into a 200.
+            if (ifNoneMatchTag != null) {
+                requestBuilder.header("If-None-Match", ifNoneMatchTag);
+            } else if (ifModifiedSinceDate != null) {
                 requestBuilder.header("If-Modified-Since", ifModifiedSinceDate);
             }
 
-            if (ifNoneMatchTag != null) {
-                requestBuilder.header("If-None-Match", ifNoneMatchTag);
-            }
 
             HttpRequest getRequest = requestBuilder.build();
 

code/processes/converting-process/test/nu/marginalia/converting/CrawlingThenConvertingIntegrationTest.java

@@ -20,7 +20,6 @@ import nu.marginalia.model.crawldata.CrawledDocument;
 import nu.marginalia.model.crawldata.CrawledDomain;
 import nu.marginalia.model.crawldata.SerializableCrawlData;
 import nu.marginalia.parquet.crawldata.CrawledDocumentParquetRecordFileWriter;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
 import org.junit.jupiter.api.*;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -247,7 +246,7 @@ public class CrawlingThenConvertingIntegrationTest {
     private CrawledDomain crawl(CrawlerMain.CrawlSpecRecord specs, Predicate<EdgeDomain> domainBlacklist) throws Exception {
         List<SerializableCrawlData> data = new ArrayList<>();
 
-        try (var recorder = new WarcRecorder(fileName, new BasicCookieStore());
+        try (var recorder = new WarcRecorder(fileName);
              var db = new DomainStateDb(dbTempFile))
         {
             new CrawlerRetreiver(httpFetcher, new DomainProber(domainBlacklist), specs, db, recorder).crawlDomain();

code/processes/crawling-process/java/nu/marginalia/crawl/CrawlerMain.java

@@ -43,6 +43,7 @@ import java.nio.file.StandardCopyOption;
 import java.security.Security;
 import java.util.*;
 import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.LinkedBlockingQueue;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicInteger;
 
@@ -66,6 +67,8 @@ public class CrawlerMain extends ProcessMainClass {
 
     private final Map<String, CrawlTask> pendingCrawlTasks = new ConcurrentHashMap<>();
 
+    private final LinkedBlockingQueue<CrawlTask> retryQueue = new LinkedBlockingQueue<>();
+
     private final AtomicInteger tasksDone = new AtomicInteger(0);
     private final HttpFetcherImpl fetcher;
 
@@ -261,28 +264,44 @@ public class CrawlerMain extends ProcessMainClass {
                 if (workLog.isJobFinished(crawlSpec.domain))
                     continue;
 
-                var task = new CrawlTask(
+                var task = new CrawlTask(crawlSpec, anchorTagsSource, outputDir, warcArchiver, domainStateDb, workLog);
-                        crawlSpec,
-                        anchorTagsSource,
-                        outputDir,
-                        warcArchiver,
-                        domainStateDb,
-                        workLog);
 
                 // Try to run immediately, to avoid unnecessarily keeping the entire work set in RAM
                 if (!trySubmitDeferredTask(task)) {
-                    // Otherwise add to the taskList for deferred execution
+
+                    // Drain the retry queue to the taskList, and try to submit any tasks that are in the retry queue
+                    retryQueue.drainTo(taskList);
+                    taskList.removeIf(this::trySubmitDeferredTask);
+
+                    // Then add this new task to the retry queue
                     taskList.add(task);
                 }
             }
 
              // Schedule viable tasks for execution until list is empty
-            while (!taskList.isEmpty()) {
+            for (int emptyRuns = 0;emptyRuns < 300;) {
-                taskList.removeIf(this::trySubmitDeferredTask);
+                boolean hasTasks = !taskList.isEmpty();
 
-                // Add a small pause here to avoid busy looping toward the end of the execution cycle when
+                // The order of these checks  very important to avoid a race condition
-                // we might have no new viable tasks to run for hours on end
+                // where we miss a task that is put into the retry queue
-                TimeUnit.MILLISECONDS.sleep(50);
+                boolean hasRunningTasks = pool.getActiveCount() > 0;
+                boolean hasRetryTasks = !retryQueue.isEmpty();
+
+                if (hasTasks || hasRetryTasks || hasRunningTasks) {
+                    retryQueue.drainTo(taskList);
+
+                    // Try to submit any tasks that are in the retry queue (this will block if the pool is full)
+                    taskList.removeIf(this::trySubmitDeferredTask);
+
+                    // Add a small pause here to avoid busy looping toward the end of the execution cycle when
+                    // we might have no new viable tasks to run for hours on end
+                    TimeUnit.MILLISECONDS.sleep(5);
+                } else {
+                    // We have no tasks to run, and no tasks in the retry queue
+                    // but we wait a bit to see if any new tasks come in via the retry queue
+                    emptyRuns++;
+                    TimeUnit.SECONDS.sleep(1);
+                }
             }
 
             logger.info("Shutting down the pool, waiting for tasks to complete...");
@@ -414,7 +433,7 @@ public class CrawlerMain extends ProcessMainClass {
         /** Best effort indicator whether we could start this now without getting stuck in
          * DomainLocks purgatory */
         public boolean canRun() {
-            return domainLocks.canLock(new EdgeDomain(domain));
+            return domainLocks.isLockableHint(new EdgeDomain(domain));
         }
 
         @Override
@@ -425,66 +444,82 @@ public class CrawlerMain extends ProcessMainClass {
                 return;
             }
 
-            Path newWarcFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.LIVE);
+            Optional<DomainLocks.DomainLock> lock = domainLocks.tryLockDomain(new EdgeDomain(domain));
-            Path tempFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.TEMP);
+            // We don't have a lock, so we can't run this task
-            Path slopFile = CrawlerOutputFile.createSlopPath(outputDir, id, domain);
+            // we return to avoid blocking the pool for too long
-
+            if (lock.isEmpty()) {
-            // Move the WARC file to a temp file if it exists, so we can resume the crawl using the old data
+                if (retryQueue.remainingCapacity() > 0) {
-            // while writing to the same file name as before
+                    // Sleep a moment to avoid busy looping via the retry queue
-            if (Files.exists(newWarcFile)) {
+                    // in the case when few tasks remain and almost all are ineligible for
-                Files.move(newWarcFile, tempFile, StandardCopyOption.REPLACE_EXISTING);
+                    // immediate restart
-            }
+                    Thread.sleep(5);
-            else {
-                Files.deleteIfExists(tempFile);
-            }
-
-            try (var warcRecorder = new WarcRecorder(newWarcFile, fetcher); // write to a temp file for now
-                 var retriever = new CrawlerRetreiver(fetcher, domainProber, specification, domainStateDb, warcRecorder);
-                 CrawlDataReference reference = getReference()
-            )
-            {
-                // Resume the crawl if it was aborted
-                if (Files.exists(tempFile)) {
-                    retriever.syncAbortedRun(tempFile);
-                    Files.delete(tempFile);
                 }
 
-                DomainLinks domainLinks = anchorTagsSource.getAnchorTags(domain);
+                retryQueue.put(this);
+                return;
+            }
+            DomainLocks.DomainLock domainLock = lock.get();
 
-                int size;
+            try (domainLock) {
-                try (var lock = domainLocks.lockDomain(new EdgeDomain(domain))) {
+                Thread.currentThread().setName("crawling:" + domain);
-                    size = retriever.crawlDomain(domainLinks, reference);
+
+                Path newWarcFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.LIVE);
+                Path tempFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.TEMP);
+                Path slopFile = CrawlerOutputFile.createSlopPath(outputDir, id, domain);
+
+                // Move the WARC file to a temp file if it exists, so we can resume the crawl using the old data
+                // while writing to the same file name as before
+                if (Files.exists(newWarcFile)) {
+                    Files.move(newWarcFile, tempFile, StandardCopyOption.REPLACE_EXISTING);
+                }
+                else {
+                    Files.deleteIfExists(tempFile);
                 }
 
-                // Delete the reference crawl data if it's not the same as the new one
+                try (var warcRecorder = new WarcRecorder(newWarcFile); // write to a temp file for now
-                // (mostly a case when migrating from legacy->warc)
+                     var retriever = new CrawlerRetreiver(fetcher, domainProber, specification, domainStateDb, warcRecorder);
-                reference.delete();
+                     CrawlDataReference reference = getReference())
+                {
+                    // Resume the crawl if it was aborted
+                    if (Files.exists(tempFile)) {
+                        retriever.syncAbortedRun(tempFile);
+                        Files.delete(tempFile);
+                    }
 
-                // Convert the WARC file to Parquet
+                    DomainLinks domainLinks = anchorTagsSource.getAnchorTags(domain);
-                SlopCrawlDataRecord
-                        .convertWarc(domain, userAgent, newWarcFile, slopFile);
 
-                // Optionally archive the WARC file if full retention is enabled,
+                    int size = retriever.crawlDomain(domainLinks, reference);
-                // otherwise delete it:
-                warcArchiver.consumeWarc(newWarcFile, domain);
 
-                // Mark the domain as finished in the work log
+                    // Delete the reference crawl data if it's not the same as the new one
-                workLog.setJobToFinished(domain, slopFile.toString(), size);
+                    // (mostly a case when migrating from legacy->warc)
+                    reference.delete();
 
-                // Update the progress bar
+                    // Convert the WARC file to Slop
-                heartbeat.setProgress(tasksDone.incrementAndGet() / (double) totalTasks);
+                    SlopCrawlDataRecord
+                            .convertWarc(domain, userAgent, newWarcFile, slopFile);
 
-                logger.info("Fetched {}", domain);
+                    // Optionally archive the WARC file if full retention is enabled,
-            } catch (Exception e) {
+                    // otherwise delete it:
-                logger.error("Error fetching domain " + domain, e);
+                    warcArchiver.consumeWarc(newWarcFile, domain);
-            }
-            finally {
-                // We don't need to double-count these; it's also kept in the workLog
-                pendingCrawlTasks.remove(domain);
-                Thread.currentThread().setName("[idle]");
 
-                Files.deleteIfExists(newWarcFile);
+                    // Mark the domain as finished in the work log
-                Files.deleteIfExists(tempFile);
+                    workLog.setJobToFinished(domain, slopFile.toString(), size);
+
+                    // Update the progress bar
+                    heartbeat.setProgress(tasksDone.incrementAndGet() / (double) totalTasks);
+
+                    logger.info("Fetched {}", domain);
+                } catch (Exception e) {
+                    logger.error("Error fetching domain " + domain, e);
+                }
+                finally {
+                    // We don't need to double-count these; it's also kept in the workLog
+                    pendingCrawlTasks.remove(domain);
+                    Thread.currentThread().setName("[idle]");
+
+                    Files.deleteIfExists(newWarcFile);
+                    Files.deleteIfExists(tempFile);
+                }
             }
         }
 

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/ContentTags.java

@@ -1,6 +1,6 @@
 package nu.marginalia.crawl.fetcher;
 
-import org.apache.hc.core5.http.io.support.ClassicRequestBuilder;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
 
 /** Encapsulates request modifiers; the ETag and Last-Modified tags for a resource */
 public record ContentTags(String etag, String lastMod) {
@@ -17,14 +17,16 @@ public record ContentTags(String etag, String lastMod) {
     }
 
     /** Paints the tags onto the request builder. */
-    public void paint(ClassicRequestBuilder getBuilder) {
+    public void paint(HttpGet request) {
+
+        // Paint the ETag header if present,
+        // otherwise paint the Last-Modified header
+        // (but not both at the same time due to some servers not liking it)
 
         if (etag != null) {
-            getBuilder.addHeader("If-None-Match", etag);
+            request.addHeader("If-None-Match", etag);
-        }
+        } else if (lastMod != null) {
-
+            request.addHeader("If-Modified-Since", lastMod);
-        if (lastMod != null) {
-            getBuilder.addHeader("If-Modified-Since", lastMod);
         }
     }
 }

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/Cookies.java

@@ -1,34 +0,0 @@
-package nu.marginalia.crawl.fetcher;
-
-import java.io.IOException;
-import java.net.CookieHandler;
-import java.net.URI;
-import java.util.List;
-import java.util.Map;
-import java.util.concurrent.ConcurrentHashMap;
-
-public class Cookies extends CookieHandler {
-    final ThreadLocal<ConcurrentHashMap<String, List<String>>> cookieJar = ThreadLocal.withInitial(ConcurrentHashMap::new);
-
-    public void clear() {
-        cookieJar.get().clear();
-    }
-
-    public boolean hasCookies() {
-        return !cookieJar.get().isEmpty();
-    }
-
-    public List<String> getCookies() {
-        return cookieJar.get().values().stream().flatMap(List::stream).toList();
-    }
-
-    @Override
-    public Map<String, List<String>> get(URI uri, Map<String, List<String>> requestHeaders) throws IOException {
-        return cookieJar.get();
-    }
-
-    @Override
-    public void put(URI uri, Map<String, List<String>> responseHeaders) throws IOException {
-        cookieJar.get().putAll(responseHeaders);
-    }
-}

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/DomainCookies.java

@@ -0,0 +1,56 @@
+package nu.marginalia.crawl.fetcher;
+
+import org.apache.hc.client5.http.classic.methods.HttpUriRequestBase;
+import org.apache.hc.core5.http.ClassicHttpRequest;
+import org.apache.hc.core5.http.HttpResponse;
+
+import java.util.HashMap;
+import java.util.Map;
+import java.util.StringJoiner;
+
+public class DomainCookies {
+    private final Map<String, String> cookies = new HashMap<>();
+
+    public boolean hasCookies() {
+        return !cookies.isEmpty();
+    }
+
+    public void updateCookieStore(HttpResponse response) {
+        for (var header : response.getHeaders()) {
+            if (header.getName().equalsIgnoreCase("Set-Cookie")) {
+                parseCookieHeader(header.getValue());
+            }
+        }
+    }
+
+    private void parseCookieHeader(String value) {
+        // Parse the Set-Cookie header value and extract the cookies
+
+        String[] parts = value.split(";");
+        String cookie = parts[0].trim();
+
+        if (cookie.contains("=")) {
+            String[] cookieParts = cookie.split("=");
+            String name = cookieParts[0].trim();
+            String val = cookieParts[1].trim();
+            cookies.put(name, val);
+        }
+    }
+
+    public void paintRequest(HttpUriRequestBase request) {
+        request.addHeader("Cookie", createCookieHeader());
+    }
+
+    public void paintRequest(ClassicHttpRequest request) {
+        request.addHeader("Cookie", createCookieHeader());
+    }
+
+    private String createCookieHeader() {
+        StringJoiner sj = new StringJoiner("; ");
+        for (var cookie : cookies.entrySet()) {
+            sj.add(cookie.getKey() + "=" + cookie.getValue());
+        }
+        return sj.toString();
+    }
+
+}

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/HttpFetcher.java

@@ -23,6 +23,7 @@ public interface HttpFetcher extends AutoCloseable {
 
     HttpFetchResult fetchContent(EdgeUrl url,
                                  WarcRecorder recorder,
+                                 DomainCookies cookies,
                                  CrawlDelayTimer timer,
                                  ContentTags tags,
                                  ProbeType probeType);

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/HttpFetcherImpl.java

@@ -17,6 +17,7 @@ import nu.marginalia.model.crawldata.CrawlerDomainStatus;
 import org.apache.hc.client5.http.ConnectionKeepAliveStrategy;
 import org.apache.hc.client5.http.HttpRequestRetryStrategy;
 import org.apache.hc.client5.http.classic.HttpClient;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
 import org.apache.hc.client5.http.config.ConnectionConfig;
 import org.apache.hc.client5.http.config.RequestConfig;
 import org.apache.hc.client5.http.cookie.BasicCookieStore;
@@ -34,6 +35,7 @@ import org.apache.hc.core5.http.io.entity.EntityUtils;
 import org.apache.hc.core5.http.io.support.ClassicRequestBuilder;
 import org.apache.hc.core5.http.message.MessageSupport;
 import org.apache.hc.core5.http.protocol.HttpContext;
+import org.apache.hc.core5.pool.PoolStats;
 import org.apache.hc.core5.util.TimeValue;
 import org.apache.hc.core5.util.Timeout;
 import org.jsoup.Jsoup;
@@ -45,11 +47,14 @@ import org.slf4j.Marker;
 import org.slf4j.MarkerFactory;
 
 import javax.net.ssl.SSLContext;
+import javax.net.ssl.SSLException;
 import java.io.IOException;
 import java.net.SocketTimeoutException;
 import java.net.URISyntaxException;
+import java.net.UnknownHostException;
 import java.security.NoSuchAlgorithmException;
 import java.time.Duration;
+import java.time.Instant;
 import java.util.*;
 import java.util.concurrent.Semaphore;
 import java.util.concurrent.TimeUnit;
@@ -76,14 +81,20 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
     }
 
     private final CloseableHttpClient client;
+    private PoolingHttpClientConnectionManager connectionManager;
+
+    public PoolStats getPoolStats() {
+        return connectionManager.getTotalStats();
+    }
 
     private CloseableHttpClient createClient() throws NoSuchAlgorithmException {
         final ConnectionConfig connectionConfig = ConnectionConfig.custom()
                 .setSocketTimeout(10, TimeUnit.SECONDS)
                 .setConnectTimeout(30, TimeUnit.SECONDS)
+                .setValidateAfterInactivity(TimeValue.ofSeconds(5))
                 .build();
 
