(crawler) Add request time to crawl data

This is an interesting indicator of website quality.

code/index/index-forward/java/nu/marginalia/index/forward/construction/ForwardIndexConverter.java

@@ -84,7 +84,7 @@ public class ForwardIndexConverter {
 
             LongArray docFileData = LongArrayFactory.mmapForWritingConfined(outputFileDocsData, ForwardIndexParameters.ENTRY_SIZE * docsFileId.size());
 
-            ByteBuffer workArea = ByteBuffer.allocate(65536);
+            ByteBuffer workArea = ByteBuffer.allocate(1024*1024*100);
             for (var instance : journal.pages()) {
                 try (var slopTable = new SlopTable(instance.baseDir(), instance.page()))
                 {

code/processes/converting-process/java/nu/marginalia/converting/sideload/SideloaderProcessing.java

@@ -53,6 +53,7 @@ public class SideloaderProcessing {
                 "",
                 body.getBytes(StandardCharsets.UTF_8),
                 false,
+                -1,
                 null,
                 null
         );

code/processes/converting-process/java/org/apache/pdfbox/text/HeadingAwarePDFTextStripper.java

@@ -2002,12 +2002,11 @@ public class HeadingAwarePDFTextStripper extends LegacyPDFStreamEngine
         float minFontWeight = Integer.MAX_VALUE;
         for (var word : line)
         {
-            int i = 0;
             for (var textPosition : word.getTextPositions())
             {
-                if (word.text.charAt(i++) == ' ') {
-                    continue;
-                }
+                // Skip empty text positions as they may have a different font
+                if (word.text.isBlank()) continue;
+
                 var font = textPosition.getFont();
                 if (font == null) continue;
                 var descriptor = font.getFontDescriptor();

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

@@ -148,6 +148,7 @@ public class ConvertingIntegrationTest {
                     "",
                     readClassPathFile(p.toString()).getBytes(),
                     false,
+                    -1,
                     null,
                     null
                     );

code/processes/converting-process/test/nu/marginalia/converting/processor/plugin/PdfDocumentProcessorPluginTest.java

@@ -50,7 +50,7 @@ class PdfDocumentProcessorPluginTest {
                         ));
     }
     public AbstractDocumentProcessorPlugin.DetailsWithWords testPdfFile(byte[] pdfBytes) throws Exception {
-        var doc = new CrawledDocument("test", "https://www.example.com/sample.pdf", "application/pdf", Instant.now().toString(), 200, "OK", "OK", "", pdfBytes, false, null, null);
+        var doc = new CrawledDocument("test", "https://www.example.com/sample.pdf", "application/pdf", Instant.now().toString(), 200, "OK", "OK", "", pdfBytes, false, -1, null, null);
         return plugin.createDetails(doc, new LinkTexts(), DocumentClass.NORMAL);
     }
 

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

@@ -10,6 +10,7 @@ import java.net.http.HttpClient;
 import java.net.http.HttpHeaders;
 import java.net.http.HttpResponse;
 import java.nio.charset.StandardCharsets;
+import java.time.Duration;
 import java.util.*;
 import java.util.stream.Collectors;
 
@@ -90,8 +91,8 @@ public class WarcProtocolReconstructor {
         return "HTTP/" + version + " " + statusCode + " " + statusMessage + "\r\n" + headerString + "\r\n\r\n";
     }
 
-    static String getResponseHeader(ClassicHttpResponse response, long size) {
-        String headerString = getHeadersAsString(response.getHeaders(), size);
+    static String getResponseHeader(ClassicHttpResponse response, Duration responseDuration, long size) {
+        String headerString = getHeadersAsString(response.getHeaders(), responseDuration, size);
 
         return response.getVersion().format() + " " + response.getCode() + " " + response.getReasonPhrase() + "\r\n" + headerString + "\r\n\r\n";
     }
@@ -160,7 +161,7 @@ public class WarcProtocolReconstructor {
 
 
 
-    static private String getHeadersAsString(Header[] headers, long responseSize) {
+    static private String getHeadersAsString(Header[] headers, Duration responseDuration, long responseSize) {
         StringJoiner joiner = new StringJoiner("\r\n");
 
         for (var header : headers) {
@@ -176,6 +177,7 @@ public class WarcProtocolReconstructor {
             if (headerCapitalized.equals("Content-Encoding"))
                 continue;
 
+
             // Since we're transparently decoding gzip, we need to update the Content-Length header
             // to reflect the actual size of the response body. We'll do this at the end.
             if (headerCapitalized.equals("Content-Length"))
@@ -184,6 +186,7 @@ public class WarcProtocolReconstructor {
             joiner.add(headerCapitalized + ": " + header.getValue());
         }
 
+        joiner.add("X-Marginalia-Response-Time: " + responseDuration.toMillis());
         joiner.add("Content-Length: " + responseSize);
 
         return joiner.toString();

