RED-5718: Run the ocr service as specific user

Merge in RED/ocr-service from RED-6126 to master

* commit '0bc4fea2a52c92efaaaf8cf93c2ae02766168a80':
  RED-6126: In the OCRService, OCR Text is not applied to Document *removed unnecessary getXObject() call, since it fails for inline_images

ocr-service-image-v1/src/main/docker/Dockerfile

@@ -1,4 +1,4 @@
-FROM red/base-image:2.0.0
+FROM red/base-image:baseImageUser5
 
 COPY "libs/pdftron/OCRModuleLinux.tar.gz" .
 RUN tar xvzf OCRModuleLinux.tar.gz

ocr-service-v1/ocr-service-server-v1/src/main/java/com/iqser/red/service/ocr/v1/server/service/ImagePositionRetrievalService.java

@@ -27,9 +27,12 @@ public class ImagePositionRetrievalService {
 
     private static final double TOLERANCE = 1e-1;
 
+    // any image with smaller height and width than this gets thrown out, see everyPointInDashedLineIsImage.pdf
+    private static final int PIXEL_THRESHOLD = 10;
+
 
     /**
-     * Iterates over all elements in a PDF Document and retrieves the bounding box for each image,
+     * Iterates over all elements in a PDF Document and retrieves the bounding box for each image, that is larger than the pixel threshold of 10 in either dimension.
      * Then it adjusts the bounding boxes for the page rotation.
      * If the mirrorY flag is set, the Y Coordinates are mirrored and moved up by the page height. This is required for PDFTrons OCRModule.
      *
@@ -63,7 +66,12 @@ public class ImagePositionRetrievalService {
         Element element;
         while ((element = reader.next()) != null) {
             switch (element.getType()) {
-                case Element.e_image, Element.e_inline_image -> imagePositions.addRect(toRotationAdjustedRect(element.getBBox(), currentPage, mirrorY));
+                case Element.e_image, Element.e_inline_image -> {
+                    // see everyPointInDashedLineIsImage.pdf TestFile
+                    if (element.getImageHeight() > PIXEL_THRESHOLD || element.getImageWidth() > PIXEL_THRESHOLD) {
+                        imagePositions.addRect(toRotationAdjustedRect(element.getBBox(), currentPage, mirrorY));
+                    }
+                }
                 case Element.e_form -> {
                     reader.formBegin();
                     findImagePositionsOnPage(reader, imagePositions, currentPage, mirrorY);
@@ -77,39 +85,49 @@ public class ImagePositionRetrievalService {
     @SneakyThrows
     public RectCollection mergeOverlappingRects(RectCollection imagePositions) {
 
-        if (imagePositions.getNumRects() == 1) {
+        if (imagePositions.getNumRects() < 2) {
             return imagePositions;
         }
 
         List<Rectangle2D> rectangleList = toSortedRectangleList(imagePositions);
 
-        rectangleList = mergeRectangleListRecursive(rectangleList, 0);
+        mergeRectangleList(rectangleList);
+
         return toRectCollection(rectangleList);
     }
 
 
     // Sometimes images are split up into stripes, here we merge the positions of aligned and intersecting rectangles into one larger rectangle
-    private List<Rectangle2D> mergeRectangleListRecursive(List<Rectangle2D> rectangleList, int currentIdx) {
+    private void mergeRectangleList(List<Rectangle2D> rectangleList) {
 
-        if (rectangleList.size() < currentIdx + 2) {
-            return rectangleList;
+        for (int idx = 0; rectangleList.size() >= idx + 2; ) {
+
+            var rect1 = rectangleList.get(idx);
+            var rect2 = rectangleList.get(idx + 1);
+
+            if (intersects(rect1, rect2) && isAlignedXOrY(rect1, rect2)) {
+                rectangleList.remove(idx + 1);
+                rectangleList.remove(idx);
+                rectangleList.add(idx, rect1.createUnion(rect2));
+            } else {
+                ++idx;
+            }
         }
+    }
 