-        final PoolingHttpClientConnectionManager connectionManager = PoolingHttpClientConnectionManagerBuilder.create()
+        connectionManager = PoolingHttpClientConnectionManagerBuilder.create()
                 .setMaxConnPerRoute(2)
                 .setMaxConnTotal(5000)
                 .setDefaultConnectionConfig(connectionConfig)
@@ -91,15 +102,27 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
                 .build();
 
         connectionManager.setDefaultSocketConfig(SocketConfig.custom()
-                .setSoLinger(TimeValue.ofSeconds(15))
+                .setSoLinger(TimeValue.ofSeconds(-1))
                 .setSoTimeout(Timeout.ofSeconds(10))
                 .build()
         );
 
+        Thread.ofPlatform().daemon(true).start(() -> {
+            try {
+                for (;;) {
+                    TimeUnit.SECONDS.sleep(15);
+                    logger.info("Connection pool stats: {}", connectionManager.getTotalStats());
+                }
+            }
+            catch (InterruptedException e) {
+                Thread.currentThread().interrupt();
+            }
+        });
+
         final RequestConfig defaultRequestConfig = RequestConfig.custom()
                 .setCookieSpec(StandardCookieSpec.RELAXED)
                 .setResponseTimeout(10, TimeUnit.SECONDS)
-                .setConnectionRequestTimeout(8, TimeUnit.SECONDS)
+                .setConnectionRequestTimeout(5, TimeUnit.MINUTES)
                 .build();
 
         return HttpClients.custom()
@@ -287,6 +310,7 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
      * recorded in the WARC file on failure.
      */
     public ContentTypeProbeResult probeContentType(EdgeUrl url,
+                                                   DomainCookies cookies,
                                                    CrawlDelayTimer timer,
                                                    ContentTags tags) {
         if (!tags.isEmpty() || !contentTypeLogic.isUrlLikeBinary(url)) {
@@ -299,9 +323,11 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
                     .addHeader("Accept-Encoding", "gzip")
                     .build();
 
-            var result = SendLock.wrapSend(client, head, (rsp) -> {
+            cookies.paintRequest(head);
-                EntityUtils.consume(rsp.getEntity());
 
+            return SendLock.wrapSend(client, head, (rsp) -> {
+                cookies.updateCookieStore(rsp);
+                EntityUtils.consume(rsp.getEntity());
                 int statusCode = rsp.getCode();
 
                 // Handle redirects
@@ -339,8 +365,6 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
                     return new ContentTypeProbeResult.BadContentType(contentType, statusCode);
                 }
             });
-
-            return result;
         }
         catch (SocketTimeoutException ex) {
 
@@ -362,6 +386,7 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
     @Override
     public HttpFetchResult fetchContent(EdgeUrl url,
                                            WarcRecorder warcRecorder,
+                                           DomainCookies cookies,
                                            CrawlDelayTimer timer,
                                            ContentTags contentTags,
                                            ProbeType probeType)
@@ -369,26 +394,32 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
         try {
             if (probeType == HttpFetcher.ProbeType.FULL) {
                 try {
-                    var probeResult = probeContentType(url, timer, contentTags);
+                    var probeResult = probeContentType(url, cookies, timer, contentTags);
-                    logger.info(crawlerAuditMarker, "Probe result {} for {}", probeResult.getClass().getSimpleName(), url);
+
                     switch (probeResult) {
                         case HttpFetcher.ContentTypeProbeResult.NoOp():
                             break; //
                         case HttpFetcher.ContentTypeProbeResult.Ok(EdgeUrl resolvedUrl):
+                            logger.info(crawlerAuditMarker, "Probe result OK for {}", url);
                             url = resolvedUrl; // If we were redirected while probing, use the final URL for fetching
                             break;
                         case ContentTypeProbeResult.BadContentType badContentType:
                             warcRecorder.flagAsFailedContentTypeProbe(url, badContentType.contentType(), badContentType.statusCode());
+                            logger.info(crawlerAuditMarker, "Probe result Bad ContenType ({}) for {}", badContentType.contentType(), url);
                             return new HttpFetchResult.ResultNone();
                         case ContentTypeProbeResult.BadContentType.Timeout(Exception ex):
+                            logger.info(crawlerAuditMarker, "Probe result Timeout for {}", url);
                             warcRecorder.flagAsTimeout(url);
                             return new HttpFetchResult.ResultException(ex);
                         case ContentTypeProbeResult.Exception(Exception ex):
+                            logger.info(crawlerAuditMarker, "Probe result Exception({}) for {}", ex.getClass().getSimpleName(), url);
                             warcRecorder.flagAsError(url, ex);
                             return new HttpFetchResult.ResultException(ex);
                         case ContentTypeProbeResult.HttpError httpError:
+                            logger.info(crawlerAuditMarker, "Probe result HTTP Error ({}) for {}", httpError.statusCode(), url);
                             return new HttpFetchResult.ResultException(new HttpException("HTTP status code " + httpError.statusCode() + ": " + httpError.message()));
                         case ContentTypeProbeResult.Redirect redirect:
+                            logger.info(crawlerAuditMarker, "Probe result redirect for {} -> {}", url, redirect.location());
                             return new HttpFetchResult.ResultRedirect(redirect.location());
                     }
                 } catch (Exception ex) {
@@ -398,36 +429,41 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
 
             }
 
-            ClassicRequestBuilder getBuilder = ClassicRequestBuilder.get(url.asURI())
+            HttpGet request = new HttpGet(url.asURI());
-                    .addHeader("User-Agent", userAgentString)
+            request.addHeader("User-Agent", userAgentString);
-                    .addHeader("Accept-Encoding", "gzip")
+            request.addHeader("Accept-Encoding", "gzip");
-                    .addHeader("Accept-Language", "en,*;q=0.5")
+            request.addHeader("Accept-Language", "en,*;q=0.5");
-                    .addHeader("Accept", "text/html, application/xhtml+xml, text/*;q=0.8");
+            request.addHeader("Accept", "text/html, application/xhtml+xml, text/*;q=0.8");
 
-            contentTags.paint(getBuilder);
+            contentTags.paint(request);
 
             try (var sl = new SendLock()) {
-                HttpFetchResult result = warcRecorder.fetch(client, getBuilder.build());
+                Instant start = Instant.now();
+                HttpFetchResult result = warcRecorder.fetch(client, cookies, request);
+
+                Duration fetchDuration = Duration.between(start, Instant.now());
 
                 if (result instanceof HttpFetchResult.ResultOk ok) {
                     if (ok.statusCode() == 304) {
-                        return new HttpFetchResult.Result304Raw();
+                        result = new HttpFetchResult.Result304Raw();
                     }
                 }
 
                 switch (result) {
-                    case HttpFetchResult.ResultOk ok -> logger.info(crawlerAuditMarker, "Fetch result OK {} for {}", ok.statusCode(), url);
+                    case HttpFetchResult.ResultOk ok -> logger.info(crawlerAuditMarker, "Fetch result OK {} for {} ({} ms)", ok.statusCode(), url, fetchDuration.toMillis());
                     case HttpFetchResult.ResultRedirect redirect -> logger.info(crawlerAuditMarker, "Fetch result redirect: {}  for {}", redirect.url(), url);
-                    case HttpFetchResult.ResultNone none -> logger.info(crawlerAuditMarker, "Fetch result none  for {}", url);
+                    case HttpFetchResult.ResultNone none -> logger.info(crawlerAuditMarker, "Fetch result none for {}", url);
-                    case HttpFetchResult.ResultException ex -> logger.error(crawlerAuditMarker, "Fetch result exception: {}  for {}", ex.getClass().getSimpleName(), url);
+                    case HttpFetchResult.ResultException ex -> logger.error(crawlerAuditMarker, "Fetch result exception for {}", url, ex.ex());
                     case HttpFetchResult.Result304Raw raw -> logger.info(crawlerAuditMarker, "Fetch result: 304 Raw for {}", url);
                     case HttpFetchResult.Result304ReplacedWithReference ref -> logger.info(crawlerAuditMarker, "Fetch result: 304 With reference for {}", url);
                 }
+
                 return result;
             }
         }
         catch (Exception ex) {
-            ex.printStackTrace();
+            logger.error(crawlerAuditMarker, "Fetch result exception for {}", url, ex);
+
             return new HttpFetchResult.ResultException(ex);
         }
 
@@ -494,56 +530,61 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
     }
 
 
-    private SitemapResult fetchSingleSitemap(EdgeUrl sitemapUrl) throws URISyntaxException, IOException, InterruptedException {
+    private SitemapResult fetchSingleSitemap(EdgeUrl sitemapUrl) throws URISyntaxException {
-        ClassicHttpRequest getRequest = ClassicRequestBuilder.get(sitemapUrl.asURI())
+        HttpGet getRequest = new HttpGet(sitemapUrl.asURI());
-                .addHeader("User-Agent", userAgentString)
+
-                .addHeader("Accept-Encoding", "gzip")
+        getRequest.addHeader("User-Agent", userAgentString);
-                .addHeader("Accept", "text/*, */*;q=0.9")
+        getRequest.addHeader("Accept-Encoding", "gzip");
-                .addHeader("User-Agent", userAgentString)
+        getRequest.addHeader("Accept", "text/*, */*;q=0.9");
-                .build();
+        getRequest.addHeader("User-Agent", userAgentString);
 
         try (var sl = new SendLock()) {
             return client.execute(getRequest, response -> {
-                if (response.getCode() != 200) {
+                try {
-                    return new SitemapResult.SitemapError();
+                    if (response.getCode() != 200) {
+                        return new SitemapResult.SitemapError();
+                    }
+
+                    Document parsedSitemap = Jsoup.parse(
+                            EntityUtils.toString(response.getEntity()),
+                            sitemapUrl.toString(),
+                            Parser.xmlParser()
+                    );
+
+                    if (parsedSitemap.childrenSize() == 0) {
+                        return new SitemapResult.SitemapError();
+                    }
+
+                    String rootTagName = parsedSitemap.child(0).tagName();
+
+                    return switch (rootTagName.toLowerCase()) {
+                        case "sitemapindex" -> {
+                            List<String> references = new ArrayList<>();
+                            for (var locTag : parsedSitemap.getElementsByTag("loc")) {
+                                references.add(locTag.text().trim());
+                            }
+                            yield new SitemapResult.SitemapReferences(Collections.unmodifiableList(references));
+                        }
+                        case "urlset" -> {
+                            List<String> urls = new ArrayList<>();
+                            for (var locTag : parsedSitemap.select("url > loc")) {
+                                urls.add(locTag.text().trim());
+                            }
+                            yield new SitemapResult.SitemapUrls(Collections.unmodifiableList(urls));
+                        }
+                        case "rss", "atom" -> {
+                            List<String> urls = new ArrayList<>();
+                            for (var locTag : parsedSitemap.select("link, url")) {
+                                urls.add(locTag.text().trim());
+                            }
+                            yield new SitemapResult.SitemapUrls(Collections.unmodifiableList(urls));
+                        }
+                        default -> new SitemapResult.SitemapError();
+                    };
                 }
-
+                finally {
-                Document parsedSitemap = Jsoup.parse(
+                    EntityUtils.consume(response.getEntity());
-                        EntityUtils.toString(response.getEntity()),
-                        sitemapUrl.toString(),
-                        Parser.xmlParser()
-                );
-
-                if (parsedSitemap.childrenSize() == 0) {
-                    return new SitemapResult.SitemapError();
                 }
-
-                String rootTagName = parsedSitemap.child(0).tagName();
-
-                return switch (rootTagName.toLowerCase()) {
-                    case "sitemapindex" -> {
-                        List<String> references = new ArrayList<>();
-                        for (var locTag : parsedSitemap.getElementsByTag("loc")) {
-                            references.add(locTag.text().trim());
-                        }
-                        yield new SitemapResult.SitemapReferences(Collections.unmodifiableList(references));
-                    }
-                    case "urlset" -> {
-                        List<String> urls = new ArrayList<>();
-                        for (var locTag : parsedSitemap.select("url > loc")) {
-                            urls.add(locTag.text().trim());
-                        }
-                        yield new SitemapResult.SitemapUrls(Collections.unmodifiableList(urls));
-                    }
-                    case "rss", "atom" -> {
-                        List<String> urls = new ArrayList<>();
-                        for (var locTag : parsedSitemap.select("link, url")) {
-                            urls.add(locTag.text().trim());
-                        }
-                        yield new SitemapResult.SitemapUrls(Collections.unmodifiableList(urls));
-                    }
-                    default -> new SitemapResult.SitemapError();
-                };
             });
         }
         catch (Exception ex) {
@@ -574,13 +615,12 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
     private Optional<SimpleRobotRules> fetchAndParseRobotsTxt(EdgeUrl url, WarcRecorder recorder) {
         try (var sl = new SendLock()) {
 
-            ClassicHttpRequest request = ClassicRequestBuilder.get(url.asURI())
+            HttpGet request = new HttpGet(url.asURI());
-                    .addHeader("User-Agent", userAgentString)
+            request.addHeader("User-Agent", userAgentString);
-                    .addHeader("Accept-Encoding", "gzip")
+            request.addHeader("Accept-Encoding", "gzip");
-                    .addHeader("Accept", "text/*, */*;q=0.9")
+            request.addHeader("Accept", "text/*, */*;q=0.9");
-                    .build();
 
-            HttpFetchResult result = recorder.fetch(client, request);
+            HttpFetchResult result = recorder.fetch(client, new DomainCookies(), request);
 
             return DocumentBodyExtractor.asBytes(result).mapOpt((contentType, body) ->
                 robotsParser.parseContent(url.toString(),
@@ -596,18 +636,19 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
 
     @Override
     public boolean retryRequest(HttpRequest request, IOException exception, int executionCount, HttpContext context) {
-        if (exception instanceof SocketTimeoutException ex) {
+        return switch (exception) {
-            return false;
+            case SocketTimeoutException ste -> false;
-        }
+            case SSLException ssle -> false;
-
+            case UnknownHostException uhe -> false;
-        return executionCount < 3;
+            default -> executionCount <= 3;
+        };
     }
 
     @Override
     public boolean retryRequest(HttpResponse response, int executionCount, HttpContext context) {
         return switch (response.getCode()) {
-            case 500, 503 -> executionCount < 2;
+            case 500, 503 -> executionCount <= 2;
-            case 429 -> executionCount < 3;
+            case 429 -> executionCount <= 3;
             default -> false;
         };
     }

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/warc/WarcInputBuffer.java

@@ -2,6 +2,7 @@ package nu.marginalia.crawl.fetcher.warc;
 
 import org.apache.commons.io.IOUtils;
 import org.apache.commons.io.input.BOMInputStream;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
 import org.apache.hc.core5.http.ClassicHttpResponse;
 import org.apache.hc.core5.http.Header;
 import org.netpreserve.jwarc.WarcTruncationReason;
@@ -43,7 +44,9 @@ public abstract class WarcInputBuffer implements AutoCloseable {
      *  and suppressed from the headers.
      *  If an error occurs, a buffer will be created with no content and an error status.
      */
-    static WarcInputBuffer forResponse(ClassicHttpResponse response, Duration timeLimit) throws IOException {
+    static WarcInputBuffer forResponse(ClassicHttpResponse response,
+                                       HttpGet request,
+                                       Duration timeLimit) throws IOException {
         if (response == null)
             return new ErrorBuffer();
 
@@ -54,16 +57,47 @@ public abstract class WarcInputBuffer implements AutoCloseable {
             return new ErrorBuffer();
         }
 
-        InputStream is = entity.getContent();
+        Instant start = Instant.now();
-        long length = entity.getContentLength();
+        InputStream is = null;
+        try {
+            is = entity.getContent();
+            long length = entity.getContentLength();
 
-        try (response) {
             if (length > 0 && length < 8192) {
                 // If the content is small and not compressed, we can just read it into memory
-                return new MemoryBuffer(response.getHeaders(), timeLimit, is, (int) length);
+                return new MemoryBuffer(response.getHeaders(), request, timeLimit, is, (int) length);
             } else {
                 // Otherwise, we unpack it into a file and read it from there
-                return new FileBuffer(response.getHeaders(), timeLimit, is);
+                return new FileBuffer(response.getHeaders(), request, timeLimit, is);
+            }
+        }
+        finally {
+            // We're required to consume the stream to avoid leaking connections,
+            // but we also don't want to get stuck on slow or malicious connections
+            // forever, so we set a time limit on this phase and call abort() if it's exceeded.
+            try {
+                while (is != null) {
+                    // Consume some data
+                    if (is.skip(65536) == 0) {
+                        // Note that skip may return 0 if the stream is empty
+                        // or for other unspecified reasons, so we need to check
+                        // with read() as well to determine if the stream is done
+                        if (is.read() == -1)
+                            is = null;
+                    }
+                    // Check if the time limit has been exceeded
+                    else if (Duration.between(start, Instant.now()).compareTo(timeLimit) > 0) {
+                        request.abort();
+                        is = null;
+                    }
+                }
+            }
+            catch (IOException e) {
+                // Ignore the exception
+            }
+            finally {
+                // Close the input stream
+                IOUtils.closeQuietly(is);
             }
         }
 