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

@@ -93,7 +93,7 @@ public class WarcRecorder implements AutoCloseable {
         WarcDigestBuilder responseDigestBuilder = new WarcDigestBuilder();
         WarcDigestBuilder payloadDigestBuilder = new WarcDigestBuilder();
 
-        Instant date = Instant.now();
+        Instant requestDate = Instant.now();
 
         // Not entirely sure why we need to do this, but keeping it due to Chesterton's Fence
         Map<String, List<String>> extraHeaders = new HashMap<>(request.getHeaders().length);
@@ -108,6 +108,8 @@ public class WarcRecorder implements AutoCloseable {
                 try (WarcInputBuffer inputBuffer = WarcInputBuffer.forResponse(response, request, timeout);
                      InputStream inputStream = inputBuffer.read()) {
 
+                    Instant responseDate = Instant.now();
+
                     cookies.updateCookieStore(response);
 
                     // Build and write the request
@@ -126,7 +128,7 @@ public class WarcRecorder implements AutoCloseable {
 
                     WarcRequest warcRequest = new WarcRequest.Builder(requestUri)
                             .blockDigest(requestDigestBuilder.build())
-                            .date(date)
+                            .date(requestDate)
                             .body(MediaType.HTTP_REQUEST, httpRequestString)
                             .build();
 
@@ -138,7 +140,9 @@ public class WarcRecorder implements AutoCloseable {
                         response.addHeader("X-Has-Cookies", 1);
                     }
 
-                    byte[] responseHeaders = WarcProtocolReconstructor.getResponseHeader(response, inputBuffer.size()).getBytes(StandardCharsets.UTF_8);
+                    byte[] responseHeaders = WarcProtocolReconstructor.getResponseHeader(response,
+                            Duration.between(requestDate, responseDate),
+                            inputBuffer.size()).getBytes(StandardCharsets.UTF_8);
 
                     ResponseDataBuffer responseDataBuffer = new ResponseDataBuffer(inputBuffer.size() + responseHeaders.length);
 
@@ -169,7 +173,7 @@ public class WarcRecorder implements AutoCloseable {
 
                     WarcResponse.Builder responseBuilder = new WarcResponse.Builder(responseUri)
                             .blockDigest(responseDigestBuilder.build())
-                            .date(date)
+                            .date(responseDate)
                             .concurrentTo(warcRequest.id())
                             .body(MediaType.HTTP_RESPONSE, responseDataBuffer.copyBytes());
 
@@ -184,7 +188,7 @@ public class WarcRecorder implements AutoCloseable {
                     warcResponse.http(); // force HTTP header to be parsed before body is consumed so that caller can use it
                     writer.write(warcResponse);
 
-                    if (Duration.between(date, Instant.now()).compareTo(Duration.ofSeconds(9)) > 0
+                    if (Duration.between(requestDate, Instant.now()).compareTo(Duration.ofSeconds(9)) > 0
                             && inputBuffer.size() < 2048
                             && !requestUri.getPath().endsWith("robots.txt")) // don't bail on robots.txt
                     {
@@ -196,7 +200,7 @@ public class WarcRecorder implements AutoCloseable {
 
                         logger.warn("URL {} took too long to fetch ({}s) and was too small for the effort ({}b)",
                                 requestUri,
-                                Duration.between(date, Instant.now()).getSeconds(),
+                                Duration.between(requestDate, Instant.now()).getSeconds(),
                                 inputBuffer.size()
                         );
 

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

@@ -148,6 +148,7 @@ public class ParquetSerializableCrawlDataStream implements AutoCloseable, Serial
                 nextRecord.body,
                 // this field isn't actually used, maybe we can skip calculating it?
                 nextRecord.cookies,
+                -1,
                 lastModified,
                 etag));
     }

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

@@ -166,6 +166,7 @@ public class SlopSerializableCrawlDataStream implements AutoCloseable, Serializa
                 nextRecord.body(),
                 // this field isn't actually used, maybe we can skip calculating it?
                 nextRecord.cookies(),
+                nextRecord.requestTimeMs(),
                 null,
                 null));
     }

code/processes/crawling-process/model/java/nu/marginalia/model/crawldata/CrawledDocument.java

@@ -23,6 +23,7 @@ public final class CrawledDocument implements SerializableCrawlData {
 
     public String crawlerStatus;
     public String crawlerStatusDesc;
+    public int requestTimeMs;
 
     @Nullable
     public String headers;
@@ -82,7 +83,7 @@ public final class CrawledDocument implements SerializableCrawlData {
     public String lastModifiedMaybe;
     public String etagMaybe;
 