-        var rect1 = rectangleList.get(currentIdx);
-        var rect2 = rectangleList.get(currentIdx + 1);
+
+    private boolean intersects(Rectangle2D rect1, Rectangle2D rect2) {
+
+        return rect1.intersects(rect2.getMinX() - TOLERANCE, rect2.getMinY() - TOLERANCE, rect2.getWidth() + (2 * TOLERANCE), rect2.getHeight() + (2 * TOLERANCE));
+    }
+
+
+    private boolean isAlignedXOrY(Rectangle2D rect1, Rectangle2D rect2) {
 
         boolean isAlignedX = Math.abs(rect1.getMinX() - rect2.getMinX()) < TOLERANCE && Math.abs(rect1.getMaxX() - rect2.getMaxX()) < TOLERANCE;
         boolean isAlignedY = Math.abs(rect1.getMinY() - rect2.getMinY()) < TOLERANCE && Math.abs(rect1.getMaxY() - rect2.getMaxY()) < TOLERANCE;
-        boolean intersects = rect1.intersects(rect2.getMinX() - TOLERANCE, rect2.getMinY() - TOLERANCE, rect2.getWidth() + (2 * TOLERANCE), rect2.getHeight() + (2 * TOLERANCE));
 
-        if (intersects && (isAlignedX || isAlignedY)) {
-            rectangleList.remove(currentIdx + 1);
-            rectangleList.remove(currentIdx);
-            rectangleList.add(currentIdx, rect1.createUnion(rect2));
-            return mergeRectangleListRecursive(rectangleList, currentIdx);
-        } else {
-            return mergeRectangleListRecursive(rectangleList, currentIdx + 1);
-        }
+        return isAlignedX || isAlignedY;
     }
 
 

ocr-service-v1/ocr-service-server-v1/src/main/java/com/iqser/red/service/ocr/v1/server/service/InvisibleElementRemovalService.java

@@ -76,6 +76,8 @@ public class InvisibleElementRemovalService {
             Page page = iterator.next();
 
             visitedXObjIds.add(page.getSDFObj().getObjNum());
+
+
             InvisibleElementRemovalContext context = InvisibleElementRemovalContext.builder()
                     .reader(reader)
                     .clippingPathStack(new ClippingPathStack(page.getMediaBox()))
@@ -221,8 +223,14 @@ public class InvisibleElementRemovalService {
 
 
     private void processPath(Element pathElement, ElementWriter writer, InvisibleElementRemovalContext context) throws PDFNetException {
+        PathData pathData = pathElement.getPathData();
 
-        GeneralPath linePath = convertToGeneralPath(pathElement.getPathData());
+        if (pathData.getOperators().length == 0 && pathData.getPoints().length == 0) {
+            writer.writeGStateChanges(pathElement);
+            return;
+        }
+
+        GeneralPath linePath = convertToGeneralPath(pathData);
 
         //transform path to initial user space
         var ctm = pathElement.getCTM();

ocr-service-v1/ocr-service-server-v1/src/main/java/com/iqser/red/service/ocr/v1/server/service/OCRService.java

@@ -1,7 +1,10 @@
 package com.iqser.red.service.ocr.v1.server.service;
 
+import static java.lang.String.format;
+
 import java.io.ByteArrayInputStream;
 import java.io.ByteArrayOutputStream;
+import java.io.IOException;
 import java.io.InputStream;
 import java.io.OutputStream;
 import java.util.Map;
@@ -59,16 +62,22 @@ public class OCRService {
      * @param fileId    The file id
      * @param out       OutputStream to write the file to
      */
-    @SneakyThrows
     @Timed("redactmanager_runOcrOnDocument")
-    public void runOcrOnDocument(String dossierId, String fileId, OutputStream out) {
+    public void runOcrOnDocument(String dossierId, String fileId, OutputStream out) throws IOException {
 
         try (ByteArrayOutputStream transferOutputStream = new ByteArrayOutputStream()) {
             try (InputStream fileStream = fileStorageService.getOriginalFileAsStream(dossierId, fileId)) {
+                long start = System.currentTimeMillis();
+                log.debug("Start invisible element removal for file with dossierId {} and fileId {}", dossierId, fileId);
                 invisibleElementRemovalService.removeInvisibleElements(fileStream, transferOutputStream, false);
+                long end = System.currentTimeMillis();
+                log.debug("Invisible element removal successful for file with dossierId {} and fileId {}, took {}s", dossierId, fileId, format("%.1f", (end - start) / 1000.0));
             }
             try (InputStream transferInputStream = new ByteArrayInputStream(transferOutputStream.toByteArray())) {
+                long start = System.currentTimeMillis();
                 runOcr(transferInputStream, out, fileId);
+                long end = System.currentTimeMillis();
+                log.info("ocr successful for file with dossierId {} and fileId {}, took {}s", dossierId, fileId, format("%.1f", (end - start) / 1000.0));
             }
         }
     }
@@ -81,36 +90,26 @@ public class OCRService {
 