@@ -71,7 +105,7 @@ public abstract class WarcInputBuffer implements AutoCloseable {
     }
 
     /** Copy an input stream to an output stream, with a maximum size and time limit */
-    protected void copy(InputStream is, OutputStream os, Duration timeLimit) {
+    protected void copy(InputStream is, HttpGet request, OutputStream os, Duration timeLimit) {
         Instant start = Instant.now();
         Instant timeout = start.plus(timeLimit);
         long size = 0;
@@ -86,6 +120,11 @@ public abstract class WarcInputBuffer implements AutoCloseable {
                 Duration remaining = Duration.between(Instant.now(), timeout);
                 if (remaining.isNegative()) {
                     truncationReason = WarcTruncationReason.TIME;
+                    // Abort the request if the time limit is exceeded
+                    // so we don't keep the connection open forever or are forced to consume
+                    // the stream to the end
+
+                    request.abort();
                     break;
                 }
 
@@ -104,6 +143,7 @@ public abstract class WarcInputBuffer implements AutoCloseable {
                 }
                 else if (truncationReason != WarcTruncationReason.LENGTH) {
                     truncationReason = WarcTruncationReason.LENGTH;
+                    break;
                 }
 
             } catch (IOException e) {
@@ -111,13 +151,6 @@ public abstract class WarcInputBuffer implements AutoCloseable {
             }
         }
 
-        // Try to close the connection as long as we haven't timed out.
-        // As per Apache HttpClient's semantics, this will reset the connection
-        // and close the stream if we have timed out.
-
-        if (truncationReason != WarcTruncationReason.TIME) {
-            IOUtils.closeQuietly(is);
-        }
     }
 
     /** Takes a Content-Range header and checks if it is complete.
@@ -218,7 +251,7 @@ class ErrorBuffer extends WarcInputBuffer {
 /** Buffer for when we have the response in memory */
 class MemoryBuffer extends WarcInputBuffer {
     byte[] data;
-    public MemoryBuffer(Header[] headers, Duration timeLimit, InputStream responseStream, int size) {
+    public MemoryBuffer(Header[] headers, HttpGet request, Duration timeLimit, InputStream responseStream, int size) {
         super(suppressContentEncoding(headers));
 
         if (!isRangeComplete(headers)) {
@@ -229,7 +262,7 @@ class MemoryBuffer extends WarcInputBuffer {
 
         var outputStream = new ByteArrayOutputStream(size);
 
-        copy(responseStream, outputStream, timeLimit);
+        copy(responseStream, request, outputStream, timeLimit);
 
         data = outputStream.toByteArray();
     }
@@ -253,7 +286,7 @@ class MemoryBuffer extends WarcInputBuffer {
 class FileBuffer extends WarcInputBuffer {
     private final Path tempFile;
 
-    public FileBuffer(Header[] headers, Duration timeLimit, InputStream responseStream) throws IOException {
+    public FileBuffer(Header[] headers, HttpGet request, Duration timeLimit, InputStream responseStream) throws IOException {
         super(suppressContentEncoding(headers));
 
         if (!isRangeComplete(headers)) {
@@ -265,7 +298,7 @@ class FileBuffer extends WarcInputBuffer {
         this.tempFile = Files.createTempFile("rsp", ".html");
 
         try (var out = Files.newOutputStream(tempFile)) {
-            copy(responseStream, out, timeLimit);
+            copy(responseStream, request, out, timeLimit);
         }
         catch (Exception ex) {
             truncationReason = WarcTruncationReason.UNSPECIFIED;

code/processes/crawling-process/java/nu/marginalia/crawl/fetcher/warc/WarcRecorder.java

@@ -1,6 +1,7 @@
 package nu.marginalia.crawl.fetcher.warc;
 
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.HttpFetcher;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 import nu.marginalia.link_parser.LinkParser;
@@ -8,9 +9,7 @@ import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.body.HttpFetchResult;
 import org.apache.hc.client5.http.classic.HttpClient;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
-import org.apache.hc.client5.http.cookie.CookieStore;
-import org.apache.hc.core5.http.ClassicHttpRequest;
 import org.apache.hc.core5.http.NameValuePair;
 import org.jetbrains.annotations.Nullable;
 import org.netpreserve.jwarc.*;
@@ -42,7 +41,7 @@ public class WarcRecorder implements AutoCloseable {
     static final int MAX_TIME = 30_000;
 
     /** Maximum (decompressed) size we'll save */
-    static final int MAX_SIZE = Integer.getInteger("crawler.maxFetchSize", 10 * 1024 * 1024);
+    static final int MAX_SIZE = Integer.getInteger("crawler.maxFetchSize", 32 * 1024 * 1024);
 
     private final WarcWriter writer;
     private final Path warcFile;
@@ -53,23 +52,15 @@ public class WarcRecorder implements AutoCloseable {
     // Affix a version string in case we need to change the format in the future
     // in some way
     private final String warcRecorderVersion = "1.0";
-    private final CookieStore cookies;
     private final LinkParser linkParser = new LinkParser();
     /**
      * Create a new WarcRecorder that will write to the given file
      *
      * @param warcFile The file to write to
      */
-    public WarcRecorder(Path warcFile, HttpFetcherImpl fetcher) throws IOException {
+    public WarcRecorder(Path warcFile) throws IOException {
         this.warcFile = warcFile;
         this.writer = new WarcWriter(warcFile);
-        this.cookies = fetcher.getCookies();
-    }
-
-    public WarcRecorder(Path warcFile, CookieStore cookies) throws IOException {
-        this.warcFile = warcFile;
-        this.writer = new WarcWriter(warcFile);
-        this.cookies = cookies;
     }
 
     /**
@@ -79,24 +70,21 @@ public class WarcRecorder implements AutoCloseable {
     public WarcRecorder() throws IOException {
         this.warcFile = Files.createTempFile("warc", ".warc.gz");
         this.writer = new WarcWriter(this.warcFile);
-        this.cookies = new BasicCookieStore();
 
         temporaryFile = true;
     }
 
-    private boolean hasCookies() {
-        return !cookies.getCookies().isEmpty();
-    }
-
     public HttpFetchResult fetch(HttpClient client,
-                                 ClassicHttpRequest request)
+                                 DomainCookies cookies,
+                                 HttpGet request)
             throws NoSuchAlgorithmException, IOException, URISyntaxException, InterruptedException
     {
-        return fetch(client, request, Duration.ofMillis(MAX_TIME));
+        return fetch(client, cookies, request, Duration.ofMillis(MAX_TIME));
     }
 
     public HttpFetchResult fetch(HttpClient client,
-                                 ClassicHttpRequest request,
+                                 DomainCookies cookies,
+                                 HttpGet request,
                                  Duration timeout)
             throws NoSuchAlgorithmException, IOException, URISyntaxException, InterruptedException
     {
@@ -113,13 +101,15 @@ public class WarcRecorder implements AutoCloseable {
         // Inject a range header to attempt to limit the size of the response
         // to the maximum size we want to store, if the server supports it.
         request.addHeader("Range", "bytes=0-"+MAX_SIZE);
-
+        cookies.paintRequest(request);
         try {
-            return client.execute(request, response -> {
+            return client.execute(request,response -> {
 
-                try (WarcInputBuffer inputBuffer = WarcInputBuffer.forResponse(response, timeout);
+                try (WarcInputBuffer inputBuffer = WarcInputBuffer.forResponse(response, request, timeout);
                      InputStream inputStream = inputBuffer.read()) {
 
+                    cookies.updateCookieStore(response);
+
                     // Build and write the request
 
                     WarcDigestBuilder requestDigestBuilder = new WarcDigestBuilder();
@@ -143,8 +133,9 @@ public class WarcRecorder implements AutoCloseable {
                     warcRequest.http(); // force HTTP header to be parsed before body is consumed so that caller can use it
                     writer.write(warcRequest);
 
-                    if (hasCookies()) {
+
-                        extraHeaders.put("X-Has-Cookies", List.of("1"));
+                    if (cookies.hasCookies()) {
+                        response.addHeader("X-Has-Cookies", 1);
                     }
 
                     byte[] responseHeaders = WarcProtocolReconstructor.getResponseHeader(response, inputBuffer.size()).getBytes(StandardCharsets.UTF_8);
@@ -259,7 +250,7 @@ public class WarcRecorder implements AutoCloseable {
         writer.write(item);
     }
 
-    private void saveOldResponse(EdgeUrl url, String contentType, int statusCode, byte[] documentBody, @Nullable String headers, ContentTags contentTags) {
+    private void saveOldResponse(EdgeUrl url, DomainCookies domainCookies, String contentType, int statusCode, byte[] documentBody, @Nullable String headers, ContentTags contentTags) {
         try {
             WarcDigestBuilder responseDigestBuilder = new WarcDigestBuilder();
             WarcDigestBuilder payloadDigestBuilder = new WarcDigestBuilder();
@@ -320,7 +311,7 @@ public class WarcRecorder implements AutoCloseable {
                     .date(Instant.now())
                     .body(MediaType.HTTP_RESPONSE, responseDataBuffer.copyBytes());
 
-            if (hasCookies()) {
+            if (domainCookies.hasCookies() || (headers != null && headers.contains("Set-Cookie:"))) {
                 builder.addHeader("X-Has-Cookies", "1");
             }
 
@@ -340,8 +331,8 @@ public class WarcRecorder implements AutoCloseable {
      * an E-Tag or Last-Modified header, and the server responds with a 304 Not Modified.  In this
      * scenario we want to record the data as it was in the previous crawl, but not re-fetch it.
      */
-    public void writeReferenceCopy(EdgeUrl url, String contentType, int statusCode, byte[] documentBody, @Nullable String headers, ContentTags ctags) {
+    public void writeReferenceCopy(EdgeUrl url, DomainCookies cookies, String contentType, int statusCode, byte[] documentBody, @Nullable String headers, ContentTags ctags) {
-        saveOldResponse(url, contentType, statusCode, documentBody, headers, ctags);
+        saveOldResponse(url, cookies, contentType, statusCode, documentBody, headers, ctags);
     }
 
     public void writeWarcinfoHeader(String ip, EdgeDomain domain, HttpFetcherImpl.DomainProbeResult result) throws IOException {

code/processes/crawling-process/java/nu/marginalia/crawl/logic/DomainLocks.java

@@ -3,6 +3,7 @@ package nu.marginalia.crawl.logic;
 import nu.marginalia.model.EdgeDomain;
 
 import java.util.Map;
+import java.util.Optional;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.Semaphore;
 
@@ -19,8 +20,22 @@ public class DomainLocks {
      * and may be held by another thread.  The caller is responsible for locking and  releasing the lock.
      */
     public DomainLock lockDomain(EdgeDomain domain) throws InterruptedException {
-        return new DomainLock(domain.toString(),
+        var sem = locks.computeIfAbsent(domain.topDomain.toLowerCase(), this::defaultPermits);
-                locks.computeIfAbsent(domain.topDomain.toLowerCase(), this::defaultPermits));
+
+        sem.acquire();
+
+        return new DomainLock(sem);
+    }
+
+    public Optional<DomainLock> tryLockDomain(EdgeDomain domain) {
+        var sem = locks.computeIfAbsent(domain.topDomain.toLowerCase(), this::defaultPermits);
+        if (sem.tryAcquire(1)) {
+            return Optional.of(new DomainLock(sem));
+        }
+        else {
+            // We don't have a lock, so we return an empty optional
+            return Optional.empty();
+        }
     }
 
     private Semaphore defaultPermits(String topDomain) {
@@ -28,23 +43,27 @@ public class DomainLocks {
             return new Semaphore(16);
         if (topDomain.equals("blogspot.com"))
             return new Semaphore(8);
-
+        if (topDomain.equals("tumblr.com"))
+            return new Semaphore(8);
         if (topDomain.equals("neocities.org"))
-            return new Semaphore(4);
+            return new Semaphore(8);
         if (topDomain.equals("github.io"))
-            return new Semaphore(4);
+            return new Semaphore(8);
 
+        // Substack really dislikes broad-scale crawlers, so we need to be careful
+        // to not get blocked.
         if (topDomain.equals("substack.com")) {
             return new Semaphore(1);
         }
-        if (topDomain.endsWith(".edu")) {
-            return new Semaphore(1);
-        }
 
         return new Semaphore(2);
     }
 
-    public boolean canLock(EdgeDomain domain) {
+    /** Returns true if the domain is lockable, i.e. if it is not already locked by another thread.
+     * (this is just a hint, and does not guarantee that the domain is actually lockable any time
+     * after this method returns true)
+     */
+    public boolean isLockableHint(EdgeDomain domain) {
         Semaphore sem = locks.get(domain.topDomain.toLowerCase());
         if (null == sem)
             return true;
@@ -53,22 +72,16 @@ public class DomainLocks {
     }
 
     public static class DomainLock implements AutoCloseable {
-        private final String domainName;
         private final Semaphore semaphore;
 
-        DomainLock(String domainName, Semaphore semaphore) throws InterruptedException {
+        DomainLock(Semaphore semaphore) {
-            this.domainName = domainName;
             this.semaphore = semaphore;
-
-            Thread.currentThread().setName("crawling:" + domainName + " [await domain lock]");
-            semaphore.acquire();
-            Thread.currentThread().setName("crawling:" + domainName);
         }
 
         @Override
         public void close() throws Exception {
             semaphore.release();
-            Thread.currentThread().setName("crawling:" + domainName + " [wrapping up]");
+            Thread.currentThread().setName("[idle]");
         }
     }
 }

code/processes/crawling-process/java/nu/marginalia/crawl/retreival/CrawlerRetreiver.java

@@ -6,6 +6,7 @@ import nu.marginalia.contenttype.ContentType;
 import nu.marginalia.crawl.CrawlerMain;
 import nu.marginalia.crawl.DomainStateDb;
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.HttpFetcher;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
 import nu.marginalia.crawl.logic.LinkFilterSelector;
@@ -51,9 +52,10 @@ public class CrawlerRetreiver implements AutoCloseable {
     private final DomainStateDb domainStateDb;
     private final WarcRecorder warcRecorder;
     private final CrawlerRevisitor crawlerRevisitor;
+    private final DomainCookies cookies = new DomainCookies();
 
     private static final CrawlerConnectionThrottle connectionThrottle = new CrawlerConnectionThrottle(
-            Duration.ofMillis(50) // pace the connections to avoid network congestion at startup
+            Duration.ofSeconds(1) // pace the connections to avoid network congestion at startup
     );
 
     int errorCount = 0;
@@ -124,7 +126,7 @@ public class CrawlerRetreiver implements AutoCloseable {
                     }
 
                     Instant recrawlStart = Instant.now();
-                    CrawlerRevisitor.RecrawlMetadata recrawlMetadata = crawlerRevisitor.recrawl(oldCrawlData, robotsRules, delayTimer);
+                    CrawlerRevisitor.RecrawlMetadata recrawlMetadata = crawlerRevisitor.recrawl(oldCrawlData, cookies, robotsRules, delayTimer);
                     Duration recrawlTime = Duration.between(recrawlStart, Instant.now());
 
                     // Play back the old crawl data (if present) and fetch the documents comparing etags and last-modified
@@ -274,7 +276,7 @@ public class CrawlerRetreiver implements AutoCloseable {
         try {
             var url = rootUrl.withPathAndParam("/", null);
 
-            HttpFetchResult result = fetcher.fetchContent(url, warcRecorder, timer, ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+            HttpFetchResult result = fetcher.fetchContent(url, warcRecorder, cookies, timer, ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
             timer.waitFetchDelay(0);
 
             if (result instanceof HttpFetchResult.ResultRedirect(EdgeUrl location)) {
@@ -337,7 +339,7 @@ public class CrawlerRetreiver implements AutoCloseable {
 
             // Grab the favicon if it exists
 
-            if (fetcher.fetchContent(faviconUrl, warcRecorder, timer, ContentTags.empty(), HttpFetcher.ProbeType.DISABLED) instanceof HttpFetchResult.ResultOk iconResult) {
+            if (fetcher.fetchContent(faviconUrl, warcRecorder, cookies, timer, ContentTags.empty(), HttpFetcher.ProbeType.DISABLED) instanceof HttpFetchResult.ResultOk iconResult) {
                 String contentType = iconResult.header("Content-Type");
                 byte[] iconData = iconResult.getBodyBytes();
 
@@ -407,7 +409,7 @@ public class CrawlerRetreiver implements AutoCloseable {
         if (parsedOpt.isEmpty())
             return false;
 
-        HttpFetchResult result = fetcher.fetchContent(parsedOpt.get(), warcRecorder, timer, ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        HttpFetchResult result = fetcher.fetchContent(parsedOpt.get(), warcRecorder, cookies, timer, ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
         timer.waitFetchDelay(0);
 
         if (!(result instanceof HttpFetchResult.ResultOk ok)) {
@@ -435,7 +437,7 @@ public class CrawlerRetreiver implements AutoCloseable {
     {
         var contentTags = reference.getContentTags();
 