-    public CrawledDocument(String crawlId, String url, String contentType, String timestamp, int httpStatus, String crawlerStatus, String crawlerStatusDesc, @Nullable String headers, byte[] documentBodyBytes, Boolean hasCookies, String lastModifiedMaybe, String etagMaybe) {
+    public CrawledDocument(String crawlId, String url, String contentType, String timestamp, int httpStatus, String crawlerStatus, String crawlerStatusDesc, @Nullable String headers, byte[] documentBodyBytes, Boolean hasCookies, int requestTimeMs, String lastModifiedMaybe, String etagMaybe) {
         this.crawlId = crawlId;
         this.url = url;
         this.contentType = contentType;
@@ -94,6 +95,7 @@ public final class CrawledDocument implements SerializableCrawlData {
         this.documentBodyBytes = Objects.requireNonNullElse(documentBodyBytes, new byte[] {});
         this.hasCookies = hasCookies;
         this.lastModifiedMaybe = lastModifiedMaybe;
+        this.requestTimeMs = requestTimeMs;
         this.etagMaybe = etagMaybe;
     }
 
@@ -173,6 +175,7 @@ public final class CrawledDocument implements SerializableCrawlData {
         private byte[] documentBodyBytes = new byte[0];
         private String recrawlState;
         private Boolean hasCookies;
+        private int requestTimeMs;
         private String lastModifiedMaybe;
         private String etagMaybe;
 
@@ -248,8 +251,13 @@ public final class CrawledDocument implements SerializableCrawlData {
             return this;
         }
 
+        public CrawledDocumentBuilder requestTimeMs(int requestTimeMs) {
+            this.requestTimeMs = requestTimeMs;
+            return this;
+        }
+
         public CrawledDocument build() {
-            return new CrawledDocument(this.crawlId, this.url, this.contentType, this.timestamp, this.httpStatus, this.crawlerStatus, this.crawlerStatusDesc, this.headers, this.documentBodyBytes, this.hasCookies, this.lastModifiedMaybe, this.etagMaybe);
+            return new CrawledDocument(this.crawlId, this.url, this.contentType, this.timestamp, this.httpStatus, this.crawlerStatus, this.crawlerStatusDesc, this.headers, this.documentBodyBytes, this.hasCookies, this.requestTimeMs, this.lastModifiedMaybe, this.etagMaybe);
         }
 
         public String toString() {

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

@@ -9,6 +9,7 @@ import nu.marginalia.parquet.crawldata.CrawledDocumentParquetRecord;
 import nu.marginalia.parquet.crawldata.CrawledDocumentParquetRecordFileReader;
 import nu.marginalia.slop.column.array.ByteArrayColumn;
 import nu.marginalia.slop.column.primitive.ByteColumn;
+import nu.marginalia.slop.column.primitive.IntColumn;
 import nu.marginalia.slop.column.primitive.LongColumn;
 import nu.marginalia.slop.column.primitive.ShortColumn;
 import nu.marginalia.slop.column.string.EnumColumn;
@@ -39,6 +40,7 @@ public record SlopCrawlDataRecord(String domain,
                                   long timestamp,
                                   String contentType,
                                   byte[] body,
+                                  int requestTimeMs,
                                   String headers)
 {
     private static final EnumColumn domainColumn = new EnumColumn("domain", StandardCharsets.UTF_8, StorageType.ZSTD);
@@ -49,6 +51,7 @@ public record SlopCrawlDataRecord(String domain,
     private static final LongColumn timestampColumn = new LongColumn("timestamp");
     private static final EnumColumn contentTypeColumn = new EnumColumn("contentType", StandardCharsets.UTF_8);
     private static final ByteArrayColumn bodyColumn = new ByteArrayColumn("body", StorageType.ZSTD);
+    private static final ShortColumn requestTimeColumn = new ShortColumn("requestTimeMs");
     private static final StringColumn headerColumn = new StringColumn("header", StandardCharsets.UTF_8, StorageType.ZSTD);
 
     public SlopCrawlDataRecord(CrawledDocumentParquetRecord parquetRecord) {
@@ -60,6 +63,7 @@ public record SlopCrawlDataRecord(String domain,
                 parquetRecord.timestamp.toEpochMilli(),
                 parquetRecord.contentType,
                 parquetRecord.body,
+                -1,
                 parquetRecord.headers
                 );
     }
@@ -74,6 +78,7 @@ public record SlopCrawlDataRecord(String domain,
                 date.toEpochMilli(),
                 "x-marginalia/advisory;state=redirect",
                 new byte[0],
+                -1,
                 ""
         );
     }
@@ -87,6 +92,7 @@ public record SlopCrawlDataRecord(String domain,
                 date.toEpochMilli(),
                 "x-marginalia/advisory;state=error",
                 errorStatus.getBytes(),
+                -1,
                 ""
         );
     }
@@ -100,6 +106,7 @@ public record SlopCrawlDataRecord(String domain,
                 date.toEpochMilli(),
                 errorStatus,
                 new byte[0],
+                -1,
                 ""
         );
     }
@@ -391,10 +398,20 @@ public record SlopCrawlDataRecord(String domain,
 
