Merge branch 'RED-8933-4.0' into 'release/0.89.x'

RED-8933: Fixed bugs in DocumineClassificationService

See merge request fforesight/layout-parser!149

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/LayoutParsingPipeline.java

@@ -28,6 +28,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
@@ -94,8 +95,10 @@ public class LayoutParsingPipeline {
         log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());
 
         File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
-        File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId())
-                .orElse(originFile);
+//        File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId())
+//                .orElse(originFile);
+        File viewerDocumentFile = originFile;
+
 
         ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
         if (layoutParsingRequest.imagesFileStorageId()
@@ -248,8 +251,10 @@ public class LayoutParsingPipeline {
             PDRectangle cropbox = pdPage.getCropBox();
             CleanRulings cleanRulings = rulingCleaningService.getCleanRulings(pdfTableCells.get(pageNumber), stripper.getRulings());
 
+            List<Cell> emptyTableCells = TableExtractionService.findCells(cleanRulings.getHorizontal(), cleanRulings.getVertical());
+
             ClassificationPage classificationPage = switch (layoutParsingType) {
-                case REDACT_MANAGER -> redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
+                case REDACT_MANAGER -> redactManagerBlockificationService.blockify(stripper.getTextPositionSequences(), emptyTableCells);
                 case TAAS -> taasBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
                 case DOCUMINE -> docuMineBlockificationService.blockify(stripper.getTextPositionSequences(), cleanRulings.getHorizontal(), cleanRulings.getVertical());
             };

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/graph/nodes/Image.java

@@ -12,7 +12,6 @@ import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.No
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.entity.RedactionEntity;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
-import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlockCollector;
 
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -36,6 +35,8 @@ public class Image implements GenericSemanticNode {
     boolean transparent;
     Rectangle2D position;
 
+    TextBlock leafTextBlock;
+
     boolean redaction;
     boolean ignored;
     @Builder.Default
@@ -66,7 +67,7 @@ public class Image implements GenericSemanticNode {
     @Override
     public TextBlock getTextBlock() {
 
-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return leafTextBlock;
     }
 
 
@@ -92,4 +93,11 @@ public class Image implements GenericSemanticNode {
         return bBoxPerPage;
     }
 
+
+    @Override
+    public boolean isLeaf() {
+
+        return true;
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/RedTextPosition.java

@@ -1,7 +1,6 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.text;
 
 import org.apache.pdfbox.text.TextPosition;
-import org.springframework.beans.BeanUtils;
 
 import com.fasterxml.jackson.annotation.JsonIgnore;
 
@@ -50,7 +49,13 @@ public class RedTextPosition {
     public static RedTextPosition fromTextPosition(TextPosition textPosition) {
 
         var pos = new RedTextPosition();
-        BeanUtils.copyProperties(textPosition, pos);
+        pos.setRotation(textPosition.getRotation());
+        pos.setPageHeight(textPosition.getPageHeight());
+        pos.setPageWidth(textPosition.getPageWidth());
+        pos.setUnicode(textPosition.getUnicode());
+        pos.setDir(textPosition.getDir());
+        pos.setWidthOfSpace(textPosition.getWidthOfSpace());
+        pos.setFontSizeInPt(textPosition.getFontSizeInPt());
         pos.setFontName(textPosition.getFont().getName());
 
         pos.setFontSizeInPt(textPosition.getFontSizeInPt());

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/text/TextPositionSequence.java

@@ -190,6 +190,12 @@ public class TextPositionSequence implements CharSequence {
     }
 
 
+    public float getTextHeightNoPadding() {
+
+        return textPositions.get(0).getHeightDir();
+    }
+
+
     public float getTextHeight() {
 
         return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/SectionsBuilderService.java

@@ -43,7 +43,6 @@ public class SectionsBuilderService {
         for (ClassificationPage page : document.getPages()) {
             List<TextPageBlock> header = new ArrayList<>();
             List<TextPageBlock> footer = new ArrayList<>();
-            List<TextPageBlock> unclassifiedText = new ArrayList<>();
             for (AbstractPageBlock current : page.getTextBlocks()) {
 
                 if (current.getClassification() == null) {
@@ -62,11 +61,6 @@ public class SectionsBuilderService {
                     continue;
                 }
 
-                if (current.getClassification().equals(PageBlockType.OTHER)) {
-                    unclassifiedText.add((TextPageBlock) current);
-                    continue;
-                }
-
                 if (prev != null && current.getClassification().isHeadline() && !prev.getClassification().isHeadline() || !document.isHeadlines()) {
                     ClassificationSection chunkBlock = buildTextBlock(chunkWords, lastHeadline);
                     chunkBlock.setHeadline(lastHeadline);
@@ -94,9 +88,6 @@ public class SectionsBuilderService {
             if (!footer.isEmpty()) {
                 footers.add(new ClassificationFooter(footer));
             }
-            if (!unclassifiedText.isEmpty()) {
-                unclassifiedTexts.add(new UnclassifiedText(unclassifiedText));
-            }
         }
 