-        HttpFetchResult fetchedDoc = fetcher.fetchContent(top, warcRecorder, timer, contentTags, HttpFetcher.ProbeType.FULL);
+        HttpFetchResult fetchedDoc = fetcher.fetchContent(top, warcRecorder, cookies, timer, contentTags, HttpFetcher.ProbeType.FULL);
         timer.waitFetchDelay();
 
         if (Thread.interrupted()) {
@@ -461,7 +463,7 @@ public class CrawlerRetreiver implements AutoCloseable {
                 {
                     var doc = reference.doc();
 
-                    warcRecorder.writeReferenceCopy(top, doc.contentType, doc.httpStatus, doc.documentBodyBytes, doc.headers, contentTags);
+                    warcRecorder.writeReferenceCopy(top, cookies, doc.contentType, doc.httpStatus, doc.documentBodyBytes, doc.headers, contentTags);
 
                     fetchedDoc = new HttpFetchResult.Result304ReplacedWithReference(doc.url,
                             new ContentType(doc.contentType, "UTF-8"),

code/processes/crawling-process/java/nu/marginalia/crawl/retreival/revisit/CrawlerRevisitor.java

@@ -2,6 +2,7 @@ package nu.marginalia.crawl.retreival.revisit;
 
 import crawlercommons.robots.SimpleRobotRules;
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
 import nu.marginalia.crawl.retreival.CrawlDataReference;
 import nu.marginalia.crawl.retreival.CrawlDelayTimer;
@@ -37,6 +38,7 @@ public class CrawlerRevisitor {
 
     /** Performs a re-crawl of old documents, comparing etags and last-modified */
     public RecrawlMetadata recrawl(CrawlDataReference oldCrawlData,
+                       DomainCookies cookies,
                        SimpleRobotRules robotsRules,
                        CrawlDelayTimer delayTimer)
     throws InterruptedException {
@@ -132,6 +134,7 @@ public class CrawlerRevisitor {
                 }
                 // Add a WARC record so we don't repeat this
                 warcRecorder.writeReferenceCopy(url,
+                        cookies,
                         doc.contentType,
                         doc.httpStatus,
                         doc.documentBodyBytes,

code/processes/crawling-process/model/java/nu/marginalia/ContentTypes.java

@@ -6,6 +6,7 @@ public class ContentTypes {
     public static final Set<String> acceptedContentTypes = Set.of("application/xhtml+xml",
             "application/xhtml",
             "text/html",
+            "application/pdf",
             "image/x-icon",
             "text/plain");
 
@@ -19,4 +20,9 @@ public class ContentTypes {
         return false;
     }
 
+    public static boolean isBinary(String contentTypeHeader) {
+        String lcHeader = contentTypeHeader.toLowerCase();
+        return lcHeader.startsWith("application/pdf");
+    }
+
 }

code/processes/crawling-process/model/java/nu/marginalia/io/crawldata/format/SlopSerializableCrawlDataStream.java

@@ -37,8 +37,12 @@ public class SlopSerializableCrawlDataStream implements AutoCloseable, Serializa
             public boolean filter(String url, int status, String contentType) {
                 String ctLc = contentType.toLowerCase();
 
+                // Permit all plain text content types
                 if (ctLc.startsWith("text/"))
                     return true;
+                // PDF
+                else if (ctLc.startsWith("application/pdf"))
+                    return true;
                 else if (ctLc.startsWith("x-marginalia/"))
                     return true;
 

code/processes/crawling-process/model/java/nu/marginalia/model/body/ContentTypeLogic.java

@@ -10,7 +10,7 @@ import java.util.regex.Pattern;
 
 public class ContentTypeLogic {
 
-    private static final Predicate<String> probableHtmlPattern = Pattern.compile("^.*\\.(htm|html|php|txt|md)$").asMatchPredicate();
+    private static final Predicate<String> probableGoodPattern = Pattern.compile("^.*\\.(htm|html|php|txt|md|pdf)$").asMatchPredicate();
     private static final Predicate<String> probableBinaryPattern = Pattern.compile("^.*\\.[a-z]+$").asMatchPredicate();
     private static final Set<String> blockedContentTypes = Set.of("text/css", "text/javascript");
     private static final List<String> acceptedContentTypePrefixes = List.of(
@@ -22,6 +22,7 @@ public class ContentTypeLogic {
             "application/rss+xml",
             "application/x-rss+xml",
             "application/rdf+xml",
+            "application/pdf",
             "x-rss+xml"
     );
     private boolean allowAllContentTypes = false;
@@ -34,7 +35,7 @@ public class ContentTypeLogic {
     public boolean isUrlLikeBinary(EdgeUrl url) {
         String pathLowerCase = url.path.toLowerCase();
 
-        if (probableHtmlPattern.test(pathLowerCase))
+        if (probableGoodPattern.test(pathLowerCase))
             return false;
 
         return probableBinaryPattern.test(pathLowerCase);

code/processes/crawling-process/model/java/nu/marginalia/slop/SlopCrawlDataRecord.java

@@ -216,6 +216,11 @@ public record SlopCrawlDataRecord(String domain,
             return false;
         }
 
+        // If the format is binary, we don't want to translate it if the response is truncated
+        if (response.truncated() != WarcTruncationReason.NOT_TRUNCATED && ContentTypes.isBinary(contentType)) {
+            return false;
+        }
+
         return true;
     }
 

code/processes/crawling-process/test/nu/marginalia/crawl/fetcher/HttpFetcherImplContentTypeProbeTest.java

@@ -11,6 +11,8 @@ import org.junit.jupiter.api.*;
 import java.io.IOException;
 import java.net.URISyntaxException;
 
+import static org.junit.jupiter.api.Assertions.assertEquals;
+
 @Tag("slow")
 class HttpFetcherImplContentTypeProbeTest {
 
@@ -85,55 +87,59 @@ class HttpFetcherImplContentTypeProbeTest {
 
     @AfterEach
     public void tearDown() throws IOException {
+        var stats = fetcher.getPoolStats();
+        assertEquals(0, stats.getLeased());
+        assertEquals(0, stats.getPending());
+
         fetcher.close();
     }
 
     @Test
     public void testProbeContentTypeHtmlShortcircuitPath() throws URISyntaxException {
-        var result = fetcher.probeContentType(new EdgeUrl("https://localhost/test.html"), new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(new EdgeUrl("https://localhost/test.html"), new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
-        Assertions.assertInstanceOf(HttpFetcher.ContentTypeProbeResult.Ok.class, result);
+        Assertions.assertInstanceOf(HttpFetcher.ContentTypeProbeResult.NoOp.class, result);
     }
 
 
     @Test
     public void testProbeContentTypeHtmlShortcircuitTags() {
-        var result = fetcher.probeContentType(contentTypeBinaryUrl, new CrawlDelayTimer(50), new ContentTags("a", "b"));
+        var result = fetcher.probeContentType(contentTypeBinaryUrl, new DomainCookies(), new CrawlDelayTimer(50), new ContentTags("a", "b"));
-        Assertions.assertInstanceOf(HttpFetcher.ContentTypeProbeResult.Ok.class, result);
+        Assertions.assertInstanceOf(HttpFetcher.ContentTypeProbeResult.NoOp.class, result);
     }
 
     @Test
     public void testProbeContentTypeHtml() {
-        var result = fetcher.probeContentType(contentTypeHtmlUrl, new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(contentTypeHtmlUrl, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
         Assertions.assertEquals(new HttpFetcher.ContentTypeProbeResult.Ok(contentTypeHtmlUrl), result);
     }
 
     @Test
     public void testProbeContentTypeBinary() {
-        var result = fetcher.probeContentType(contentTypeBinaryUrl, new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(contentTypeBinaryUrl, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
         Assertions.assertEquals(new HttpFetcher.ContentTypeProbeResult.BadContentType("application/octet-stream", 200), result);
     }
 
     @Test
     public void testProbeContentTypeRedirect() {
-        var result = fetcher.probeContentType(redirectUrl, new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(redirectUrl, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
         Assertions.assertEquals(new HttpFetcher.ContentTypeProbeResult.Redirect(contentTypeHtmlUrl), result);
     }
 
     @Test
     public void testProbeContentTypeBadHttpStatus() {
-        var result = fetcher.probeContentType(badHttpStatusUrl, new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(badHttpStatusUrl, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
         Assertions.assertEquals(new HttpFetcher.ContentTypeProbeResult.HttpError(500, "Bad status code"), result);
     }
 
     @Test
     public void testOnlyGetAllowed() {
-        var result = fetcher.probeContentType(onlyGetAllowedUrl, new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(onlyGetAllowedUrl, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
         Assertions.assertEquals(new HttpFetcher.ContentTypeProbeResult.Ok(onlyGetAllowedUrl), result);
     }
 
     @Test
     public void testTimeout() {
-        var result = fetcher.probeContentType(timeoutUrl, new CrawlDelayTimer(50), ContentTags.empty());
+        var result = fetcher.probeContentType(timeoutUrl, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
         Assertions.assertInstanceOf(HttpFetcher.ContentTypeProbeResult.Timeout.class, result);
     }
 

code/processes/crawling-process/test/nu/marginalia/crawl/fetcher/HttpFetcherImplDomainProbeTest.java

@@ -12,6 +12,8 @@ import org.junit.jupiter.api.*;
 import java.io.IOException;
 import java.net.URISyntaxException;
 
+import static org.junit.jupiter.api.Assertions.assertEquals;
+
 @Tag("slow")
 class HttpFetcherImplDomainProbeTest {
 
@@ -47,6 +49,10 @@ class HttpFetcherImplDomainProbeTest {
 
     @AfterEach
     public void tearDown() throws IOException {
+        var stats = fetcher.getPoolStats();
+        assertEquals(0, stats.getLeased());
+        assertEquals(0, stats.getPending());
+
         fetcher.close();
     }
 

code/processes/crawling-process/test/nu/marginalia/crawl/fetcher/HttpFetcherImplFetchTest.java

@@ -31,6 +31,7 @@ class HttpFetcherImplFetchTest {
     private static String lastModified = "Wed, 21 Oct 2024 07:28:00 GMT";
 
     private static EdgeUrl okUrl;
+    private static EdgeUrl okUrlSetsCookie;
     private static EdgeUrl okRangeResponseUrl;
     private static EdgeUrl okUrlWith304;
 
@@ -39,6 +40,8 @@ class HttpFetcherImplFetchTest {
     private static EdgeUrl badHttpStatusUrl;
     private static EdgeUrl keepAliveUrl;
 
+    private static EdgeUrl pdfUrl;
+
     @BeforeAll
     public static void setupAll() throws URISyntaxException {
         wireMockServer =
@@ -88,6 +91,19 @@ class HttpFetcherImplFetchTest {
                         .withStatus(200)
                         .withBody("Hello World")));
 
+        okUrlSetsCookie = new EdgeUrl("http://localhost:18089/okSetCookie.bin");
+        wireMockServer.stubFor(WireMock.head(WireMock.urlEqualTo(okUrlSetsCookie.path))
+                .willReturn(WireMock.aResponse()
+                        .withHeader("Content-Type", "text/html")
+                        .withHeader("Set-Cookie", "test=1")
+                        .withStatus(200)));
+        wireMockServer.stubFor(WireMock.get(WireMock.urlEqualTo(okUrlSetsCookie.path))
+                .willReturn(WireMock.aResponse()
+                        .withHeader("Content-Type", "text/html")
+                        .withHeader("Set-Cookie", "test=1")
+                        .withStatus(200)
+                        .withBody("Hello World")));
+
         okUrlWith304 = new EdgeUrl("http://localhost:18089/ok304.bin");
         wireMockServer.stubFor(WireMock.head(WireMock.urlEqualTo(okUrlWith304.path))
                 .willReturn(WireMock.aResponse()
@@ -117,6 +133,15 @@ class HttpFetcherImplFetchTest {
                         .withHeader("Keep-Alive", "max=4, timeout=30")
                         .withBody("Hello")
                         ));
+
+
+        pdfUrl = new EdgeUrl("http://localhost:18089/test.pdf");
+        wireMockServer.stubFor(WireMock.get(WireMock.urlEqualTo(pdfUrl.path))
+                .willReturn(WireMock.aResponse()
+                        .withHeader("Content-Type", "application/pdf")
+                        .withStatus(200)
+                        .withBody("Hello World")));
+
         wireMockServer.start();
 
     }
@@ -134,20 +159,31 @@ class HttpFetcherImplFetchTest {
     public void setUp() throws IOException {
         fetcher = new HttpFetcherImpl(new UserAgent("test.marginalia.nu", "test.marginalia.nu"));
         warcFile = Files.createTempFile(getClass().getSimpleName(), ".warc");
-        warcRecorder = new WarcRecorder(warcFile, fetcher);
+        warcRecorder = new WarcRecorder(warcFile);
     }
 
     @AfterEach
     public void tearDown() throws IOException {
+        var stats = fetcher.getPoolStats();
+        assertEquals(0, stats.getLeased());
+        assertEquals(0, stats.getPending());
+
+        System.out.println(stats);
+
         fetcher.close();
         warcRecorder.close();
         Files.deleteIfExists(warcFile);
     }
 
 
+    @Test
+    public void testFoo() {
+        fetcher.fetchSitemapUrls("https://www.marginalia.nu/sitemap.xml", new CrawlDelayTimer(100));
+    }
+
     @Test
     public void testOk_NoProbe() throws IOException {
-        var result = fetcher.fetchContent(okUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(okUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
         Assertions.assertTrue(result.isOk());
@@ -158,12 +194,29 @@ class HttpFetcherImplFetchTest {
         Assertions.assertInstanceOf(WarcResponse.class, warcRecords.get(1));
 
         WarcResponse response = (WarcResponse) warcRecords.get(1);
-        assertEquals("0", response.headers().first("X-Has-Cookies").orElse("0"));
+        assertEquals("0", response.http().headers().first("X-Has-Cookies").orElse("0"));
+    }
+
+    @Test
+    public void testOkSetsCookie() throws IOException {
+        var cookies = new DomainCookies();
+        var result = fetcher.fetchContent(okUrlSetsCookie, warcRecorder, cookies, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+
+        Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
+        Assertions.assertTrue(result.isOk());
+
+        List<WarcRecord> warcRecords = getWarcRecords();
+        assertEquals(2, warcRecords.size());
+        Assertions.assertInstanceOf(WarcRequest.class, warcRecords.get(0));
+        Assertions.assertInstanceOf(WarcResponse.class, warcRecords.get(1));
+
+        WarcResponse response = (WarcResponse) warcRecords.get(1);
+        assertEquals("1", response.http().headers().first("X-Has-Cookies").orElse("0"));
     }
 
     @Test
     public void testOk_FullProbe() {
-        var result = fetcher.fetchContent(okUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
+        var result = fetcher.fetchContent(okUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
         Assertions.assertTrue(result.isOk());
@@ -171,7 +224,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testOk304_NoProbe() {
-        var result = fetcher.fetchContent(okUrlWith304, warcRecorder, new CrawlDelayTimer(1000), new ContentTags(etag, lastModified), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(okUrlWith304, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), new ContentTags(etag, lastModified), HttpFetcher.ProbeType.DISABLED);
 
         Assertions.assertInstanceOf(HttpFetchResult.Result304Raw.class, result);
         System.out.println(result);
@@ -180,7 +233,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testOk304_FullProbe() {
-        var result = fetcher.fetchContent(okUrlWith304, warcRecorder, new CrawlDelayTimer(1000), new ContentTags(etag, lastModified), HttpFetcher.ProbeType.FULL);
+        var result = fetcher.fetchContent(okUrlWith304, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), new ContentTags(etag, lastModified), HttpFetcher.ProbeType.FULL);
 
         Assertions.assertInstanceOf(HttpFetchResult.Result304Raw.class, result);
         System.out.println(result);
@@ -188,7 +241,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testBadStatus_NoProbe() throws IOException {
-        var result = fetcher.fetchContent(badHttpStatusUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(badHttpStatusUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
         Assertions.assertFalse(result.isOk());
@@ -202,7 +255,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testBadStatus_FullProbe() {
-        var result = fetcher.fetchContent(badHttpStatusUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
+        var result = fetcher.fetchContent(badHttpStatusUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
         Assertions.assertFalse(result.isOk());
@@ -212,7 +265,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testRedirect_NoProbe() throws URISyntaxException, IOException {
-        var result = fetcher.fetchContent(redirectUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(redirectUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultRedirect.class, result);
         assertEquals(new EdgeUrl("http://localhost:18089/test.html.bin"), ((HttpFetchResult.ResultRedirect) result).url());
@@ -225,7 +278,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testRedirect_FullProbe() throws URISyntaxException {
-        var result = fetcher.fetchContent(redirectUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
+        var result = fetcher.fetchContent(redirectUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultRedirect.class, result);
         assertEquals(new EdgeUrl("http://localhost:18089/test.html.bin"), ((HttpFetchResult.ResultRedirect) result).url());
@@ -238,7 +291,7 @@ class HttpFetcherImplFetchTest {
     public void testFetchTimeout_NoProbe() throws IOException, URISyntaxException {
         Instant requestStart = Instant.now();
 
-        var result = fetcher.fetchContent(timeoutUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(timeoutUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultException.class, result);
 