             String headersStr;
             StringJoiner headersStrBuilder = new StringJoiner("\n");
+            int requestTimeMs = -1;
             for (var header : headers) {
                 if (header.getName().equalsIgnoreCase("X-Cookies") && "1".equals(header.getValue())) {
                     hasCookies = true;
                 }
+                if (header.getName().equals("X-Marginalia-Response-Time")) {
+                    try {
+                        requestTimeMs = Integer.parseInt(header.getValue());
+                    }
+                    catch (NumberFormatException ex) {
+                        logger.warn("Failed to parse X-Marginalia-Response-Time header: {}", header.getValue());
+                    }
+                    continue;
+                }
                 headersStrBuilder.add(header.getName() + ": " + header.getValue());
             }
             headersStr = headersStrBuilder.toString();
@@ -409,6 +426,7 @@ public record SlopCrawlDataRecord(String domain,
                     response.date().toEpochMilli(),
                     contentType,
                     bodyBytes,
+                    requestTimeMs,
                     headersStr
                 )
             );
@@ -461,6 +479,7 @@ public record SlopCrawlDataRecord(String domain,
         private final LongColumn.Reader timestampColumnReader;
         private final EnumColumn.Reader contentTypeColumnReader;
         private final ByteArrayColumn.Reader bodyColumnReader;
+        private final ShortColumn.Reader requestTimeColumnReader;
         private final StringColumn.Reader headerColumnReader;
 
         public Reader(Path path) throws IOException {
@@ -474,6 +493,7 @@ public record SlopCrawlDataRecord(String domain,
             timestampColumnReader = timestampColumn.open(this);
             contentTypeColumnReader = contentTypeColumn.open(this);
             bodyColumnReader = bodyColumn.open(this);
+            requestTimeColumnReader = requestTimeColumn.open(this);
             headerColumnReader = headerColumn.open(this);
         }
 
@@ -487,6 +507,7 @@ public record SlopCrawlDataRecord(String domain,
                     timestampColumnReader.get(),
                     contentTypeColumnReader.get(),
                     bodyColumnReader.get(),
+                    requestTimeColumnReader.get(),
                     headerColumnReader.get()
             );
         }
@@ -506,6 +527,7 @@ public record SlopCrawlDataRecord(String domain,
         private final LongColumn.Reader timestampColumnReader;
         private final EnumColumn.Reader contentTypeColumnReader;
         private final ByteArrayColumn.Reader bodyColumnReader;
+        private final ShortColumn.Reader requestTimeColumnReader;
         private final StringColumn.Reader headerColumnReader;
 
         private SlopCrawlDataRecord next = null;
@@ -521,6 +543,7 @@ public record SlopCrawlDataRecord(String domain,
             timestampColumnReader = timestampColumn.open(this);
             contentTypeColumnReader = contentTypeColumn.open(this);
             bodyColumnReader = bodyColumn.open(this);
+            requestTimeColumnReader = requestTimeColumn.open(this);
             headerColumnReader = headerColumn.open(this);
         }
 
@@ -548,6 +571,7 @@ public record SlopCrawlDataRecord(String domain,
                 boolean cookies = cookiesColumnReader.get() == 1;
                 int status = statusColumnReader.get();
                 long timestamp = timestampColumnReader.get();
+                int requestTimeMs = requestTimeColumnReader.get();
                 String contentType = contentTypeColumnReader.get();
 
                 LargeItem<byte[]> body = bodyColumnReader.getLarge();
@@ -555,7 +579,7 @@ public record SlopCrawlDataRecord(String domain,
 
                 if (filter(url, status, contentType)) {
                     next = new SlopCrawlDataRecord(
-                            domain, url, ip, cookies, status, timestamp, contentType, body.get(), headers.get()
+                            domain, url, ip, cookies, status, timestamp, contentType, body.get(), requestTimeMs, headers.get()
                     );
                     return true;
                 }

code/processes/live-crawling-process/java/nu/marginalia/livecrawler/LiveCrawlDataSet.java

@@ -195,6 +195,7 @@ public class LiveCrawlDataSet implements AutoCloseable {
                             headers,
                             body,
                             false,
+                            -1,
                             "",
                             ""
                     ));

deploy.txt

@@ -7,4 +7,5 @@
 2025-05-04:  Deploy qs, search and api-services.
 2025-05-05:  Deploy executor partition 4.
 2025-05-05:  Deploy control.
-2025-05-08:  Deploy assistant.
+2025-05-08:  Deploy assistant.
+2025-05-17:  Redeploy all.