         ClassificationSection chunkBlock = buildTextBlock(chunkWords, lastHeadline);

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/blockification/DocuMineBlockificationService.java

@@ -61,8 +61,8 @@ public class DocuMineBlockificationService {
             boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
             boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
             boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
-            boolean splitByOtherFontAndOtherY = prev != null && prev.getMaxYDirAdj() != word.getMaxYDirAdj() && (word.getFontStyle().contains("bold") && !prev.getFontStyle()
-                    .contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
+            boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 && (word.getFontStyle()
+                    .contains("bold") && !prev.getFontStyle().contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
 
             Matcher matcher = pattern.matcher(chunkWords.stream().collect(Collectors.joining(" ")).toString());
             boolean startsOnSameX = Math.abs(minX - word.getMinXDirAdj()) < 5 && matcher.matches();

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/blockification/RedactManagerBlockificationService.java

@@ -2,6 +2,7 @@ package com.knecon.fforesight.service.layoutparser.processor.services.blockifica
 
 import static java.util.stream.Collectors.toSet;
 
+import java.awt.geom.Point2D;
 import java.util.ArrayList;
 import java.util.Comparator;
 import java.util.Iterator;
@@ -13,10 +14,13 @@ import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBl
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.Orientation;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RulingTextDirAdjustUtil;
 
 @SuppressWarnings("all")
@@ -31,12 +35,12 @@ public class RedactManagerBlockificationService {
      * This method must use text direction adjusted postions (DirAdj). Where {0,0} is on the upper left. Never try to change this!
      * Rulings (TablePageBlock lines) must be adjusted to the text directions as well, when checking if a block is split by a ruling.
      *
-     * @param textPositions         The words of a page.
-     * @param horizontalRulingLines Horizontal table lines.
-     * @param verticalRulingLines   Vertical table lines.
+     * @param textPositions The words of a page.
      * @return Page object that contains the Textblock and text statistics.
      */
-    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+    public ClassificationPage blockify(List<TextPositionSequence> textPositions, List<Cell> emptyCells) {
+
+        CleanRulings usedRulings = RectangleTransformations.extractRulings(emptyCells);
 
         int indexOnPage = 0;
         List<TextPositionSequence> chunkWords = new ArrayList<>();
@@ -54,7 +58,7 @@ public class RedactManagerBlockificationService {
             boolean splitByX = prev != null && maxX + 50 < word.getMinXDirAdj() && prev.getMinYDirAdj() == word.getMinYDirAdj();
             boolean xIsBeforeFirstX = prev != null && word.getMinXDirAdj() < minX;
             boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
-            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
+            boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, usedRulings.getHorizontal(), usedRulings.getVertical());
             boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
 
             if (prev != null && (lineSeparation || startFromTop || splitByX || splitByDir || isSplitByRuling)) {

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/classification/DocuMineClassificationService.java

@@ -5,7 +5,6 @@ import java.util.Locale;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 
-import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
@@ -13,6 +12,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
 import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PositionUtils;
 
 import lombok.RequiredArgsConstructor;
@@ -23,7 +23,7 @@ import lombok.extern.slf4j.Slf4j;
 @RequiredArgsConstructor
 public class DocuMineClassificationService {
 
-    private static final Pattern pattern = Pattern.compile("^(\\d{1,2}\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
+    private static final Pattern pattern = Pattern.compile("^([1-9]\\d?\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
     private static final Pattern pattern2 = Pattern.compile("\\p{L}{3,}", Pattern.CASE_INSENSITIVE);
     private static final Pattern pattern3 = Pattern.compile("^(\\d{1,1}\\.){1,3}\\d{1,2}\\.?\\s[a-z]{1,2}\\/[a-z]{1,2}.*");
 
@@ -64,46 +64,54 @@ public class DocuMineClassificationService {
             return;
         }
         if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
-                || PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
-                .getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())
-        ) {
+            || PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null
+                                                                                                   || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
             textBlock.setClassification(PageBlockType.HEADER);
 
         } else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER)
-                || PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
-                .getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())
-        ) {
+                   || PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null
+                                                                                                           || textBlock.getHighestFontSize() <= document.getFontSizeCounter()
+                .getMostPopular())) {
             textBlock.setClassification(PageBlockType.FOOTER);
-        } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
-                document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()
-                .size() == 1)) {
+        } else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock, document.getTextHeightCounter().getMostPopular()) > 2.5
+                                                 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks().size() == 1)) {
             if (!Pattern.matches("[0-9]+", textBlock.toString())) {
                 textBlock.setClassification(PageBlockType.TITLE);
             }
-        } else if (textBlock.getText().length() > 5 && (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter()
-                .getMostPopular() || textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular()) && PositionUtils.getApproxLineCount(textBlock) < 5.9
-
-                && (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && !matcher2.matches() && !textBlock.toString()
-                .contains(":") || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && !matcher2.matches() && !textBlock.toString()
-                .contains(":") || textBlock.toString().startsWith("APPENDIX") || textBlock.toString().startsWith("FIGURE") || textBlock.toString()
-                .startsWith("TABLE")) && !textBlock.toString().endsWith(":") && matcher2.find()) {
+        } else if (textBlock.getText().length() > 5
+                   && (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter().getMostPopular()
+                       || textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular())
+                   && PositionUtils.getApproxLineCount(textBlock) < 5.9
+                   && (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && matcher2.reset().find() && !textBlock.toString()
+                .contains(":")
+                       || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && matcher2.reset().find() && !textBlock.toString().contains(":")
+                       || textBlock.toString().startsWith("APPENDIX")
+                       || textBlock.toString().startsWith("FIGURE")
+                       || textBlock.toString().startsWith("TABLE"))
+                   && !textBlock.toString().endsWith(":")
+                   && matcher2.reset().find()) {
             textBlock.setClassification(PageBlockType.getHeadlineType(1));
             document.setHeadlines(true);
 