@@ -262,7 +315,7 @@ class HttpFetcherImplFetchTest {
 
     @Test
     public void testRangeResponse() throws IOException {
-        var result = fetcher.fetchContent(okRangeResponseUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(okRangeResponseUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
         Assertions.assertTrue(result.isOk());
@@ -279,7 +332,7 @@ class HttpFetcherImplFetchTest {
     @Test
     public void testFetchTimeout_Probe() throws IOException, URISyntaxException {
         Instant requestStart = Instant.now();
-        var result = fetcher.fetchContent(timeoutUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
+        var result = fetcher.fetchContent(timeoutUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
         Instant requestEnd = Instant.now();
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultException.class, result);
@@ -302,7 +355,15 @@ class HttpFetcherImplFetchTest {
     @Test
     public void testKeepaliveUrl() {
         // mostly for smoke testing and debugger utility
-        var result = fetcher.fetchContent(keepAliveUrl, warcRecorder, new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+        var result = fetcher.fetchContent(keepAliveUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.DISABLED);
+
+        Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
+        Assertions.assertTrue(result.isOk());
+    }
+
+    @Test
+    public void testPdf() {
+        var result = fetcher.fetchContent(pdfUrl, warcRecorder, new DomainCookies(), new CrawlDelayTimer(1000), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
 
         Assertions.assertInstanceOf(HttpFetchResult.ResultOk.class, result);
         Assertions.assertTrue(result.isOk());
@@ -319,6 +380,13 @@ class HttpFetcherImplFetchTest {
             WarcXEntityRefused.register(reader);
 
             for (var record : reader) {
+                // Load the body, we need to do this before we close the reader to have access to the content.
+                if (record instanceof WarcRequest req) {
+                    req.http();
+                } else if (record instanceof WarcResponse rsp) {
+                    rsp.http();
+                }
+
                 records.add(record);
             }
         }

code/processes/crawling-process/test/nu/marginalia/crawl/retreival/CrawlerWarcResynchronizerTest.java

@@ -1,12 +1,12 @@
 package nu.marginalia.crawl.retreival;
 
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
 import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.model.EdgeUrl;
 import org.apache.hc.client5.http.classic.HttpClient;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
 import org.apache.hc.client5.http.impl.classic.HttpClients;
-import org.apache.hc.core5.http.io.support.ClassicRequestBuilder;
 import org.junit.jupiter.api.AfterEach;
 import org.junit.jupiter.api.BeforeEach;
 import org.junit.jupiter.api.Test;
@@ -45,7 +45,7 @@ class CrawlerWarcResynchronizerTest {
 
     @Test
     void run() throws IOException, URISyntaxException {
-        try (var oldRecorder = new WarcRecorder(fileName, new BasicCookieStore())) {
+        try (var oldRecorder = new WarcRecorder(fileName)) {
             fetchUrl(oldRecorder, "https://www.marginalia.nu/");
             fetchUrl(oldRecorder, "https://www.marginalia.nu/log/");
             fetchUrl(oldRecorder, "https://www.marginalia.nu/feed/");
@@ -55,7 +55,7 @@ class CrawlerWarcResynchronizerTest {
 
         var crawlFrontier = new DomainCrawlFrontier(new EdgeDomain("www.marginalia.nu"), List.of(), 100);
 
-        try (var newRecorder = new WarcRecorder(outputFile, new BasicCookieStore())) {
+        try (var newRecorder = new WarcRecorder(outputFile)) {
             new CrawlerWarcResynchronizer(crawlFrontier, newRecorder).run(fileName);
         }
 
@@ -78,10 +78,10 @@ class CrawlerWarcResynchronizerTest {
     }
 
     void fetchUrl(WarcRecorder recorder, String url) throws NoSuchAlgorithmException, IOException, URISyntaxException, InterruptedException {
-        var req = ClassicRequestBuilder.get(new java.net.URI(url))
+        HttpGet request = new HttpGet(url);
-                .addHeader("User-agent", "test.marginalia.nu")
+        request.addHeader("User-agent", "test.marginalia.nu");
-                .addHeader("Accept-Encoding", "gzip")
+        request.addHeader("Accept-Encoding", "gzip");
-                .build();
+
-        recorder.fetch(httpClient, req);
+        recorder.fetch(httpClient, new DomainCookies(), request);
     }
 }

code/processes/crawling-process/test/nu/marginalia/crawl/retreival/fetcher/ContentTypeProberTest.java

@@ -2,6 +2,7 @@ package nu.marginalia.crawl.retreival.fetcher;
 
 import com.sun.net.httpserver.HttpServer;
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.HttpFetcher;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 import nu.marginalia.crawl.retreival.CrawlDelayTimer;
@@ -88,7 +89,7 @@ class ContentTypeProberTest {
 
     @Test
     void probeContentTypeOk() throws Exception {
-        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(htmlEndpoint, new CrawlDelayTimer(50), ContentTags.empty());
+        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(htmlEndpoint, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
 
         System.out.println(result);
 
@@ -97,7 +98,7 @@ class ContentTypeProberTest {
 
     @Test
     void probeContentTypeRedir() throws Exception {
-        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(htmlRedirEndpoint, new CrawlDelayTimer(50), ContentTags.empty());
+        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(htmlRedirEndpoint, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
 
         System.out.println(result);
 
@@ -106,7 +107,7 @@ class ContentTypeProberTest {
 
     @Test
     void probeContentTypeBad() throws Exception {
-        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(binaryEndpoint, new CrawlDelayTimer(50), ContentTags.empty());
+        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(binaryEndpoint, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
 
         System.out.println(result);
 
@@ -115,7 +116,7 @@ class ContentTypeProberTest {
 
     @Test
     void probeContentTypeTimeout() throws Exception {
-        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(timeoutEndpoint, new CrawlDelayTimer(50), ContentTags.empty());
+        HttpFetcher.ContentTypeProbeResult result = fetcher.probeContentType(timeoutEndpoint, new DomainCookies(), new CrawlDelayTimer(50), ContentTags.empty());
 
         System.out.println(result);
 

code/processes/crawling-process/test/nu/marginalia/crawl/retreival/fetcher/WarcRecorderFakeServerTest.java

@@ -1,11 +1,11 @@
 package nu.marginalia.crawl.retreival.fetcher;
 
 import com.sun.net.httpserver.HttpServer;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
 import org.apache.hc.client5.http.classic.HttpClient;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
 import org.apache.hc.client5.http.impl.classic.HttpClients;
-import org.apache.hc.core5.http.io.support.ClassicRequestBuilder;
 import org.junit.jupiter.api.*;
 import org.netpreserve.jwarc.WarcReader;
 import org.netpreserve.jwarc.WarcRequest;
@@ -51,14 +51,14 @@ class WarcRecorderFakeServerTest {
                 os.write("<html><body>hello</body></html>".getBytes());
                 os.flush();
                 try {
-                    TimeUnit.SECONDS.sleep(1);
+                    TimeUnit.SECONDS.sleep(2);
                 } catch (InterruptedException e) {
                     throw new RuntimeException(e);
                 }
                 os.write(":".getBytes());
                 os.flush();
                 try {
-                    TimeUnit.SECONDS.sleep(1);
+                    TimeUnit.SECONDS.sleep(2);
                 } catch (InterruptedException e) {
                     throw new RuntimeException(e);
                 }
@@ -89,24 +89,22 @@ class WarcRecorderFakeServerTest {
         fileNameWarc = Files.createTempFile("test", ".warc");
         fileNameParquet = Files.createTempFile("test", ".parquet");
 
-        client = new WarcRecorder(fileNameWarc, new BasicCookieStore());
+        client = new WarcRecorder(fileNameWarc);
     }
 
     @AfterEach
     public void tearDown() throws Exception {
+
         client.close();
         Files.delete(fileNameWarc);
     }
 
     @Test
     public void fetchFast() throws Exception {
-        client.fetch(httpClient,
+        HttpGet request = new HttpGet("http://localhost:14510/fast");
-                ClassicRequestBuilder
+        request.addHeader("User-agent", "test.marginalia.nu");
-                        .get(new java.net.URI("http://localhost:14510/fast"))
+        request.addHeader("Accept-Encoding", "gzip");
-                        .addHeader("User-agent", "test.marginalia.nu")
+        client.fetch(httpClient, new DomainCookies(), request);
-                        .addHeader("Accept-Encoding", "gzip")
-                        .build()
-        );
 
         Map<String, String> sampleData = new HashMap<>();
         try (var warcReader = new WarcReader(fileNameWarc)) {
@@ -127,11 +125,13 @@ class WarcRecorderFakeServerTest {
     public void fetchSlow() throws Exception {
         Instant start = Instant.now();
 
+        HttpGet request = new HttpGet("http://localhost:14510/slow");
+        request.addHeader("User-agent", "test.marginalia.nu");
+        request.addHeader("Accept-Encoding", "gzip");
+
         client.fetch(httpClient,
-                ClassicRequestBuilder.get(new java.net.URI("http://localhost:14510/slow"))
+                new DomainCookies(),
-                        .addHeader("User-agent", "test.marginalia.nu")
+                request,
-                        .addHeader("Accept-Encoding", "gzip")
-                        .build(),
                 Duration.ofSeconds(1)
         );
         Instant end = Instant.now();
@@ -149,6 +149,8 @@ class WarcRecorderFakeServerTest {
             });
         }
 
+        System.out.println(
+                Files.readString(fileNameWarc));
         System.out.println(sampleData);
 
         // Timeout is set to 1 second, but the server will take 5 seconds to respond,

code/processes/crawling-process/test/nu/marginalia/crawl/retreival/fetcher/WarcRecorderTest.java

@@ -2,14 +2,14 @@ package nu.marginalia.crawl.retreival.fetcher;
 
 import nu.marginalia.UserAgent;
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
+import nu.marginalia.io.SerializableCrawlDataStream;
 import nu.marginalia.model.EdgeUrl;
-import nu.marginalia.parquet.crawldata.CrawledDocumentParquetRecordFileReader;
+import nu.marginalia.slop.SlopCrawlDataRecord;
-import nu.marginalia.parquet.crawldata.CrawledDocumentParquetRecordFileWriter;
 import org.apache.hc.client5.http.classic.HttpClient;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
+import org.apache.hc.client5.http.classic.methods.HttpGet;
 import org.apache.hc.client5.http.impl.classic.HttpClients;
-import org.apache.hc.core5.http.io.support.ClassicRequestBuilder;
 import org.junit.jupiter.api.AfterEach;
 import org.junit.jupiter.api.BeforeEach;
 import org.junit.jupiter.api.Test;
@@ -24,13 +24,14 @@ import java.nio.file.Files;
 import java.nio.file.Path;
 import java.security.NoSuchAlgorithmException;
 import java.util.HashMap;
+import java.util.List;
 import java.util.Map;
 
 import static org.junit.jupiter.api.Assertions.assertEquals;
 
 class WarcRecorderTest {
     Path fileNameWarc;
-    Path fileNameParquet;
+    Path fileNameSlop;
     WarcRecorder client;
 
     HttpClient httpClient;
@@ -39,9 +40,9 @@ class WarcRecorderTest {
         httpClient = HttpClients.createDefault();
 
         fileNameWarc = Files.createTempFile("test", ".warc");
-        fileNameParquet = Files.createTempFile("test", ".parquet");
+        fileNameSlop = Files.createTempFile("test", ".slop.zip");
 
-        client = new WarcRecorder(fileNameWarc, new BasicCookieStore());
+        client = new WarcRecorder(fileNameWarc);
     }
 
     @AfterEach
@@ -52,12 +53,12 @@ class WarcRecorderTest {
 
     @Test
     void fetch() throws NoSuchAlgorithmException, IOException, URISyntaxException, InterruptedException {
-        client.fetch(httpClient,
+
-                ClassicRequestBuilder.get(new java.net.URI("https://www.marginalia.nu/"))
+        HttpGet request = new HttpGet("https://www.marginalia.nu/");
-                        .addHeader("User-agent", "test.marginalia.nu")
+        request.addHeader("User-agent", "test.marginalia.nu");
-                        .addHeader("Accept-Encoding", "gzip")
+        request.addHeader("Accept-Encoding", "gzip");
-                        .build()
+
-        );
+        client.fetch(httpClient, new DomainCookies(), request);
 
         Map<String, String> sampleData = new HashMap<>();
         try (var warcReader = new WarcReader(fileNameWarc)) {
@@ -78,8 +79,9 @@ class WarcRecorderTest {
     @Test
     public void flagAsSkipped() throws IOException, URISyntaxException {
 
-        try (var recorder = new WarcRecorder(fileNameWarc, new BasicCookieStore())) {
+        try (var recorder = new WarcRecorder(fileNameWarc)) {
             recorder.writeReferenceCopy(new EdgeUrl("https://www.marginalia.nu/"),
+                    new DomainCookies(),
                     "text/html",
                     200,
                     "<?doctype html><html><body>test</body></html>".getBytes(),
@@ -102,8 +104,9 @@ class WarcRecorderTest {
     @Test
     public void flagAsSkippedNullBody() throws IOException, URISyntaxException {
 
-        try (var recorder = new WarcRecorder(fileNameWarc, new BasicCookieStore())) {
+        try (var recorder = new WarcRecorder(fileNameWarc)) {
             recorder.writeReferenceCopy(new EdgeUrl("https://www.marginalia.nu/"),
+                    new DomainCookies(),
                     "text/html",
                     200,
                     null,
@@ -114,8 +117,9 @@ class WarcRecorderTest {
 
     @Test
     public void testSaveImport() throws URISyntaxException, IOException {
-        try (var recorder = new WarcRecorder(fileNameWarc, new BasicCookieStore())) {
+        try (var recorder = new WarcRecorder(fileNameWarc)) {
             recorder.writeReferenceCopy(new EdgeUrl("https://www.marginalia.nu/"),
+                    new DomainCookies(),
                     "text/html",
                     200,
                     "<?doctype html><html><body>test</body></html>".getBytes(),
@@ -138,35 +142,46 @@ class WarcRecorderTest {
 
     @Test
     public void testConvertToParquet() throws NoSuchAlgorithmException, IOException, URISyntaxException, InterruptedException {
-        client.fetch(httpClient, ClassicRequestBuilder
+        HttpGet request1 = new HttpGet("https://www.marginalia.nu/");
-                .get(new java.net.URI("https://www.marginalia.nu/"))
+        request1.addHeader("User-agent", "test.marginalia.nu");
-                .addHeader("User-agent", "test.marginalia.nu")
+        request1.addHeader("Accept-Encoding", "gzip");
-                .addHeader("Accept-Encoding", "gzip")
-                .build());
 
-        client.fetch(httpClient, ClassicRequestBuilder
+        client.fetch(httpClient, new DomainCookies(), request1);
-                .get(new java.net.URI("https://www.marginalia.nu/log/"))
-                .addHeader("User-agent", "test.marginalia.nu")
-                .addHeader("Accept-Encoding", "gzip")
-                .build());
 
-        client.fetch(httpClient, ClassicRequestBuilder
+        HttpGet request2 = new HttpGet("https://www.marginalia.nu/log/");
-                .get(new java.net.URI("https://www.marginalia.nu/sanic.png"))
+        request2.addHeader("User-agent", "test.marginalia.nu");
-                .addHeader("User-agent", "test.marginalia.nu")
+        request2.addHeader("Accept-Encoding", "gzip");
-                .addHeader("Accept-Encoding", "gzip")
-                .build());
 
-        CrawledDocumentParquetRecordFileWriter.convertWarc(
+        client.fetch(httpClient, new DomainCookies(), request2);
+
+        HttpGet request3 = new HttpGet("https://www.marginalia.nu/sanic.png");
+        request3.addHeader("User-agent", "test.marginalia.nu");
+        request3.addHeader("Accept-Encoding", "gzip");
+
+        client.fetch(httpClient, new DomainCookies(), request3);
+
+        HttpGet request4 = new HttpGet("https://downloads.marginalia.nu/test.pdf");
+        request4.addHeader("User-agent", "test.marginalia.nu");
+        request4.addHeader("Accept-Encoding", "gzip");
+
+        client.fetch(httpClient, new DomainCookies(), request4);
+
+        SlopCrawlDataRecord.convertWarc(
                 "www.marginalia.nu",
                 new UserAgent("test", "test"),
                 fileNameWarc,
-                fileNameParquet);
+                fileNameSlop);
 
-        var urls = CrawledDocumentParquetRecordFileReader.stream(fileNameParquet).map(doc -> doc.url).toList();
+        List<String> urls;
-        assertEquals(2, urls.size());
+        try (var stream = SerializableCrawlDataStream.openDataStream(fileNameSlop)) {
+            urls = stream.docsAsList().stream().map(doc -> doc.url.toString()).toList();
+        }
+
+        assertEquals(3, urls.size());
         assertEquals("https://www.marginalia.nu/", urls.get(0));
         assertEquals("https://www.marginalia.nu/log/", urls.get(1));
         // sanic.jpg gets filtered out for its bad mime type
+        assertEquals("https://downloads.marginalia.nu/test.pdf", urls.get(2));
 
     }
 

code/processes/crawling-process/test/nu/marginalia/crawling/HttpFetcherTest.java

@@ -1,6 +1,7 @@
 package nu.marginalia.crawling;
 