         Map<Integer, RectCollection> pageIdToRectCollection = imagePositionRetrievalService.getImagePositionPerPage(pdfDoc, true);
 
+        // Optimization:
+        // When a page does not have a TextZone, PDFTron whites out the page. But, PDFTron scans it anyway, resulting in a longer runtime.
+        // So, we need to remove pages without images.
+        // Furthermore, creating a new document is *much* faster than reusing the same document and adding/removing pages one by one.
+        // Therefore, we create a new Document with a single page for every page that contains text.
         int numProcessedPages = 0;
         for (Integer pageId : pageIdToRectCollection.keySet()) {
             try {
-                // optimization: creating a new document is faster than reusing the same and adding/removing pages one by one
-                OCROptions options = new OCROptions();
-                PDFDoc ocrPageDoc = new PDFDoc();
-                // optimization: only scanning pages that contain images
-                Page pdfPage = pdfDoc.getPage(pageId);
-                // optimization: this line ensures the ocr text is placed correctly by PDFTron
-                pdfPage.setMediaBox(pdfPage.getCropBox());
-                ocrPageDoc.pagePushBack(pdfPage);
-                options.addTextZonesForPage(pageIdToRectCollection.get(pageId), 1);
-                options.addLang(ENGLISH);
-                options.addDPI(settings.getOcrDPI());
-
-                OCRModule.processPDF(ocrPageDoc, options);
+                PDFDoc singlePagePdfDoc = extractSinglePagePdfDoc(pdfDoc, pageId);
+                processOcr(pageIdToRectCollection, pageId, singlePagePdfDoc);
                 ++numProcessedPages;
 
-                StringBuilder zonesString = new StringBuilder();
-                for (int j = 0; j < pageIdToRectCollection.get(pageId).getNumRects(); ++j) {
-                    var r = pageIdToRectCollection.get(pageId).getRectAt(j);
-                    zonesString.append(String.format("[lower left (%.1f|%.1f) upper right (%.1f|%.1f)]", r.getX1(), r.getY1(), r.getX2(), r.getY2()));
-                }
-                log.info("{}/{} Page {} done, OCR regions {}", numProcessedPages, pageIdToRectCollection.size(), pageId, zonesString);
+                log.info("{}/{} Page {} done, OCR regions {}",
+                        numProcessedPages,
+                        pageIdToRectCollection.size(),
+                        pageId,
+                        getAllOcrTextZonesAsString(pageIdToRectCollection, pageId));
 
-                // re-adding OCR pages
-                Page ocrPage = ocrPageDoc.getPage(1);
-                pdfDoc.pageInsert(pdfDoc.getPageIterator(pageId), ocrPage);
-                pdfDoc.pageRemove(pdfDoc.getPageIterator(pageId + 1));
-                ocrPageDoc.close();
+                replaceOriginalPageWithOcrPage(pdfDoc, pageId, singlePagePdfDoc);
+                singlePagePdfDoc.close();
 
                 rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_QUEUE,
                         objectMapper.writeValueAsString(OCRStatusUpdateResponse.builder()
@@ -120,7 +119,7 @@ public class OCRService {
                                 .build()));
 
             } catch (PDFNetException e) {
-                log.error("failed to process page {}", pageId);
+                log.error("Failed to process page {}", pageId);
                 throw new RuntimeException(e);
             }
         }
@@ -134,7 +133,52 @@ public class OCRService {
                         .build()));
 