-        } else if (matcher.find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.find() && !matcher3.matches()) {
+        } else if (matcher.reset().find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.reset().find() && !matcher3.reset().matches()) {
             textBlock.setClassification(PageBlockType.getHeadlineType(2));
             document.setHeadlines(true);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
-                .getMostPopular() && textBlock.getMostPopularWordStyle().equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
+                   && textBlock.getMostPopularWordStyle().equals("bold")
+                   && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_BOLD);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFont()
-                .equals(document.getFontCounter().getMostPopular()) && textBlock.getMostPopularWordStyle()
-                .equals(document.getFontStyleCounter().getMostPopular()) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordFont().equals(document.getFontCounter().getMostPopular())
+                   && textBlock.getMostPopularWordStyle().equals(document.getFontStyleCounter().getMostPopular())
+                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
             textBlock.setClassification(PageBlockType.PARAGRAPH);
-        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
-                .getMostPopular() && textBlock.getMostPopularWordStyle().equals("italic") && !document.getFontStyleCounter()
-                .getMostPopular()
-                .equals("italic") && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
+        } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
+                   && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
+                   && textBlock.getMostPopularWordStyle().equals("italic")
+                   && !document.getFontStyleCounter().getMostPopular().equals("italic")
+                   && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_ITALIC);
         } else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
             textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/DocumentGraphFactory.java

@@ -18,8 +18,6 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationFooter;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationHeader;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
-import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
-import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.DocumentTree;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Footer;
@@ -31,6 +29,8 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Pa
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Paragraph;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Section;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.utils.IdBuilder;
 import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
 
@@ -49,9 +49,14 @@ public class DocumentGraphFactory {
         Document documentGraph = new Document();
         Context context = new Context(documentGraph);
 
-        document.getPages().forEach(context::buildAndAddPageWithCounter);
-        document.getSections().stream().flatMap(section -> section.getImages().stream()).forEach(image -> context.getImages().add(image));
-        addSections(document, context);
+        document.getPages()
+                .forEach(context::buildAndAddPageWithCounter);
+        document.getSections()
+                .stream()
+                .flatMap(section -> section.getImages()
+                        .stream())
+                .forEach(image -> context.getImages().add(image));
+        addSections(document, context, documentGraph);
         addHeaderAndFooterToEachPage(document, context);
 
         documentGraph.setNumberOfPages(context.pages.size());
@@ -62,9 +67,10 @@ public class DocumentGraphFactory {
     }
 
 
-    private void addSections(ClassificationDocument document, Context context) {
+    private void addSections(ClassificationDocument classificationDocument, Context context, Document document) {
 
-        document.getSections().forEach(section -> SectionNodeFactory.addSection(null, section.getNonEmptyPageBlocks(), section.getImages(), context));
+        classificationDocument.getSections()
+                .forEach(section -> SectionNodeFactory.addSection(null, section.getNonEmptyPageBlocks(), section.getImages(), context, document));
     }
 
 
@@ -74,9 +80,11 @@ public class DocumentGraphFactory {
 
         GenericSemanticNode node;
         if (originalTextBlock.isHeadline()) {
-            node = Headline.builder().documentTree(context.getDocumentTree()).build();
+            node = Headline.builder().documentTree(context.getDocumentTree())
+                    .build();
         } else {
-            node = Paragraph.builder().documentTree(context.getDocumentTree()).build();
+            node = Paragraph.builder().documentTree(context.getDocumentTree())
+                    .build();
         }
 
         page.getMainBody().add(node);
@@ -91,7 +99,16 @@ public class DocumentGraphFactory {
     }
 
 
-    public void addImage(Section section, ClassifiedImage image, Context context) {
+    public void addImage(GenericSemanticNode parent, ClassifiedImage image, Context context) {
+
+        Image imageNode = createImage(image, context);
+        List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(parent, imageNode);
+        imageNode.setTreeId(treeId);
+        imageNode.setLeafTextBlock(context.textBlockFactory.emptyTextBlock(parent, context, context.getPage(image.getPage())));
+    }
+
+
+    private Image createImage(ClassifiedImage image, Context context) {
 