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.HttpFetcher;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
@@ -31,7 +32,7 @@ class HttpFetcherTest {
     void fetchUTF8() throws Exception {
         var fetcher = new HttpFetcherImpl("nu.marginalia.edge-crawler");
         try (var recorder = new WarcRecorder()) {
-            var result = fetcher.fetchContent(new EdgeUrl("https://www.marginalia.nu"), recorder, new CrawlDelayTimer(100), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
+            var result = fetcher.fetchContent(new EdgeUrl("https://www.marginalia.nu"), recorder, new DomainCookies(), new CrawlDelayTimer(100), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
             if (DocumentBodyExtractor.asString(result) instanceof DocumentBodyResult.Ok bodyOk) {
                 System.out.println(bodyOk.contentType());
             }
@@ -49,7 +50,7 @@ class HttpFetcherTest {
         var fetcher = new HttpFetcherImpl("nu.marginalia.edge-crawler");
 
         try (var recorder = new WarcRecorder()) {
-            var result = fetcher.fetchContent(new EdgeUrl("https://www.marginalia.nu/robots.txt"), recorder, new CrawlDelayTimer(100), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
+            var result = fetcher.fetchContent(new EdgeUrl("https://www.marginalia.nu/robots.txt"), recorder, new DomainCookies(), new CrawlDelayTimer(100), ContentTags.empty(), HttpFetcher.ProbeType.FULL);
             if (DocumentBodyExtractor.asString(result) instanceof DocumentBodyResult.Ok bodyOk) {
                 System.out.println(bodyOk.contentType());
             }

code/processes/crawling-process/test/nu/marginalia/crawling/retreival/CrawlerMockFetcherTest.java

@@ -3,10 +3,7 @@ package nu.marginalia.crawling.retreival;
 import crawlercommons.robots.SimpleRobotRules;
 import nu.marginalia.crawl.CrawlerMain;
 import nu.marginalia.crawl.DomainStateDb;
-import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.*;
-import nu.marginalia.crawl.fetcher.HttpFetcher;
-import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
-import nu.marginalia.crawl.fetcher.SitemapRetriever;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
 import nu.marginalia.crawl.retreival.CrawlDelayTimer;
 import nu.marginalia.crawl.retreival.CrawlerRetreiver;
@@ -137,7 +134,7 @@ public class CrawlerMockFetcherTest {
         }
 
         @Override
-        public HttpFetchResult fetchContent(EdgeUrl url, WarcRecorder recorder, CrawlDelayTimer timer, ContentTags tags, ProbeType probeType) {
+        public HttpFetchResult fetchContent(EdgeUrl url, WarcRecorder recorder, DomainCookies cookies, CrawlDelayTimer timer, ContentTags tags, ProbeType probeType) {
             logger.info("Fetching {}", url);
             if (mockData.containsKey(url)) {
                 byte[] bodyBytes = mockData.get(url).documentBodyBytes;

code/processes/crawling-process/test/nu/marginalia/crawling/retreival/CrawlerRetreiverTest.java

@@ -16,7 +16,6 @@ import nu.marginalia.model.crawldata.CrawledDocument;
 import nu.marginalia.model.crawldata.CrawledDomain;
 import nu.marginalia.model.crawldata.SerializableCrawlData;
 import nu.marginalia.slop.SlopCrawlDataRecord;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
 import org.jetbrains.annotations.NotNull;
 import org.junit.jupiter.api.*;
 import org.netpreserve.jwarc.*;
@@ -180,7 +179,7 @@ class CrawlerRetreiverTest {
                 new EdgeDomain("www.marginalia.nu"),
                 List.of(), 100);
         var resync = new CrawlerWarcResynchronizer(revisitCrawlFrontier,
-                new WarcRecorder(tempFileWarc2, new BasicCookieStore())
+                new WarcRecorder(tempFileWarc2)
         );
 
         // truncate the size of the file to simulate a crash
@@ -456,7 +455,7 @@ class CrawlerRetreiverTest {
                 List.of(), 100);
 
         var resync = new CrawlerWarcResynchronizer(revisitCrawlFrontier,
-                new WarcRecorder(tempFileWarc3, new BasicCookieStore())
+                new WarcRecorder(tempFileWarc3)
         );
 
         // truncate the size of the file to simulate a crash
@@ -507,7 +506,7 @@ class CrawlerRetreiverTest {
     }
 
     private void doCrawlWithReferenceStream(CrawlerMain.CrawlSpecRecord specs, CrawlDataReference reference) {
-        try (var recorder = new WarcRecorder(tempFileWarc2, new BasicCookieStore());
+        try (var recorder = new WarcRecorder(tempFileWarc2);
              var db = new DomainStateDb(tempFileDb)
         ) {
             new CrawlerRetreiver(httpFetcher, new DomainProber(d -> true), specs, db, recorder).crawlDomain(new DomainLinks(), reference);
@@ -519,7 +518,7 @@ class CrawlerRetreiverTest {
 
     @NotNull
     private DomainCrawlFrontier doCrawl(Path tempFileWarc1, CrawlerMain.CrawlSpecRecord specs) {
-        try (var recorder = new WarcRecorder(tempFileWarc1, new BasicCookieStore());
+        try (var recorder = new WarcRecorder(tempFileWarc1);
              var db = new DomainStateDb(tempFileDb)
         ) {
             var crawler = new CrawlerRetreiver(httpFetcher, new DomainProber(d -> true), specs, db, recorder);

code/processes/export-task-process/build.gradle

@@ -53,6 +53,8 @@ dependencies {
     implementation libs.commons.compress
     implementation libs.commons.codec
     implementation libs.jsoup
+    implementation libs.slop
+    implementation libs.jwarc
 
 
 

code/processes/export-task-process/java/nu/marginalia/extractor/SampleDataExporter.java

@@ -3,11 +3,15 @@ package nu.marginalia.extractor;
 import com.google.inject.Inject;
 import nu.marginalia.process.log.WorkLog;
 import nu.marginalia.process.log.WorkLogEntry;
+import nu.marginalia.slop.SlopCrawlDataRecord;
+import nu.marginalia.slop.SlopTablePacker;
 import nu.marginalia.storage.FileStorageService;
 import nu.marginalia.storage.model.FileStorage;
 import nu.marginalia.storage.model.FileStorageId;
 import org.apache.commons.compress.archivers.tar.TarArchiveOutputStream;
 import org.apache.commons.compress.utils.IOUtils;
+import org.apache.commons.io.FileUtils;
+import org.apache.commons.lang3.StringUtils;
 
 import java.io.IOException;
 import java.nio.file.Files;
@@ -27,7 +31,7 @@ public class SampleDataExporter {
     public SampleDataExporter(FileStorageService storageService) {
         this.storageService = storageService;
     }
-    public void export(FileStorageId crawlId, FileStorageId destId, int size, String name) throws SQLException, IOException {
+    public void export(FileStorageId crawlId, FileStorageId destId, int size, String ctFilter, String name) throws SQLException, IOException {
         FileStorage destStorage = storageService.getStorage(destId);
         Path inputDir = storageService.getStorage(crawlId).asPath();
 
@@ -54,6 +58,7 @@ public class SampleDataExporter {
 
         Path newCrawlerLogFile = Files.createTempFile(destStorage.asPath(), "crawler", ".log",
                 PosixFilePermissions.asFileAttribute(PosixFilePermissions.fromString("rw-r--r--")));
+
         try (var bw = Files.newBufferedWriter(newCrawlerLogFile, StandardOpenOption.CREATE, StandardOpenOption.TRUNCATE_EXISTING)) {
             for (var item : entriesAll) {
                 bw.write(item.id() + " " + item.ts() + " " + item.relPath() + " " + item.cnt() + "\n");
@@ -72,7 +77,22 @@ public class SampleDataExporter {
                 Path crawlDataPath = inputDir.resolve(item.relPath());
                 if (!Files.exists(crawlDataPath)) continue;
 
-                addFileToTar(stream, crawlDataPath, item.relPath());
+                if (StringUtils.isBlank(ctFilter)) {
+                    addFileToTar(stream, crawlDataPath, item.relPath());
+                }
+                else /* filter != null */ {
+                    boolean didFilterData = false;
+                    try {
+                        crawlDataPath = filterEntries(crawlDataPath, ctFilter);
+                        didFilterData = true;
+                        addFileToTar(stream, crawlDataPath, item.relPath());
+                    }
+                    finally {
+                        if (didFilterData) {
+                            Files.deleteIfExists(crawlDataPath);
+                        }
+                    }
+                }
             }
 
             addFileToTar(stream, newCrawlerLogFile, "crawler.log");
@@ -86,6 +106,46 @@ public class SampleDataExporter {
         Files.move(tmpTarFile, destStorage.asPath().resolve("crawl-data.tar"), StandardCopyOption.ATOMIC_MOVE, StandardCopyOption.REPLACE_EXISTING);
     }
 
+    /** Filters the entries in the crawl data file based on the content type.
+     * @param crawlDataPath The path to the crawl data file.
+     * @param contentTypeFilter The content type to filter by.
+     * @return The path to the filtered crawl data file, or null if an error occurred.
+     */
+    private Path filterEntries(Path crawlDataPath, String contentTypeFilter) throws IOException {
+        Path tempDir = crawlDataPath.resolveSibling(crawlDataPath.getFileName() + ".filtered");
+        Path tempFile = crawlDataPath.resolveSibling(crawlDataPath.getFileName() + ".filtered.slop.zip");
+
+        Files.createDirectory(tempDir);
+
+        try (var writer = new SlopCrawlDataRecord.Writer(tempDir);
+             var reader = new SlopCrawlDataRecord.FilteringReader(crawlDataPath) {
+                 @Override
+                 public boolean filter(String url, int status, String contentType) {
+                     if (contentTypeFilter.equals(contentType))
+                         return true;
+                     else if (contentType.startsWith("x-marginalia/"))
+                         // This is a metadata entry, typically domain or redirect information
+                         // let's keep those to not confuse the consumer of the data, which might
+                         // expect at least the domain summary
+                         return true;
+                     return false;
+                 }
+             }
+        ) {
+            while (reader.hasRemaining()) {
+                writer.write(reader.get());
+            }
+
+            SlopTablePacker.packToSlopZip(tempDir, tempFile);
+        }
+        finally {
+            FileUtils.deleteDirectory(tempDir.toFile());
+        }
+
+
+        return tempFile;
+    }
+
     private void addFileToTar(TarArchiveOutputStream outputStream, Path file, String fileName) throws IOException {
         var entry = outputStream.createArchiveEntry(file.toFile(), fileName);
         entry.setSize(Files.size(file));

code/processes/export-task-process/java/nu/marginalia/task/ExportTasksMain.java

@@ -92,7 +92,7 @@ public class ExportTasksMain extends ProcessMainClass {
                     termFrequencyExporter.export(request.crawlId, request.destId);
                     break;
                 case SAMPLE_DATA:
-                    sampleDataExporter.export(request.crawlId, request.destId, request.size, request.name);
+                    sampleDataExporter.export(request.crawlId, request.destId, request.size, request.ctFilter, request.name);
                     break;
                 case ADJACENCIES:
                     websiteAdjacenciesCalculator.export();

code/processes/process-mq-api/java/nu/marginalia/mqapi/tasks/ExportTaskRequest.java

@@ -16,6 +16,7 @@ public class ExportTaskRequest {
     public FileStorageId destId;
     public int size;
     public String name;
+    public String ctFilter;
 
     public ExportTaskRequest(Task task) {
         this.task = task;
@@ -42,12 +43,13 @@ public class ExportTaskRequest {
         return request;
     }
 
-    public static ExportTaskRequest sampleData(FileStorageId crawlId, FileStorageId destId, int size, String name) {
+    public static ExportTaskRequest sampleData(FileStorageId crawlId, FileStorageId destId, String ctFilter, int size, String name) {
         ExportTaskRequest request = new ExportTaskRequest(Task.SAMPLE_DATA);
         request.crawlId = crawlId;
         request.destId = destId;
         request.size = size;
         request.name = name;
+        request.ctFilter = ctFilter;
         return request;
     }
 

code/services-application/api-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 java {

code/services-application/dating-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-application/explorer-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-application/search-service-legacy/build.gradle

@@ -5,7 +5,7 @@ plugins {
     id 'application'
     id 'jvm-test-suite'
 
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-application/search-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
     id 'application'
     id 'jvm-test-suite'
     id 'gg.jte.gradle' version '3.1.15'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-application/search-service/resources/jte/part/head.jte

@@ -26,4 +26,10 @@
 
     <link rel="search" type="application/opensearchdescription+xml" href="/opensearch.xml" title="Marginalia">
 
-</head>
+</head>
+<noscript>
+    <h1>Users of text-based browsers</h1>
+    <p>Consider using the old interface at <a href="https://old-search.marginalia.nu/">https://old-search.marginalia.nu/</a>,
+    as it uses fewer modern CSS tricks, and should work better than the new UI.  It's functionally nearly identical, but just renders it using a different layout.</p>
+    <hr>
+</noscript>

code/services-application/search-service/resources/jte/part/warmup.jte

@@ -1,9 +1,16 @@
 This is a bit of a hack!
 
 This class exists to let tailwind we're using these classes even though they aren't visible in the code,
-as we sometimes generate classes from Java code!
+as we sometimes generate classes from Java code or javascript!
 
 <i class="text-blue-800 bg-blue-50 dark:text-blue-200 dark:bg-blue-950"></i>
 <i class="text-green-800 bg-green-50 dark:text-green-200 dark:bg-green-950"></i>
 <i class="text-purple-800 bg-purple-50 dark:text-purple-200 dark:bg-purple-950"></i>
 <i class="text-blue-950 bg-gray-100 dark:text-blue-50 dark:bg-gray-900"></i>
+<span class="hover:bg-gray-300 "></span>
+
+<label class="suggestion group block relative">
+    <input type="radio" name="suggestion" class="peer hidden" checked>
+    <div class="px-4 py-2 cursor-pointer dark:peer-checked:bg-gray-700 dark:hover:bg-gray-700 peer-checked:bg-gray-300 hover:bg-gray-300 w-full">
+    </div>
+</label>

code/services-application/search-service/resources/jte/serp/error.jte

@@ -26,7 +26,7 @@
 
     <!-- Main content -->
     <main class="flex-1 p-4 max-w-2xl space-y-4">
-        <div class="border dark:border-gray-600 rounded bg-white text-black dark:bg-gray-800 dark:text-white text-m p-4">
+        <div class="border border-gray-300 dark:border-gray-600 rounded bg-white text-black dark:bg-gray-800 dark:text-white text-m p-4">
             <div class="flex space-x-3  place-items-baseline">
                 <i class="fa fa-circle-exclamation text-red-800"></i>
                 <div class="grow">${model.errorTitle()}</div>

code/services-application/search-service/resources/jte/serp/part/footerHowto.jte

@@ -80,10 +80,6 @@
             <tr><td>rank&gt;50</td><td>The ranking of the website is at least 50 in a span of 1 - 255</td></tr>
             <tr><td>rank&lt;50</td><td>The ranking of the website is at most 50 in a span of 1 - 255</td></tr>
 
-            <tr><td>count&gt;10</td><td> The search term must appear in at least 10 results form the domain</td></tr>
-            <tr><td>count&lt;10</td><td> The search term must appear in at most 10 results from the domain</td></tr>
-
-
             <tr><td>format:html5</td><td>Filter documents using the HTML5 standard. This is typically modern websites.</td></tr>
             <tr><td>format:xhtml</td><td>Filter documents using the XHTML standard</td></tr>
             <tr><td>format:html123</td><td>Filter documents using the HTML standards 1, 2, and 3. This is typically very old websites. </td></tr>

code/services-application/search-service/resources/jte/serp/part/searchform.jte

@@ -7,13 +7,13 @@
 
 <form class="flex-1 max-w-2xl" action="/search">
     <div class="flex">
-        @if (query.isBlank())
+        @if (query != null && query.isBlank())
         <%-- Add autofocus if the query is blank --%>
             <input type="text"
                    class="shadow-inner flex-1 dark:bg-black dark:text-gray-100 bg-gray-50 border dark:border-gray-600 border-gray-300 text-gray-900 text-sm rounded-sm block w-full p-2.5"
                    value="${query}"
                    autofocus
-                   placeholder="Search..."
+                   placeholder="Search the web!"
                    autocomplete="off"
                    name="query"
                    id="searchInput" />
@@ -21,13 +21,13 @@
             <input type="text"
                    class="shadow-inner flex-1 dark:bg-black dark:text-gray-100 bg-gray-50 border dark:border-gray-600 border-gray-300 text-gray-900 text-sm rounded-sm block w-full p-2.5"
                    value="${query}"
-                   placeholder="Search..."
+                   placeholder="Search the web!"
                    autocomplete="off"
                    name="query"
                    id="searchInput" />
         @endif
 
-        <div id="searchSuggestions" class="text-sm absolute top-2 mt-10 w-96 bg-white dark:bg-black border dark:border-gray-600 border-gray-200 rounded-lg shadow-lg hidden"></div>
+        <div aria-hidden="true" id="searchSuggestions" class="text-sm absolute top-3 mt-10 w-96 bg-white dark:bg-black border dark:border-gray-600 border-gray-300 rounded-lg shadow-lg hidden"></div>
 