         Optimizer.optimize(pdfDoc);
-        pdfDoc.save(out, SDFDoc.SaveMode.LINEARIZED, null);
+        try {
+            pdfDoc.save(out, SDFDoc.SaveMode.LINEARIZED, null);
+        } catch (Exception e) {
+            log.error("Processed File with fileId {} could not be saved", fileId);
+            throw new RuntimeException(e);
+        }
+    }
+
+
+    private void processOcr(Map<Integer, RectCollection> pageIdToRectCollection, Integer pageId, PDFDoc singlePagePdfDoc) throws PDFNetException {
+
+        OCROptions options = new OCROptions();
+        options.addTextZonesForPage(pageIdToRectCollection.get(pageId), 1);
+        options.addLang(ENGLISH);
+        options.addDPI(settings.getOcrDPI());
+
+        OCRModule.processPDF(singlePagePdfDoc, options);
+    }
+
+
+    private static PDFDoc extractSinglePagePdfDoc(PDFDoc pdfDoc, Integer pageId) throws PDFNetException {
+
+        PDFDoc singlePagePdfDoc = new PDFDoc();
+        Page page = pdfDoc.getPage(pageId);
+        page.setMediaBox(page.getCropBox()); // this line ensures the ocr text is placed correctly by PDFTron, see TestFile MediaBoxBiggerThanCropBox.pdf
+        singlePagePdfDoc.pagePushBack(page);
+        return singlePagePdfDoc;
+    }
+
+
+    private static void replaceOriginalPageWithOcrPage(PDFDoc pdfDoc, Integer pageId, PDFDoc ocrPageDoc) throws PDFNetException {
+
+        Page ocrPage = ocrPageDoc.getPage(1);
+        pdfDoc.pageInsert(pdfDoc.getPageIterator(pageId), ocrPage);
+        pdfDoc.pageRemove(pdfDoc.getPageIterator(pageId + 1));
+    }
+
+
+    private static StringBuilder getAllOcrTextZonesAsString(Map<Integer, RectCollection> pageIdToRectCollection, Integer pageId) throws PDFNetException {
+
+        StringBuilder zonesString = new StringBuilder();
+        for (int j = 0; j < pageIdToRectCollection.get(pageId).getNumRects(); ++j) {
+            var r = pageIdToRectCollection.get(pageId).getRectAt(j);
+            zonesString.append(format("[lower left (%.1f|%.1f) upper right (%.1f|%.1f)]", r.getX1(), r.getY1(), r.getX2(), r.getY2()));
+        }
+        return zonesString;
     }
 
 }

ocr-service-v1/ocr-service-server-v1/src/main/java/com/iqser/red/service/ocr/v1/server/service/OcrMessageReceiver.java

@@ -38,7 +38,6 @@ public class OcrMessageReceiver {
 
         DocumentRequest ocrRequestMessage = objectMapper.readValue(in, DocumentRequest.class);
 
-        long start = System.currentTimeMillis();
         log.info("Start ocr for file with dossierId {} and fileId {}", ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId());
 
         setStatusOcrProcessing(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId());
@@ -48,19 +47,18 @@ public class OcrMessageReceiver {
             fileStorageService.storeUntouchedFile(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), originalFile);
         }
 
-        try (var out = new ByteArrayOutputStream()) {
-            ocrService.runOcrOnDocument(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), out);
-
-            fileStorageService.storeOriginalFile(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), new ByteArrayInputStream(out.toByteArray()));
+        try (var transferStream = new ByteArrayOutputStream()) {
+            ocrService.runOcrOnDocument(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), transferStream);
+            try (var inputStream = new ByteArrayInputStream(transferStream.toByteArray())) {
+                fileStorageService.storeOriginalFile(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), inputStream);
+            }
         } catch (IOException e) {
+            log.error("Failed to store file with dossierId {} and fileId {}", ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId());
             throw new RuntimeException(e);
         }
 
-
-        long end = System.currentTimeMillis();
-        log.info("Successfully processed ocr for file with dossierId {} and fileId {}, took {}", ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId(), end - start);
-
         fileStatusProcessingUpdateClient.ocrSuccessful(ocrRequestMessage.getDossierId(), ocrRequestMessage.getFileId());
+
     }
 
 

ocr-service-v1/ocr-service-server-v1/src/test/java/com/iqser/red/service/ocr/v1/server/OcrServiceIntegrationTest.java

@@ -5,12 +5,12 @@ import static org.assertj.core.api.Assertions.assertThat;
 
 import java.io.FileInputStream;
 import java.io.FileOutputStream;
+import java.io.IOException;
+import java.io.InputStream;
 import java.util.ArrayList;
 import java.util.List;
 import java.util.concurrent.TimeUnit;
 
-import io.micrometer.prometheus.PrometheusMeterRegistry;
-import io.micrometer.prometheus.PrometheusTimer;
 import org.junit.jupiter.api.AfterEach;
 import org.junit.jupiter.api.BeforeEach;
 import org.junit.jupiter.api.Disabled;
@@ -36,12 +36,15 @@ import com.iqser.red.service.ocr.v1.server.utils.FileSystemBackedStorageService;
 import com.iqser.red.service.persistence.service.v1.api.model.dossiertemplate.dossier.file.FileType;
 import com.iqser.red.storage.commons.StorageAutoConfiguration;
 import com.iqser.red.storage.commons.service.StorageService;
+import com.pdftron.common.PDFNetException;
 import com.pdftron.pdf.OCRModule;
 import com.pdftron.pdf.PDFDoc;
 import com.pdftron.pdf.Page;
 import com.pdftron.pdf.PageIterator;
 import com.pdftron.pdf.TextExtractor;
 