         Rectangle2D position = image.getPosition();
         Page page = context.getPage(image.getPage());
@@ -104,9 +121,7 @@ public class DocumentGraphFactory {
                 .documentTree(context.getDocumentTree())
                 .build();
         page.getMainBody().add(imageNode);
-
-        List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(section, imageNode);
-        imageNode.setTreeId(treeId);
+        return imageNode;
     }
 
 
@@ -146,10 +161,7 @@ public class DocumentGraphFactory {
 
         Page page = context.getPage(textBlocks.get(0).getPage());
         Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks),
-                footer,
-                context,
-                page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), footer, context, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
         footer.setTreeId(tocId);
         footer.setLeafTextBlock(textBlock);
@@ -161,7 +173,7 @@ public class DocumentGraphFactory {
 
         Page page = context.getPage(textBlocks.get(0).getPage());
         Header header = Header.builder().documentTree(context.getDocumentTree()).build();
-        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), header, 0, page);
+        AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeTextPositionSequence(textBlocks), header, 0, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
         header.setTreeId(tocId);
         header.setLeafTextBlock(textBlock);
@@ -172,7 +184,8 @@ public class DocumentGraphFactory {
     private void addEmptyFooter(int pageIndex, Context context) {
 
         Page page = context.getPage(pageIndex);
-        Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
+        Footer footer = Footer.builder().documentTree(context.getDocumentTree())
+                .build();
         AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(footer, context, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
         footer.setTreeId(tocId);
@@ -184,7 +197,8 @@ public class DocumentGraphFactory {
     private void addEmptyHeader(int pageIndex, Context context) {
 
         Page page = context.getPage(pageIndex);
-        Header header = Header.builder().documentTree(context.getDocumentTree()).build();
+        Header header = Header.builder().documentTree(context.getDocumentTree())
+                .build();
         AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(header, 0, page);
         List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
         header.setTreeId(tocId);

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/SectionNodeFactory.java

@@ -11,6 +11,7 @@ import java.util.Map;
 import java.util.Set;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
@@ -24,27 +25,46 @@ import lombok.experimental.UtilityClass;
 @UtilityClass
 public class SectionNodeFactory {
 
-    public void addSection(GenericSemanticNode parentNode, List<AbstractPageBlock> pageBlocks, List<ClassifiedImage> images, DocumentGraphFactory.Context context) {
+    public void addSection(GenericSemanticNode parentNode,
+                           List<AbstractPageBlock> pageBlocks,
+                           List<ClassifiedImage> images,
+                           DocumentGraphFactory.Context context,
+                           Document document) {
+
+        // This is for the case where we have images on a page without any text/footer/header.
+        // The pageBlocks list is empty, but we still need to add those images to the document.
+        if (!images.isEmpty() && pageBlocks.isEmpty()) {
+            images.stream()
+                    .distinct()
+                    .forEach(image -> DocumentGraphFactory.addImage(document, image, context));
+            return;
+        }
 
         if (pageBlocks.isEmpty()) {
             return;
         }
-        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream().collect(groupingBy(AbstractPageBlock::getPage));
-        Section section = Section.builder().documentTree(context.getDocumentTree()).build();
+
+        Map<Integer, List<AbstractPageBlock>> blocksPerPage = pageBlocks.stream()
+                .collect(groupingBy(AbstractPageBlock::getPage));
+        Section section = Section.builder().documentTree(context.getDocumentTree())
+                .build();
 
         context.getSections().add(section);
-        blocksPerPage.keySet().forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
+        blocksPerPage.keySet()
+                .forEach(pageNumber -> addSectionNodeToPageNode(context, section, pageNumber));
 
         section.setTreeId(getTreeId(parentNode, context, section));
 
-        addFirstHeadlineDirectlyToSection(pageBlocks, context, section);
+        addFirstHeadlineDirectlyToSection(pageBlocks, context, section, document);
         if (containsTablesAndTextBlocks(pageBlocks)) {
-            splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context));
+            splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context, document));
         } else {
-            addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section);
+            addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section, document);
         }
 
-        images.stream().distinct().forEach(image -> DocumentGraphFactory.addImage(section, image, context));
+        images.stream()
+                .distinct()
+                .forEach(image -> DocumentGraphFactory.addImage(section, image, context));
     }
 
 
@@ -58,16 +78,16 @@ public class SectionNodeFactory {
     }
 
 
-    private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
+    private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section, Document document) {
 
         if (pageBlocks.get(0).isHeadline()) {
-            addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section);
+            addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section, document);
             pageBlocks.remove(0);
         }
     }
 
 
-    private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
+    private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section, Document document) {
 
         Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
         List<AbstractPageBlock> remainingBlocks = new LinkedList<>(pageBlocks);
@@ -86,7 +106,7 @@ public class SectionNodeFactory {
             } else if (abstractPageBlock instanceof TablePageBlock tablePageBlock) {
                 List<TablePageBlock> tablesToMerge = TableMergingUtility.findConsecutiveTablesWithSameColCountAndSameHeaders(tablePageBlock, remainingBlocks);
                 alreadyMerged.addAll(tablesToMerge);
-                TableNodeFactory.addTable(section, tablesToMerge, context);
+                TableNodeFactory.addTable(section, tablesToMerge, context, document);
             } else {
                 throw new RuntimeException(format("Unhandled AbstractPageBlockType %s!", abstractPageBlock.getClass()));
             }
@@ -96,7 +116,9 @@ public class SectionNodeFactory {
 
     private boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {
 
-        return pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
+        return pageBlocks.stream()
+                       .anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream()
+                       .anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
     }
 
 
@@ -112,7 +134,9 @@ public class SectionNodeFactory {
 
         List<List<AbstractPageBlock>> splitList = splitIntoCoherentList(pageBlocks);
         movePrecedingHeadlineToTableList(splitList);
-        return splitList.stream().filter(list -> !list.isEmpty()).toList();
+        return splitList.stream()
+                .filter(list -> !list.isEmpty())
+                .toList();
     }
 
 
@@ -133,7 +157,8 @@ public class SectionNodeFactory {
 
     private boolean listIsTablesOnly(List<AbstractPageBlock> abstractPageBlocks) {
 
-        return abstractPageBlocks.stream().allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
+        return abstractPageBlocks.stream()
+                .allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/factory/TableNodeFactory.java

@@ -8,6 +8,7 @@ import java.util.Set;
 import java.util.stream.Collectors;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
@@ -27,23 +28,26 @@ public class TableNodeFactory {
     public final double TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD = 0.05;
 
 
-    public void addTable(GenericSemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context) {
+    public void addTable(GenericSemanticNode parentNode, List<TablePageBlock> tablesToMerge, DocumentGraphFactory.Context context, Document document) {
 
         setPageNumberInCells(tablesToMerge);
-        Set<Page> pages = tablesToMerge.stream().map(AbstractPageBlock::getPage).map(context::getPage).collect(Collectors.toSet());
-        List<List<Cell>> mergedRows = tablesToMerge.stream().map(TablePageBlock::getRows).flatMap(Collection::stream).toList();
+        Set<Page> pages = tablesToMerge.stream()
+                .map(AbstractPageBlock::getPage)
+                .map(context::getPage)
+                .collect(Collectors.toSet());
+        List<List<Cell>> mergedRows = tablesToMerge.stream()
+                .map(TablePageBlock::getRows)
+                .flatMap(Collection::stream)
+                .toList();
 
-        Table table = Table.builder()
-                .documentTree(context.getDocumentTree())
-                .numberOfCols(mergedRows.isEmpty() ? 0 : mergedRows.get(0).size())
-                .numberOfRows(mergedRows.size())
+        Table table = Table.builder().documentTree(context.getDocumentTree()).numberOfCols(mergedRows.isEmpty() ? 0 : mergedRows.get(0).size()).numberOfRows(mergedRows.size())
                 .build();
 
         pages.forEach(page -> addTableToPage(page, parentNode, table));
 
         List<Integer> treeId = context.getDocumentTree().createNewChildEntryAndReturnId(parentNode, table);
         table.setTreeId(treeId);
-        addTableCells(mergedRows, table, context);
+        addTableCells(mergedRows, table, context, document);
 
         ifTableHasNoHeadersSetFirstRowAsHeaders(table);
     }
@@ -63,7 +67,8 @@ public class TableNodeFactory {
 
     private void setPageNumberInTextBlocksWithPageNumberSetTo0(TablePageBlock table, Cell cell) {
 
-        cell.getTextBlocks().stream()//
+        cell.getTextBlocks()
+                .stream()//
                 .filter(tb -> tb.getPage() == 0)//
                 .forEach(tb -> tb.setPage(table.getPage()));
     }
@@ -82,28 +87,32 @@ public class TableNodeFactory {
 
     private void ifTableHasNoHeadersSetFirstRowAsHeaders(Table table) {
 
-        if (table.streamHeaders().findAny().isEmpty()) {
-            table.streamRow(0).forEach(tableCellNode -> tableCellNode.setHeader(true));
+        if (table.streamHeaders()
+                .findAny().isEmpty()) {
+            table.streamRow(0)
+                    .forEach(tableCellNode -> tableCellNode.setHeader(true));
         }
     }
 
 
-    private void addTableCells(List<List<Cell>> rows, Table table, DocumentGraphFactory.Context context) {
+    private void addTableCells(List<List<Cell>> rows, Table table, DocumentGraphFactory.Context context, Document document) {
 
         for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
             for (int colIndex = 0; colIndex < rows.get(rowIndex).size(); colIndex++) {
-                addTableCell(rows.get(rowIndex).get(colIndex), rowIndex, colIndex, table, context);
+                addTableCell(rows.get(rowIndex)
+                                     .get(colIndex), rowIndex, colIndex, table, context, document);
             }
         }
     }
 