         <button class="px-4 py-2 bg-margeblue text-white ml-2 rounded whitespace-nowrap active:text-slate-200">
             <i class="fas fa-search text-sm sm:mr-3"></i>

code/services-application/search-service/resources/static/js/typeahead.js

@@ -43,13 +43,13 @@ function displaySuggestions(suggestions) {
     }
 
     suggestionsContainer.innerHTML = suggestions.map((suggestion, index) => `
-                <div 
+        <label class="suggestion group block relative">
-                    class="suggestion px-4 py-2 cursor-pointer hover:bg-gray-100 ${index === selectedIndex ? 'bg-blue-50' : ''}"
+            <input type="radio" name="suggestion" class="peer hidden" ${index === selectedIndex ? 'checked' : ''}>
-                    data-index="${index}"
+            <div class="px-4 py-2 cursor-pointer dark:peer-checked:bg-gray-700 dark:hover:bg-gray-700 peer-checked:bg-gray-300 hover:bg-gray-300 w-full" data-index="${index}">
-                >
+                ${suggestion}
-                    ${suggestion}
+            </div>
-                </div>
+        </label>
-            `).join('');
+    `).join('');
 
     suggestionsContainer.classList.remove('hidden');
 

code/services-application/status-service/build.gradle

@@ -2,7 +2,7 @@ plugins {
     id 'java'
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 java {

code/services-core/assistant-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-core/assistant-service/java/nu/marginalia/assistant/AssistantModule.java

@@ -10,7 +10,8 @@ import static com.google.inject.name.Names.named;
 
 public class AssistantModule extends AbstractModule {
     public void configure() {
-        bind(Path.class).annotatedWith(named("suggestions-file")).toInstance(WmsaHome.getHomePath().resolve("data/suggestions.txt"));
+        bind(Path.class).annotatedWith(named("suggestions-file1")).toInstance(WmsaHome.getHomePath().resolve("data/suggestions2.txt.gz"));
+        bind(Path.class).annotatedWith(named("suggestions-file2")).toInstance(WmsaHome.getHomePath().resolve("data/suggestions3.txt.gz"));
 
         bind(LanguageModels.class).toInstance(WmsaHome.getLanguageModels());
     }

code/services-core/assistant-service/java/nu/marginalia/assistant/suggest/PrefixSearchStructure.java

@@ -0,0 +1,465 @@
+package nu.marginalia.assistant.suggest;
+
+import gnu.trove.list.array.TIntArrayList;
+import org.jetbrains.annotations.NotNull;
+
+import java.util.*;
+
+/** Unhinged data structure for fast prefix searching.
+ */
+public class PrefixSearchStructure {
+    // Core data structures
+    private final HashMap<String, TIntArrayList> prefixIndex;     // Short prefix index (up to 8 chars)
+    private final HashMap<String, TIntArrayList> longPrefixIndex; // Long prefix index (9-16 chars)
+    private final ArrayList<String> words;                        // All words by ID
+    private final TIntArrayList wordScores;                       // Scores for all words
+
+    // Configuration
+    private static final int SHORT_PREFIX_LENGTH = 8;
+    private static final int MAX_INDEXED_PREFIX_LENGTH = 16;
+
+    public int size() {
+        return words.size();
+    }
+
+    // For sorting efficiency
+    private static class WordScorePair {
+        final String word;
+        final int score;
+
+        WordScorePair(String word, int score) {
+            this.word = word;
+            this.score = score;
+        }
+    }
+
+    /**
+     * Creates a new PrefixTrie for typeahead search.
+     */
+    public PrefixSearchStructure() {
+        prefixIndex = new HashMap<>(1024);
+        longPrefixIndex = new HashMap<>(1024);
+        words = new ArrayList<>(1024);
+        wordScores = new TIntArrayList(1024);
+    }
+
+    /**
+     * Adds a prefix to the index.
+     */
+    private void indexPrefix(String word, int wordId) {
+        // Index short prefixes
+        for (int i = 1; i <= Math.min(word.length(), SHORT_PREFIX_LENGTH); i++) {
+            String prefix = word.substring(0, i);
+            TIntArrayList wordIds = prefixIndex.computeIfAbsent(
+                    prefix, k -> new TIntArrayList(16));
+            wordIds.add(wordId);
+        }
+
+        // Index longer prefixes
+        for (int i = SHORT_PREFIX_LENGTH + 1; i <= Math.min(word.length(), MAX_INDEXED_PREFIX_LENGTH); i++) {
+            String prefix = word.substring(0, i);
+            TIntArrayList wordIds = longPrefixIndex.computeIfAbsent(
+                    prefix, k -> new TIntArrayList(8));
+            wordIds.add(wordId);
+        }
+
+        // If the word contains spaces, also index by each term for multi-word queries
+        if (word.contains(" ")) {
+            String[] terms = word.split("\\s+");
+            for (String term : terms) {
+                if (term.length() >= 2) {
+                    for (int i = 1; i <= Math.min(term.length(), SHORT_PREFIX_LENGTH); i++) {
+                        String termPrefix = "t:" + term.substring(0, i);
+                        TIntArrayList wordIds = prefixIndex.computeIfAbsent(
+                                termPrefix, k -> new TIntArrayList(16));
+                        wordIds.add(wordId);
+                    }
+                }
+            }
+        }
+    }
+
+    /**
+     * Inserts a word with its associated score.
+     */
+    public void insert(String word, int score) {
+        if (word == null || word.isEmpty()) {
+            return;
+        }
+
+        // Add to the word list and index
+        int wordId = words.size();
+        words.add(word);
+        wordScores.add(score);
+        indexPrefix(word, wordId);
+    }
+
+    /**
+     * Returns the top k completions for a given prefix.
+     */
+    public List<ScoredSuggestion> getTopCompletions(String prefix, int k) {
+        if (prefix == null || prefix.isEmpty()) {
+            // Return top k words by score
+            return getTopKWords(k);
+        }
+
+        // Check if this is a term search (t:) - for searching within multi-word items
+        boolean isTermSearch = false;
+        if (prefix.startsWith("t:") && prefix.length() > 2) {
+            isTermSearch = true;
+            prefix = prefix.substring(2);
+        }
+
+        // 1. Fast path for short prefixes
+        if (prefix.length() <= SHORT_PREFIX_LENGTH) {
+            String lookupPrefix = isTermSearch ? "t:" + prefix : prefix;
+            TIntArrayList wordIds = prefixIndex.get(lookupPrefix);
+            if (wordIds != null) {
+                return getTopKFromWordIds(wordIds, k);
+            }
+        }
+
+        // 2. Fast path for long prefixes (truncate to MAX_INDEXED_PREFIX_LENGTH)
+        if (prefix.length() > SHORT_PREFIX_LENGTH) {
+            // Try exact match in longPrefixIndex first
+            if (prefix.length() <= MAX_INDEXED_PREFIX_LENGTH) {
+                TIntArrayList wordIds = longPrefixIndex.get(prefix);
+                if (wordIds != null) {
+                    return getTopKFromWordIds(wordIds, k);
+                }
+            }
+
+            // If prefix is longer than MAX_INDEXED_PREFIX_LENGTH, truncate and filter
+            if (prefix.length() > MAX_INDEXED_PREFIX_LENGTH) {
+                String truncatedPrefix = prefix.substring(0, MAX_INDEXED_PREFIX_LENGTH);
+                TIntArrayList candidateIds = longPrefixIndex.get(truncatedPrefix);
+                if (candidateIds != null) {
+                    // Filter candidates by the full prefix
+                    return getFilteredTopKFromWordIds(candidateIds, prefix, k);
+                }
+            }
+        }
+
+        // 3. Optimized fallback for long prefixes - use prefix tree for segments
+        List<ScoredSuggestion> results = new ArrayList<>();
+
+        // Handle multi-segment queries by finding candidates from first 8 chars
+        if (prefix.length() > SHORT_PREFIX_LENGTH) {
+            String shortPrefix = prefix.substring(0, Math.min(prefix.length(), SHORT_PREFIX_LENGTH));
+            TIntArrayList candidates = prefixIndex.get(shortPrefix);
+
+            if (candidates != null) {
+                return getFilteredTopKFromWordIds(candidates, prefix, k);
+            }
+        }
+
+        // 4. Last resort - optimized binary search in sorted segments
+        return findByBinarySearchPrefix(prefix, k);
+    }
+
+    /**
+     * Helper to get the top k words by score.
+     */
+    private List<ScoredSuggestion> getTopKWords(int k) {
+        // Create pairs of (score, wordId)
+        int[][] pairs = new int[words.size()][2];
+        for (int i = 0; i < words.size(); i++) {
+            pairs[i][0] = wordScores.get(i);
+            pairs[i][1] = i;
+        }
+
+        // Sort by score (descending)
+        Arrays.sort(pairs, (a, b) -> Integer.compare(b[0], a[0]));
+
+        // Take top k
+        List<ScoredSuggestion> results = new ArrayList<>();
+        for (int i = 0; i < Math.min(k, pairs.length); i++) {
+            String word = words.get(pairs[i][1]);
+            int score = pairs[i][0];
+            results.add(new ScoredSuggestion(word, score));
+        }
+
+        return results;
+    }
+
+    /**
+     * Helper to get the top k words from a list of word IDs.
+     */
+    private List<ScoredSuggestion> getTopKFromWordIds(TIntArrayList wordIds, int k) {
+        if (wordIds == null || wordIds.isEmpty()) {
+            return Collections.emptyList();
+        }
+
+        // For small lists, avoid sorting
+        if (wordIds.size() <= k) {
+            List<ScoredSuggestion> results = new ArrayList<>(wordIds.size());
+            int[] ids = wordIds.toArray();
+            for (int wordId : ids) {
+                if (wordId >= 0 && wordId < words.size()) {
+                    results.add(new ScoredSuggestion(words.get(wordId), wordScores.get(wordId)));
+                }
+            }
+            results.sort((a, b) -> Integer.compare(b.getScore(), a.getScore()));
+            return results;
+        }
+
+        // For larger lists, use an array-based approach for better performance
+        // Find top k without full sorting
+        int[] topScores = new int[k];
+        int[] topWordIds = new int[k];
+        int[] ids = wordIds.toArray();
+
+        // Initialize with first k elements
+        int filledCount = Math.min(k, ids.length);
+        for (int i = 0; i < filledCount; i++) {
+            int wordId = ids[i];
+            if (wordId >= 0 && wordId < words.size()) {
+                topWordIds[i] = wordId;
+                topScores[i] = wordScores.get(wordId);
+            }
+        }
+
+        // Sort initial elements
+        for (int i = 0; i < filledCount; i++) {
+            for (int j = i + 1; j < filledCount; j++) {
+                if (topScores[j] > topScores[i]) {
+                    // Swap scores
+                    int tempScore = topScores[i];
+                    topScores[i] = topScores[j];
+                    topScores[j] = tempScore;
+
+                    // Swap word IDs
+                    int tempId = topWordIds[i];
+                    topWordIds[i] = topWordIds[j];
+                    topWordIds[j] = tempId;
+                }
+            }
+        }
+
+        // Process remaining elements
+        int minScore = filledCount > 0 ? topScores[filledCount - 1] : Integer.MIN_VALUE;
+
+        for (int i = k; i < ids.length; i++) {
+            int wordId = ids[i];
+            if (wordId >= 0 && wordId < words.size()) {
+                int score = wordScores.get(wordId);
+
+                if (score > minScore) {
+                    // Replace the lowest element
+                    topScores[filledCount - 1] = score;
+                    topWordIds[filledCount - 1] = wordId;
+
+                    // Bubble up the new element
+                    for (int j = filledCount - 1; j > 0; j--) {
+                        if (topScores[j] > topScores[j - 1]) {
+                            // Swap scores
+                            int tempScore = topScores[j];
+                            topScores[j] = topScores[j - 1];
+                            topScores[j - 1] = tempScore;
+
+                            // Swap word IDs
+                            int tempId = topWordIds[j];
+                            topWordIds[j] = topWordIds[j - 1];
+                            topWordIds[j - 1] = tempId;
+                        } else {
+                            break;
+                        }
+                    }
+
+                    // Update min score
+                    minScore = topScores[filledCount - 1];
+                }
+            }
+        }
+
+        // Create result list
+        List<ScoredSuggestion> results = new ArrayList<>(filledCount);
+        for (int i = 0; i < filledCount; i++) {
+            results.add(new ScoredSuggestion(words.get(topWordIds[i]), topScores[i]));
+        }
+
+        return results;
+    }
+
+    /**
+     * Use binary search on sorted word segments to efficiently find matches.
+     */
+    private List<ScoredSuggestion> findByBinarySearchPrefix(String prefix, int k) {
+        // If we have a lot of words, use an optimized segment approach
+        if (words.size() > 1000) {
+            // Divide words into segments for better locality
+            int segmentSize = 1000;
+            int numSegments = (words.size() + segmentSize - 1) / segmentSize;
+
+            // Find matches using binary search within each segment
+            List<WordScorePair> allMatches = new ArrayList<>();
+            for (int segment = 0; segment < numSegments; segment++) {
+                int start = segment * segmentSize;
+                int end = Math.min(start + segmentSize, words.size());
+
+                // Binary search for first potential match
+                int pos = Collections.binarySearch(
+                        words.subList(start, end),
+                        prefix,
+                        (a, b) -> a.compareTo(b)
+                );
+
+                if (pos < 0) {
+                    pos = -pos - 1;
+                }
+
+                // Collect all matches
+                for (int i = start + pos; i < end && i < words.size(); i++) {
+                    String word = words.get(i);
+                    if (word.startsWith(prefix)) {
+                        allMatches.add(new WordScorePair(word, wordScores.get(i)));
+                    } else if (word.compareTo(prefix) > 0) {
+                        break; // Past potential matches
+                    }
+                }
+            }
+
+            // Sort by score and take top k
+            allMatches.sort((a, b) -> Integer.compare(b.score, a.score));
+            List<ScoredSuggestion> results = new ArrayList<>(Math.min(k, allMatches.size()));
+            for (int i = 0; i < Math.min(k, allMatches.size()); i++) {
+                WordScorePair pair = allMatches.get(i);
+                results.add(new ScoredSuggestion(pair.word, pair.score));
+            }
+            return results;
+        }
+
+        // Fallback for small dictionaries - linear scan but optimized
+        return simpleSearchFallback(prefix, k);
+    }
+
+    /**
+     * Optimized linear scan - only used for small dictionaries.
+     */
+    private List<ScoredSuggestion> simpleSearchFallback(String prefix, int k) {
+        // Use primitive arrays for better cache locality
+        int[] matchScores = new int[Math.min(words.size(), 100)]; // Assume we won't find more than 100 matches
+        String[] matchWords = new String[matchScores.length];
+        int matchCount = 0;
+
+        for (int i = 0; i < words.size() && matchCount < matchScores.length; i++) {
+            String word = words.get(i);
+            if (word.startsWith(prefix)) {
+                matchWords[matchCount] = word;
+                matchScores[matchCount] = wordScores.get(i);
+                matchCount++;
+            }
+        }
+
+        // Sort matches by score (in-place for small arrays)
+        for (int i = 0; i < matchCount; i++) {
+            for (int j = i + 1; j < matchCount; j++) {
+                if (matchScores[j] > matchScores[i]) {
+                    // Swap scores
+                    int tempScore = matchScores[i];
+                    matchScores[i] = matchScores[j];
+                    matchScores[j] = tempScore;
+
+                    // Swap words
+                    String tempWord = matchWords[i];
+                    matchWords[i] = matchWords[j];
+                    matchWords[j] = tempWord;
+                }
+            }
+        }
+
+        // Create results
+        List<ScoredSuggestion> results = new ArrayList<>(Math.min(k, matchCount));
+        for (int i = 0; i < Math.min(k, matchCount); i++) {
+            results.add(new ScoredSuggestion(matchWords[i], matchScores[i]));
+        }
+
+        return results;
+    }
+
+    /**
+     * Get top k words from candidate IDs, filtering by the full prefix.
+     */
+    private List<ScoredSuggestion> getFilteredTopKFromWordIds(TIntArrayList wordIds, String fullPrefix, int k) {
+        if (wordIds == null || wordIds.isEmpty()) {
+            return Collections.emptyList();
+        }
+
+        // Make primitive arrays for better performance
+        String[] matchWords = new String[Math.min(wordIds.size(), 1000)];
+        int[] matchScores = new int[matchWords.length];
+        int matchCount = 0;
+
+        int[] ids = wordIds.toArray();
+        for (int i = 0; i < ids.length && matchCount < matchWords.length; i++) {
+            int wordId = ids[i];
+            if (wordId >= 0 && wordId < words.size()) {
+                String word = words.get(wordId);
+                if (word.startsWith(fullPrefix)) {
+                    matchWords[matchCount] = word;
+                    matchScores[matchCount] = wordScores.get(wordId);
+                    matchCount++;
+                }
+            }
+        }
+
+        // Sort by score (efficient insertion sort for small k)
+        for (int i = 0; i < Math.min(matchCount, k); i++) {
+            int maxPos = i;
+            for (int j = i + 1; j < matchCount; j++) {
+                if (matchScores[j] > matchScores[maxPos]) {
+                    maxPos = j;
+                }
+            }
+            if (maxPos != i) {
+                // Swap
+                int tempScore = matchScores[i];
+                matchScores[i] = matchScores[maxPos];
+                matchScores[maxPos] = tempScore;
+
+                String tempWord = matchWords[i];
+                matchWords[i] = matchWords[maxPos];
+                matchWords[maxPos] = tempWord;
+            }
+        }
+
+        // Create result list (only up to k elements)
+        List<ScoredSuggestion> results = new ArrayList<>(Math.min(k, matchCount));
+        for (int i = 0; i < Math.min(k, matchCount); i++) {
+            results.add(new ScoredSuggestion(matchWords[i], matchScores[i]));
+        }
+
+        return results;
+    }
+
+    /**
+     * Class representing a suggested completion.
+     */
+    public static class ScoredSuggestion implements Comparable<ScoredSuggestion> {
+        private final String word;
+        private final int score;
+
+        public ScoredSuggestion(String word, int score) {
+            this.word = word;
+            this.score = score;
+        }
+
+        public String getWord() {
+            return word;
+        }
+
+        public int getScore() {
+            return score;
+        }
+
+        @Override
+        public String toString() {
+            return word + " (" + score + ")";
+        }
+
+        @Override
+        public int compareTo(@NotNull PrefixSearchStructure.ScoredSuggestion o) {
+            return Integer.compare(this.score, o.score);
+        }
+    }
+}

code/services-core/assistant-service/java/nu/marginalia/assistant/suggest/Suggestions.java