+import io.micrometer.prometheus.PrometheusMeterRegistry;
+import io.micrometer.prometheus.PrometheusTimer;
 import lombok.SneakyThrows;
 
 @ExtendWith(SpringExtension.class)
@@ -80,19 +83,20 @@ public class OcrServiceIntegrationTest {
 
     @Test
     @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
-    public void testOCRMetrics(){
+    public void testOCRMetrics() {
+
         testOCR("Watermark");
         testOCR("Watermark");
         testOCR("Watermark");
 
-        var ocrOnDocumentMeter = registry.getMeters().stream()
-                .filter(m -> m.getId().getName().equalsIgnoreCase("redactmanager_runOcrOnDocument")).findAny();
+        var ocrOnDocumentMeter = registry.getMeters().stream().filter(m -> m.getId().getName().equalsIgnoreCase("redactmanager_runOcrOnDocument")).findAny();
         assertThat(ocrOnDocumentMeter.isPresent()).isTrue();
         PrometheusTimer timer = (PrometheusTimer) ocrOnDocumentMeter.get();
         assertThat(timer.count()).isEqualTo(3);
         assertThat(timer.mean(TimeUnit.SECONDS)).isGreaterThan(0.1);
     }
 
+
     @Test
     @Disabled // OCRModule is not available on build server. If you want to run the test set the property at the top.
     public void testOcr() {
@@ -153,30 +157,34 @@ public class OcrServiceIntegrationTest {
     private String testOCR(String fileName) {
 
         ClassPathResource pdfFileResource = new ClassPathResource("files/" + fileName + ".pdf");
-
         var originId = FileStorageService.getStorageId("dossier", "file", FileType.ORIGIN);
         try (var fileStream = pdfFileResource.getInputStream()) {
             storageService.storeObject(originId, fileStream);
         }
-
         try (var out = new FileOutputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
             ocrService.runOcrOnDocument("dossier", "file", out);
         }
 
+        System.out.println("File:" + getTemporaryDirectory() + "/" + fileName + ".pdf");
+        try (var fileStream = new FileInputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
+            return extractAllTextFromDocument(fileStream);
+        }
+    }
+
+
+    private static String extractAllTextFromDocument(InputStream fileStream) throws IOException, PDFNetException {
+
+        PDFDoc pdfDoc = new PDFDoc(fileStream);
         TextExtractor extractor = new TextExtractor();
         List<String> texts = new ArrayList<>();
-        PDFDoc pdfDoc;
 
-        try (var fileStream = new FileInputStream(getTemporaryDirectory() + "/" + fileName + ".pdf")) {
-            pdfDoc = new PDFDoc(fileStream);
-        }
         PageIterator iterator = pdfDoc.getPageIterator();
         while (iterator.hasNext()) {
             Page page = iterator.next();
             extractor.begin(page);
             texts.add(extractor.getAsText());
         }
-        System.out.println("File:" + getTemporaryDirectory() + "/" + fileName + ".pdf");
+
         return String.join("\n", texts);
     }
 
@@ -184,7 +192,7 @@ public class OcrServiceIntegrationTest {
     @SneakyThrows
     public void dummyTest() {
 
-        // Build needs one text to not fail.
+        // Build needs one test to not fail.
         assertThat(1).isEqualTo(1);
     }
 
@@ -204,7 +212,7 @@ public class OcrServiceIntegrationTest {
 
         @Bean
         @Primary
-        public StorageService inmemoryStorage() {
+        public StorageService inMemoryStorage() {
 
             return new FileSystemBackedStorageService();
         }

ocr-service-v1/ocr-service-server-v1/src/test/java/com/iqser/red/service/ocr/v1/server/service/ImagePositionRetrievalServiceTest.java

@@ -122,6 +122,14 @@ class ImagePositionRetrievalServiceTest {
         assertThat(allRectCoords.size()).isEqualTo(48);
     }
 
+    @Test
+    @SneakyThrows
+    public void testEveryPointInDashedLineIsImage() {
+        String fileName = "everyPointInDashedLineIsImage";
+        List<int[]> allRectCoords = testImagePositionDetection(fileName);
+        assertThat(allRectCoords.size()).isEqualTo(0);
+    }
+
 
     private List<int[]> testImagePositionDetection(String fileName) throws IOException, PDFNetException {