 
     @SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
-    private void addTableCell(Cell cell, int rowIndex, int colIndex, Table tableNode, DocumentGraphFactory.Context context) {
+    private void addTableCell(Cell cell, int rowIndex, int colIndex, Table tableNode, DocumentGraphFactory.Context context, Document document) {
 
         Page page = context.getPage(cell.getPageNumber());
 
-        TableCell tableCell = TableCell.builder().documentTree(context.getDocumentTree()).row(rowIndex).col(colIndex).header(cell.isHeaderCell()).bBox(cell.getBounds2D()).build();
+        TableCell tableCell = TableCell.builder().documentTree(context.getDocumentTree()).row(rowIndex).col(colIndex).header(cell.isHeaderCell()).bBox(cell.getBounds2D())
+                .build();
         page.getMainBody().add(tableCell);
 
         List<Integer> treeId = context.getDocumentTree().createNewTableChildEntryAndReturnId(tableNode, tableCell);
@@ -113,16 +122,26 @@ public class TableNodeFactory {
         if (cell.getTextBlocks().isEmpty()) {
             tableCell.setLeafTextBlock(context.getTextBlockFactory().emptyTextBlock(tableNode, context, page));
         } else if (cell.getTextBlocks().size() == 1) {
-            textBlock = context.getTextBlockFactory().buildAtomicTextBlock(cell.getTextBlocks().get(0).getSequences(), tableCell, context, page);
+            textBlock = context.getTextBlockFactory()
+                    .buildAtomicTextBlock(cell.getTextBlocks()
+                                                  .get(0).getSequences(), tableCell, context, page);
             tableCell.setLeafTextBlock(textBlock);
         } else if (firstTextBlockIsHeadline(cell)) {
-            SectionNodeFactory.addSection(tableCell, cell.getTextBlocks().stream().map(tb -> (AbstractPageBlock) tb).toList(), emptyList(), context);
+            SectionNodeFactory.addSection(tableCell,
+                                          cell.getTextBlocks()
+                                                  .stream()
+                                                  .map(tb -> (AbstractPageBlock) tb)
+                                                  .toList(),
+                                          emptyList(),
+                                          context,
+                                          document);
         } else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
             List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(cell.getTextBlocks());
             textBlock = context.getTextBlockFactory().buildAtomicTextBlock(sequences, tableCell, context, page);
             tableCell.setLeafTextBlock(textBlock);
         } else {
-            cell.getTextBlocks().forEach(tb -> DocumentGraphFactory.addParagraphOrHeadline(tableCell, tb, context, emptyList()));
+            cell.getTextBlocks()
+                    .forEach(tb -> DocumentGraphFactory.addParagraphOrHeadline(tableCell, tb, context, emptyList()));
         }
     }
 
@@ -135,7 +154,8 @@ public class TableNodeFactory {
 
     private boolean firstTextBlockIsHeadline(Cell cell) {
 
-        return cell.getTextBlocks().get(0).isHeadline();
+        return cell.getTextBlocks()
+                .get(0).isHeadline();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/services/visualization/LayoutGridService.java

@@ -110,11 +110,13 @@ public class LayoutGridService {
             return;
         }
         for (Page page : table.getPages()) {
+
             Optional<Integer> optionalFirstRowOnPage = table.streamCol(0).filter(tableCell -> tableCell.isOnPage(page.getNumber())).map(TableCell::getRow).findFirst();
             if (optionalFirstRowOnPage.isEmpty()) {
                 continue;
             }
             int firstRowOnPage = optionalFirstRowOnPage.get();
+
             Stream<Double> xStream = switch (page.getRotation()) {
                 case 90 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinX);
                 case 180 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMaxX);
@@ -123,6 +125,7 @@ public class LayoutGridService {
             };
             List<Double> xs = xStream.collect(Collectors.toList());
             xs.remove(0);
+
             Stream<Double> yStream = switch (page.getRotation()) {
                 case 90 -> streamBBoxOfCellsOnPage(table.streamRow(firstRowOnPage), page).map(RectangularShape::getMinY);
                 case 180 -> streamBBoxOfCellsOnPage(table.streamCol(0), page).map(RectangularShape::getMinY);
@@ -132,7 +135,7 @@ public class LayoutGridService {
             List<Double> ys = yStream.collect(Collectors.toList());
             ys.remove(0);
 
-            Rectangle2D tableBBox = table.getBBox().get(table.getFirstPage());
+            Rectangle2D tableBBox = table.getBBox().get(page);
             List<ColoredLine> coloredLines = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getColoredLines();
             xs.forEach(x -> {
                 Line2D line = new Line2D.Double(new Point2D.Double(x, tableBBox.getMaxY()), new Point2D.Double(x, tableBBox.getMinY()));

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/utils/RectangleTransformations.java

