(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

@@ -59,17 +59,14 @@ 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);
 
+        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 */
     @Override

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

@@ -264,17 +264,16 @@ 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);
                 }
             }

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

@@ -51,6 +51,7 @@ 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;
@@ -635,14 +636,12 @@ public class HttpFetcherImpl implements HttpFetcher, HttpRequestRetryStrategy {
 
     @Override
     public boolean retryRequest(HttpRequest request, IOException exception, int executionCount, HttpContext context) {
-        if (exception instanceof SocketTimeoutException) { // Timeouts are not recoverable
+        return switch (exception) {
-            return false;
+            case SocketTimeoutException ste -> false;
-        }
+            case SSLException ssle -> false;
-        if (exception instanceof SSLException) { // SSL exceptions are unlikely to be recoverable
+            case UnknownHostException uhe -> false;
-            return false;
+            default -> executionCount <= 3;
-        }
+        };
-
-        return executionCount <= 3;
     }
 
     @Override

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

@@ -57,6 +57,7 @@ public abstract class WarcInputBuffer implements AutoCloseable {
             return new ErrorBuffer();
         }
 
+        Instant start = Instant.now();
         InputStream is = null;
         try {
             is = entity.getContent();
@@ -71,8 +72,25 @@ public abstract class WarcInputBuffer implements AutoCloseable {
             }
         }
         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 {
-                is.skip(Long.MAX_VALUE);
+                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

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");
@@ -72,8 +77,23 @@ public class SampleDataExporter {
                 Path crawlDataPath = inputDir.resolve(item.relPath());
                 if (!Files.exists(crawlDataPath)) continue;
 
+                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");
             addFileToTar(stream, newManifestJsonFile, "marginalia-manifest.json");
@@ -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/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/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/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-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/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 {

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