@@ -2,74 +2,89 @@ package nu.marginalia.assistant.suggest;
 
 import com.google.inject.Inject;
 import com.google.inject.name.Named;
-import nu.marginalia.functions.math.dict.SpellChecker;
-import nu.marginalia.term_frequency_dict.TermFrequencyDict;
-import nu.marginalia.model.crawl.HtmlFeature;
-import org.apache.commons.collections4.trie.PatriciaTrie;
 import org.apache.commons.lang3.StringUtils;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import java.io.BufferedInputStream;
 import java.io.IOException;
 import java.nio.file.Files;
 import java.nio.file.Path;
+import java.nio.file.StandardOpenOption;
 import java.util.*;
-import java.util.function.Supplier;
+import java.util.zip.GZIPInputStream;
-import java.util.regex.Pattern;
-import java.util.stream.Collectors;
-import java.util.stream.Stream;
 
 public class Suggestions {
-    private PatriciaTrie<String> suggestionsTrie = null;
+    List<PrefixSearchStructure> searchStructures = new ArrayList<>();
-    private TermFrequencyDict termFrequencyDict = null;
+
-    private volatile boolean ready = false;
+    private volatile boolean ready = false;
-    private final SpellChecker spellChecker;
 
-    private static final Pattern suggestionPattern = Pattern.compile("^[a-zA-Z0-9]+( [a-zA-Z0-9]+)*$");
     private static final Logger logger = LoggerFactory.getLogger(Suggestions.class);
 
     private static final int MIN_SUGGEST_LENGTH = 3;
     @Inject
-    public Suggestions(@Named("suggestions-file") Path suggestionsFile,
+    public Suggestions(@Named("suggestions-file1") Path suggestionsFile1,
-                       SpellChecker spellChecker,
+                       @Named("suggestions-file2") Path suggestionsFile2
-                       TermFrequencyDict dict
                        ) {
-        this.spellChecker = spellChecker;
 
         Thread.ofPlatform().start(() -> {
-            suggestionsTrie = loadSuggestions(suggestionsFile);
+            searchStructures.add(loadSuggestions(suggestionsFile1));
-            termFrequencyDict = dict;
+            searchStructures.add(loadSuggestions(suggestionsFile2));
             ready = true;
-            logger.info("Loaded {} suggestions", suggestionsTrie.size());
+            logger.info("Loaded suggestions");
         });
     }
 
-    private static PatriciaTrie<String> loadSuggestions(Path file) {
+    private static PrefixSearchStructure loadSuggestions(Path file) {
+        PrefixSearchStructure ret = new PrefixSearchStructure();
+
         if (!Files.exists(file)) {
             logger.error("Suggestions file {} absent, loading empty suggestions db", file);
-            return new PatriciaTrie<>();
+            return ret;
         }
-        try (var lines = Files.lines(file)) {
-            var ret = new PatriciaTrie<String>();
 
-            lines.filter(suggestionPattern.asPredicate())
+        try (var scanner = new Scanner(new GZIPInputStream(new BufferedInputStream(Files.newInputStream(file, StandardOpenOption.READ))))) {
-                    .filter(line -> line.length()<32)
+            while (scanner.hasNextLine()) {
-                    .map(String::toLowerCase)
+                String line = scanner.nextLine().trim();
-                    .forEach(w -> ret.put(w, w));
+                String[] parts = StringUtils.split(line, " ,", 2);
+                if (parts.length != 2) {
+                    logger.warn("Invalid suggestion line: {}", line);
+                    continue;
+                }
+                int cnt = Integer.parseInt(parts[0]);
+                if (cnt > 1) {
+                    String word = parts[1];
 
-            // Add special keywords to the suggestions
+                    // Remove quotes and trailing periods if this is a CSV
-            for (var feature : HtmlFeature.values()) {
+                    if (word.startsWith("\"") && word.endsWith("\"")) {
-                String keyword = feature.getKeyword();
+                        word = word.substring(1, word.length() - 1);
+                    }
 
-                ret.put(keyword, keyword);
+                    // Remove trailing periods
-                ret.put("-" + keyword, "-" + keyword);
+                    while (word.endsWith(".")) {
+                        word = word.substring(0, word.length() - 1);
+                    }
+
+                    // Remove junk items we may have gotten from link extraction
+                    if (word.startsWith("click here"))
+                        continue;
+                    if (word.contains("new window"))
+                        continue;
+                    if (word.contains("click to"))
+                        continue;
+                    if (word.startsWith("share "))
+                        continue;
+
+                    if (word.length() > 3) {
+                        ret.insert(word, cnt);
+                    }
+                }
             }
-
             return ret;
         }
         catch (IOException ex) {
             logger.error("Failed to load suggestions file", ex);
-            return new PatriciaTrie<>();
+            return new PrefixSearchStructure();
         }
     }
 
@@ -83,96 +98,36 @@ public class Suggestions {
 
         searchWord = StringUtils.stripStart(searchWord.toLowerCase(), " ");
 
-        return Stream.of(
+        return getSuggestionsForKeyword(count, searchWord);
-                    new SuggestionStream("", getSuggestionsForKeyword(count, searchWord)),
-                    suggestionsForLastWord(count, searchWord),
-                    spellCheckStream(searchWord)
-                )
-                .flatMap(SuggestionsStreamable::stream)
-                .limit(count)
-                .collect(Collectors.toList());
     }
 
-    private SuggestionsStreamable suggestionsForLastWord(int count, String searchWord) {
+    public List<String> getSuggestionsForKeyword(int count, String prefix) {
-        int sp = searchWord.lastIndexOf(' ');
-
-        if (sp < 0) {
-            return Stream::empty;
-        }
-
-        String prefixString = searchWord.substring(0, sp+1);
-        String suggestString = searchWord.substring(sp+1);
-
-        return new SuggestionStream(prefixString, getSuggestionsForKeyword(count, suggestString));
-
-    }
-
-    private SuggestionsStreamable spellCheckStream(String word) {
-        int start = word.lastIndexOf(' ');
-        String prefix;
-        String corrWord;
-
-        if (start < 0) {
-            corrWord = word;
-            prefix = "";
-        }
-        else {
-            prefix = word.substring(0, start + 1);
-            corrWord = word.substring(start + 1);
-        }
-
-        if (corrWord.length() >= MIN_SUGGEST_LENGTH) {
-            Supplier<Stream<String>> suggestionsLazyEval = () -> spellChecker.correct(corrWord).stream();
-            return new SuggestionStream(prefix, Stream.of(suggestionsLazyEval).flatMap(Supplier::get));
-        }
-        else {
-            return Stream::empty;
-        }
-    }
-
-
-    public Stream<String> getSuggestionsForKeyword(int count, String prefix) {
         if (!ready)
-            return Stream.empty();
+            return List.of();
 
         if (prefix.length() < MIN_SUGGEST_LENGTH) {
-            return Stream.empty();
+            return List.of();
         }
 
-        var start = suggestionsTrie.select(prefix);
+        List<PrefixSearchStructure.ScoredSuggestion> resultsAll = new ArrayList<>();
 
-        if (start == null) {
+        for (var searchStructure : searchStructures) {
-            return Stream.empty();
+            resultsAll.addAll(searchStructure.getTopCompletions(prefix, count));
+        }
+        resultsAll.sort(Comparator.reverseOrder());
+        List<String> ret = new ArrayList<>(count);
+
+        Set<String> seen = new HashSet<>();
+        for (var result : resultsAll) {
+            if (seen.add(result.getWord())) {
+                ret.add(result.getWord());
+            }
+            if (ret.size() >= count) {
+                break;
+            }
         }
 
-        if (!start.getKey().startsWith(prefix)) {
+        return ret;
-            return Stream.empty();
-        }
-
-        SuggestionsValueCalculator sv = new SuggestionsValueCalculator();
-
-        return Stream.iterate(start.getKey(), Objects::nonNull, suggestionsTrie::nextKey)
-                .takeWhile(s -> s.startsWith(prefix))
-                .limit(256)
-                .sorted(Comparator.comparing(sv::get).thenComparing(String::length).thenComparing(Comparator.naturalOrder()))
-                .limit(count);
     }
 
-    private record SuggestionStream(String prefix, Stream<String> suggestionStream) implements SuggestionsStreamable {
-        public Stream<String> stream() {
-            return suggestionStream.map(s -> prefix + s);
-        }
-    }
-
-    interface SuggestionsStreamable { Stream<String> stream(); }
-
-    private class SuggestionsValueCalculator {
-
-        private final Map<String, Long> hashCache = new HashMap<>(512);
-
-        public int get(String s) {
-            long hash = hashCache.computeIfAbsent(s, TermFrequencyDict::getStringHash);
-            return -termFrequencyDict.getTermFreqHash(hash);
-        }
-    }
 }

code/services-core/control-service/build.gradle

@@ -2,7 +2,7 @@ plugins {
     id 'java'
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 java {

code/services-core/control-service/java/nu/marginalia/control/ControlMain.java

@@ -59,9 +59,14 @@ public class ControlMain extends MainClass {
             download(adblockFile, new URI("https://downloads.marginalia.nu/data/adblock.txt"));
         }
 
-        Path suggestionsFile = dataPath.resolve("suggestions.txt");
+        Path suggestionsFile = dataPath.resolve("suggestions2.txt.gz");
         if (!Files.exists(suggestionsFile)) {
-            downloadGzipped(suggestionsFile, new URI("https://downloads.marginalia.nu/data/suggestions.txt.gz"));
+            download(suggestionsFile, new URI("https://downloads.marginalia.nu/data/suggestions2.txt.gz"));
+        }
+
+        Path altSuggestionsFile = dataPath.resolve("suggestions3.txt.gz");
+        if (!Files.exists(altSuggestionsFile)) {
+            download(altSuggestionsFile, new URI("https://downloads.marginalia.nu/data/suggestions3.txt.gz"));
         }
 
         Path asnRawData = dataPath.resolve("asn-data-raw-table");

code/services-core/control-service/java/nu/marginalia/control/node/svc/ControlNodeActionsService.java

@@ -321,9 +321,10 @@ public class ControlNodeActionsService {
     private Object exportSampleData(Request req, Response rsp) {
         FileStorageId source = parseSourceFileStorageId(req.queryParams("source"));
         int size = Integer.parseInt(req.queryParams("size"));
+        String ctFilter = req.queryParams("ctFilter");
         String name = req.queryParams("name");
 
-        exportClient.exportSampleData(Integer.parseInt(req.params("id")), source, size, name);
+        exportClient.exportSampleData(Integer.parseInt(req.params("id")), source, size, ctFilter, name);
 
         return "";
     }

code/services-core/control-service/resources/templates/control/node/actions/partial-download-sample-data.hdb

@@ -24,25 +24,25 @@ This is a sample of real crawl data.  It is intended for demo, testing and devel
     <tr>
         <td><input id="sample-s" value="sample-s" name="sample" class="form-check-input" type="radio"></td>
         <td><label for="sample-s">Small</label></td>
-        <td>1000 Domains. About 2 GB. </td>
+        <td>1000 Domains. About 1 GB. </td>
     </tr>
 
     <tr>
         <td><input id="sample-m" value="sample-m" name="sample" class="form-check-input" type="radio"></td>
         <td><label for="sample-m">Medium</label></td>
-        <td>2000 Domains. About 6 GB. Recommended.</td>
+        <td>2000 Domains. About 2 GB. Recommended.</td>
     </tr>
 
     <tr>
         <td><input id="sample-l" value="sample-l" name="sample" class="form-check-input" type="radio"></td>
         <td><label for="sample-l">Large</label></td>
-        <td>5000 Domains.  About 20 GB.</td>
+        <td>5000 Domains.  About 7 GB.</td>
     </tr>
 
     <tr>
         <td><input id="sample-xl" value="sample-xl" name="sample" class="form-check-input" type="radio"></td>
         <td><label for="sample-xl">Huge</label></td>
-        <td>50,000 Domains.  Around 180 GB.  Primarily intended for pre-production like testing environments.
+        <td>50,000 Domains.  Around 80 GB.  Primarily intended for pre-production like testing environments.
             Expect hours of processing time. </td>
     </tr>
 </table>

code/services-core/control-service/resources/templates/control/node/actions/partial-export-sample-data.hdb

@@ -35,6 +35,11 @@
         <div><input type="text" name="size" id="size" pattern="\d+" /></div>
         <small class="text-muted">How many domains to include in the sample set</small>
     </div>
+    <div class="mb-3">
+        <label for="ctFilter">Content Type Filter</label>
+        <div><input type="text" name="ctFilter" id="ctFilter" /></div>
+        <small class="text-muted">If set, includes only documents with the specified content type value</small>
+    </div>
     <div class="mb-3">
         <label for="name">Name</label>
         <div><input type="text" name="name" id="name" /></div>

code/services-core/executor-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-core/index-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/services-core/query-service/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 application {

code/tools/integration-test/test/nu/marginalia/IntegrationTest.java

@@ -10,6 +10,7 @@ import nu.marginalia.api.searchquery.model.results.PrototypeRankingParameters;
 import nu.marginalia.converting.processor.DomainProcessor;
 import nu.marginalia.converting.writer.ConverterBatchWriter;
 import nu.marginalia.crawl.fetcher.ContentTags;
+import nu.marginalia.crawl.fetcher.DomainCookies;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
 import nu.marginalia.functions.searchquery.QueryFactory;
@@ -43,7 +44,6 @@ import nu.marginalia.process.control.FakeProcessHeartbeat;
 import nu.marginalia.storage.FileStorageService;
 import nu.marginalia.test.IntegrationTestModule;
 import nu.marginalia.test.TestUtil;
-import org.apache.hc.client5.http.cookie.BasicCookieStore;
 import org.junit.jupiter.api.AfterEach;
 import org.junit.jupiter.api.BeforeEach;
 import org.junit.jupiter.api.Test;
@@ -121,11 +121,12 @@ public class IntegrationTest {
     public void run() throws Exception {
 
         /** CREATE WARC */
-        try (WarcRecorder warcRecorder = new WarcRecorder(warcData, new BasicCookieStore())) {
+        try (WarcRecorder warcRecorder = new WarcRecorder(warcData)) {
             warcRecorder.writeWarcinfoHeader("127.0.0.1", new EdgeDomain("www.example.com"),
                     new HttpFetcherImpl.DomainProbeResult.Ok(new EdgeUrl("https://www.example.com/")));
 
             warcRecorder.writeReferenceCopy(new EdgeUrl("https://www.example.com/"),
+                    new DomainCookies(),
                     "text/html", 200,
                     """
                             <html>

code/tools/screenshot-capture-tool/build.gradle

@@ -3,7 +3,7 @@ plugins {
 
     id 'application'
     id 'jvm-test-suite'
-    id 'com.google.cloud.tools.jib' version '3.4.4'
+    id 'com.google.cloud.tools.jib' version '3.4.5'
 }
 
 java {

deploy.txt

@@ -1,4 +1,6 @@
-## This is a token file for automatic deployment
+## This is a token file for triggering automatic deployment when no commit is made.
 
 2025-01-08:  Deploy executor.
-2025-01-07:  Deploy executor.
+2025-01-07:  Deploy executor.
+2025-04-24:  Deploy executor.
+2025-04-24:  Deploy assistant.

gradle/wrapper/gradle-wrapper.properties

@@ -1,5 +1,5 @@
 distributionBase=GRADLE_USER_HOME
 distributionPath=wrapper/dists
-distributionUrl=https\://services.gradle.org/distributions/gradle-8.10-bin.zip
+distributionUrl=https\://services.gradle.org/distributions/gradle-8.14-bin.zip
 zipStoreBase=GRADLE_USER_HOME
 zipStorePath=wrapper/dists