@@ -3,8 +3,10 @@ package com.knecon.fforesight.service.layoutparser.processor.utils;
 import static java.lang.String.format;
 
 import java.awt.geom.Area;
+import java.awt.geom.Point2D;
 import java.awt.geom.Rectangle2D;
 import java.awt.geom.RectangularShape;
+import java.util.ArrayList;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
@@ -20,6 +22,8 @@ import org.apache.pdfbox.pdmodel.common.PDRectangle;
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Point;
 import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 
 import lombok.AllArgsConstructor;
 import lombok.NoArgsConstructor;
@@ -149,7 +153,27 @@ public class RectangleTransformations {
                 previousRectangle = currentRectangle;
             }
         }
-        return rectangleListsWithGaps.stream().map(RectangleTransformations::rectangle2DBBox).toList();
+        return rectangleListsWithGaps.stream()
+                .map(RectangleTransformations::rectangle2DBBox)
+                .toList();
+    }
+
+
+    public static CleanRulings extractRulings(List<? extends Rectangle2D.Float> rectangles) {
+
+        // Underlined or strikethrough are also in rulings but we dont want to split blocks with them so we use cells.
+        List<Ruling> horizontalRulings = new ArrayList<>();
+        List<Ruling> verticalRulings = new ArrayList<>();
+
+        rectangles.forEach(rectangle -> {
+            horizontalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y), new Point2D.Float(rectangle.x + rectangle.width, rectangle.y)));
+            horizontalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y + rectangle.height),
+                                             new Point2D.Float(rectangle.x + rectangle.width, rectangle.y + rectangle.height)));
+            verticalRulings.add(new Ruling(new Point2D.Float(rectangle.x, rectangle.y), new Point2D.Float(rectangle.x, rectangle.y + rectangle.height)));
+            verticalRulings.add(new Ruling(new Point2D.Float(rectangle.x + rectangle.width, rectangle.y),
+                                           new Point2D.Float(rectangle.x + rectangle.width, rectangle.y + rectangle.height)));
+        });
+        return CleanRulings.builder().vertical(verticalRulings).horizontal(horizontalRulings).build();
     }
 
 

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/utils/TextPositionOperations.java

@@ -23,4 +23,10 @@ public class TextPositionOperations {
         return sequence;
     }
 
+
+    public static List<TextPositionSequence> mergeTextPositionSequence(List<TextPageBlock> textBlocks) {
+
+        return textBlocks.stream().flatMap(tb -> tb.getSequences().stream()).collect(Collectors.toList());
+    }
+
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/utils/TextPositionSequenceComparator.java

@@ -28,15 +28,13 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPosit
  *
  * @author Ben Litchfield
  */
-public class TextPositionSequenceComparator implements Comparator<TextPositionSequence>
-{
+public class TextPositionSequenceComparator implements Comparator<TextPositionSequence> {
+
     @Override
-    public int compare(TextPositionSequence pos1, TextPositionSequence pos2)
-    {
+    public int compare(TextPositionSequence pos1, TextPositionSequence pos2) {
         // only compare text that is in the same direction
         int cmp1 = Float.compare(pos1.getDir().getDegrees(), pos2.getDir().getDegrees());
-        if (cmp1 != 0)
-        {
+        if (cmp1 != 0) {
             return cmp1;
         }
 
@@ -48,25 +46,19 @@ public class TextPositionSequenceComparator implements Comparator<TextPositionSe
         float pos2YBottom = pos2.getMaxYDirAdj();
 
         // note that the coordinates have been adjusted so 0,0 is in upper left
-        float pos1YTop = pos1YBottom - pos1.getTextHeight();
-        float pos2YTop = pos2YBottom - pos2.getTextHeight();
+        float pos1YTop = pos1YBottom - pos1.getTextHeightNoPadding();
+        float pos2YTop = pos2YBottom - pos2.getTextHeightNoPadding();
 
         float yDifference = Math.abs(pos1YBottom - pos2YBottom);
 
         // we will do a simple tolerance comparison
-        if (yDifference < .1 ||
-                pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom ||
-                pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom)
-        {
+        if (yDifference < .1 || pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom || pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom) {
             return Float.compare(x1, x2);
-        }
-        else if (pos1YBottom < pos2YBottom)
-        {
+        } else if (pos1YBottom < pos2YBottom) {
             return -1;
-        }
-        else
-        {
+        } else {
             return 1;
         }
     }
+
 }

layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/LayoutparserEnd2EndTest.java

@@ -28,7 +28,20 @@ public class LayoutparserEnd2EndTest extends AbstractTest {
         prepareStorage("files/bdr/Wie weiter bei Kristeneinrichtungen.pdf");
         LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(LayoutParsingType.REDACT_MANAGER);
         LayoutParsingFinishedEvent finishedEvent = layoutParsingPipeline.parseLayoutAndSaveFilesToStorage(layoutParsingRequest);
-        Arrays.stream(finishedEvent.message().split("\n")).forEach(log::info);
+        Arrays.stream(finishedEvent.message().split("\n"))
+                .forEach(log::info);
+    }
+
+
+    @Test
+    @SneakyThrows
+    public void testLayoutParserEndToEnd_RED_8747() {
+
+        prepareStorage("files/SinglePages/MergedEntities.pdf");
+        LayoutParsingRequest layoutParsingRequest = buildDefaultLayoutParsingRequest(LayoutParsingType.REDACT_MANAGER);
+        LayoutParsingFinishedEvent finishedEvent = layoutParsingPipeline.parseLayoutAndSaveFilesToStorage(layoutParsingRequest);
+        Arrays.stream(finishedEvent.message().split("\n"))
+                .forEach(log::info);
     }
 
 }

layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/segmentation/PdfSegmentationServiceTest.java

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.server.segmentation;
 
 import static org.assertj.core.api.Assertions.assertThat;
+import static org.junit.jupiter.api.Assertions.assertTrue;
 
 import java.awt.geom.Rectangle2D;
 import java.io.File;
@@ -25,14 +26,20 @@ import com.fasterxml.jackson.databind.ObjectMapper;
 import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
 import com.knecon.fforesight.service.layoutparser.processor.LayoutParsingPipeline;
 import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
+import com.knecon.fforesight.service.layoutparser.processor.model.PageContents;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.ImageType;
+import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.image.ClassifiedImage;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
+import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.image.ImageServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
+import com.knecon.fforesight.service.layoutparser.processor.services.PageContentExtractor;
 import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
 import com.knecon.fforesight.service.layoutparser.processor.services.classification.RedactManagerClassificationService;
+import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.server.utils.AbstractTest;
 
 import lombok.SneakyThrows;
@@ -752,6 +759,54 @@ public class PdfSegmentationServiceTest extends AbstractTest {
     }
 
 
+    @Test
+    public void testMergedEntities_Page26() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/MergedEntities.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 1);
+
+        validateTable(document, 0, 6, 6, 5, 0);
+
+    }
+
+
+    @Test
+    public void testHeaderAndFooter() throws IOException {
+
+        String fileName = "files/SinglePages/Page1_54 Fludioxonil - EU AIR3 - Document E1 - Listing of Community and Member States MRLs.pdf";
+        String textToSearch = "Annex to Regulation 283/2013 Annex to Regulation 284/2013";
+        ClassPathResource pdfFileResource = new ClassPathResource(fileName);
+
+        List<PageContents> textPositionPerPage = PageContentExtractor.getSortedPageContents(fileName);
+        var textPositions = textPositionPerPage.stream()
+                .flatMap(t -> t.getSortedTextPositionSequences()
+                        .stream()
+                        .map(TextPositionSequence::toString))
+                .collect(Collectors.joining(" "));
+        assertThat(textPositions.contains(textToSearch)).isFalse();
+
+        ClassificationDocument classificationDocument = buildClassificationDocument(pdfFileResource.getFile());
+
+        assertThat(classificationDocument.getHeaders()
+                           .get(0).getTextBlocks().size()).isEqualTo(3);
+        assertThat(classificationDocument.getHeaders()
+                           .get(0).getTextBlocks()
+                           .get(0).getSequences().size()).isEqualTo(8);
+        assertThat(classificationDocument.getHeaders()
+                           .get(0).getTextBlocks()
+                           .get(0).toString()).isEqualTo(textToSearch);
+
+        Document document = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
+
+        TextBlock leafTextBlock = document.getFirstPage().getHeader().getLeafTextBlock();
+        assertTrue(leafTextBlock.getSearchText().contains(textToSearch));
+
+    }
+
+
     @SneakyThrows
     private void toHtml(ClassificationDocument document, String filename) {
 

layoutparser-service/viewer-doc-processor/src/main/java/com/knecon/fforesight/service/viewerdoc/service/ViewerDocumentService.java

@@ -17,6 +17,7 @@ import java.util.stream.Collectors;
 
 import org.apache.pdfbox.Loader;
 import org.apache.pdfbox.cos.COSName;
+import org.apache.pdfbox.pdfwriter.compress.CompressParameters;
 import org.apache.pdfbox.pdmodel.PDDocument;
 import org.apache.pdfbox.pdmodel.PDDocumentCatalog;
 import org.apache.pdfbox.pdmodel.PDPage;
@@ -126,8 +127,8 @@ public class ViewerDocumentService {
                 pdDocument = openPDDocument(tmpFile.toFile());
             }
         }
-        observedIncrementalSave(pdDocument, destinationFile);
 
+        observedIncrementalSave(pdDocument, destinationFile);
         pdDocument.close();
         assert tmpFile.toFile().delete();
     }
@@ -282,10 +283,12 @@ public class ViewerDocumentService {
 
     @SneakyThrows
     private void observedIncrementalSave(PDDocument pdDocument, File outputFile) {
-
+    /*
+    Sometimes the viewer document is corrupted after saving and missing the content streams on a random page, for the files we viewed it did not seem to happen with incrementalSave. It might only be a timing issue though
+     */
         Observation.createNotStarted("ViewerDocumentService", registry).contextualName("incremental-save").observe(() -> {
             try (var out = new FileOutputStream(outputFile)) {
-                pdDocument.save(out);
+                pdDocument.save(out, CompressParameters.NO_COMPRESSION);
             } catch (IOException e) {
                 throw new RuntimeException(e);
             }