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;
@@ -94,16 +95,23 @@ 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().isPresent()) {
-            imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId().get());
+        if (layoutParsingRequest.imagesFileStorageId()
+                .isPresent()) {
+            imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId()
+                                                                                     .get());
         }
 
         TableServiceResponse tableServiceResponse = new TableServiceResponse();
-        if (layoutParsingRequest.tablesFileStorageId().isPresent()) {
-            tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId().get());
+        if (layoutParsingRequest.tablesFileStorageId()
+                .isPresent()) {
+            tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId()
+                                                                                     .get());
         }
 
         ClassificationDocument classificationDocument = parseLayout(layoutParsingRequest.layoutParsingType(),
@@ -142,25 +150,25 @@ public class LayoutParsingPipeline {
                 .numberOfPages(documentGraph.getNumberOfPages())
                 .duration(System.currentTimeMillis() - start)
                 .message(format("""
-                                Layout parsing has finished in %.02f s.
-                                identifiers: %s
-                                %s
-                                Files have been saved with Ids:
-                                Structure: %s
-                                Text: %s
-                                Positions: %s
-                                PageData: %s
-                                Simplified Text: %s
-                                Viewer Doc: %s""",
-                        ((float) (System.currentTimeMillis() - start)) / 1000,
-                        layoutParsingRequest.identifier(),
-                        buildSemanticNodeCountMessage(documentGraph.getNumberOfPages(), documentGraph.buildSemanticNodeCounts()),
-                        layoutParsingRequest.structureFileStorageId(),
-                        layoutParsingRequest.textBlockFileStorageId(),
-                        layoutParsingRequest.positionBlockFileStorageId(),
-                        layoutParsingRequest.pageFileStorageId(),
-                        layoutParsingRequest.simplifiedTextStorageId(),
-                        layoutParsingRequest.viewerDocumentStorageId()))
+                                        Layout parsing has finished in %.02f s.
+                                        identifiers: %s
+                                        %s
+                                        Files have been saved with Ids:
+                                        Structure: %s
+                                        Text: %s
+                                        Positions: %s
+                                        PageData: %s
+                                        Simplified Text: %s
+                                        Viewer Doc: %s""",
+                                ((float) (System.currentTimeMillis() - start)) / 1000,
+                                layoutParsingRequest.identifier(),
+                                buildSemanticNodeCountMessage(documentGraph.getNumberOfPages(), documentGraph.buildSemanticNodeCounts()),
+                                layoutParsingRequest.structureFileStorageId(),
+                                layoutParsingRequest.textBlockFileStorageId(),
+                                layoutParsingRequest.positionBlockFileStorageId(),
+                                layoutParsingRequest.pageFileStorageId(),
+                                layoutParsingRequest.simplifiedTextStorageId(),
+                                layoutParsingRequest.viewerDocumentStorageId()))
                 .build();
 
     }
@@ -170,9 +178,9 @@ public class LayoutParsingPipeline {
 
         AtomicReference<Document> documentReference = new AtomicReference<>();
 
-        Observation.createNotStarted("LayoutParsingPipeline", observationRegistry).contextualName("build-document-graph").observe(() -> {
-            documentReference.set(DocumentGraphFactory.buildDocumentGraph(classificationDocument));
-        });
+        Observation.createNotStarted("LayoutParsingPipeline", observationRegistry)
+                .contextualName("build-document-graph")
+                .observe(() -> documentReference.set(DocumentGraphFactory.buildDocumentGraph(classificationDocument)));
 
         return documentReference.get();
     }
@@ -181,14 +189,14 @@ public class LayoutParsingPipeline {
     private String buildSemanticNodeCountMessage(int numberOfPages, Map<NodeType, Long> semanticNodeCounts) {
 
         return String.format("%d pages with %d sections, %d headlines, %d paragraphs, %d tables with %d cells, %d headers, and %d footers parsed",
-                numberOfPages,
-                semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
-                semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
-                semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
-                semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
-                semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
-                semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
-                semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
+                             numberOfPages,
+                             semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
+                             semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
+                             semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
+                             semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
+                             semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
+                             semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
+                             semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
     }
 
 
@@ -243,8 +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());
             };
@@ -319,9 +329,7 @@ public class LayoutParsingPipeline {
 
     private void increaseDocumentStatistics(ClassificationPage classificationPage, ClassificationDocument document) {
 
-        if (!classificationPage.isLandscape()) {
-            document.getFontSizeCounter().addAll(classificationPage.getFontSizeCounter().getCountPerValue());
-        }
+        document.getFontSizeCounter().addAll(classificationPage.getFontSizeCounter().getCountPerValue());
         document.getFontCounter().addAll(classificationPage.getFontCounter().getCountPerValue());
         document.getTextHeightCounter().addAll(classificationPage.getTextHeightCounter().getCountPerValue());
         document.getFontStyleCounter().addAll(classificationPage.getFontStyleCounter().getCountPerValue());

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

@@ -6,12 +6,14 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageB
 
 import lombok.AllArgsConstructor;
 import lombok.Data;
+import lombok.EqualsAndHashCode;
 import lombok.NoArgsConstructor;
 
 @Data
 @AllArgsConstructor
 @NoArgsConstructor
-public abstract class AbstractPageBlock {
+@EqualsAndHashCode(callSuper = true)
+public abstract class AbstractPageBlock extends Rectangle {
 
     @JsonIgnore
     protected float minX;

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/graph/nodes/TableCell.java

@@ -84,14 +84,16 @@ public class TableCell implements GenericSemanticNode {
 
     private TextBlock buildTextBlock() {
 
-        return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
+        return streamAllSubNodes().filter(SemanticNode::isLeaf)
+                .map(SemanticNode::getLeafTextBlock)
+                .collect(new TextBlockCollector());
     }
 
 
     @Override
     public String toString() {
 
-        return treeId + ": " + NodeType.TABLE_CELL + ": " + this.buildTextBlock().buildSummary();
+        return treeId + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
     }
 
 }

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/Cell.java

@@ -1,6 +1,7 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.table;
 
 import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.Iterator;
 import java.util.List;
@@ -36,6 +37,12 @@ public class Cell extends Rectangle {
     }
 
 
+    public Cell(Rectangle2D r) {
+
+        super((float) r.getY(), (float) r.getX(), (float) r.getWidth(), (float) r.getHeight());
+    }
+
+
     public void addTextBlock(TextPageBlock textBlock) {
 
         textBlocks.add(textBlock);

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/Ruling.java

@@ -20,7 +20,8 @@ import lombok.extern.slf4j.Slf4j;
 @SuppressWarnings("all")
 public class Ruling extends Line2D.Float {
 
-    private static int PERPENDICULAR_PIXEL_EXPAND_AMOUNT = 2;
+    public static final int PERPENDICULAR_UNIT_EXPAND_AMOUNT = 2;
+    public static final int COLINEAR_OR_PARALLEL_UNIT_EXPAND_AMOUNT = 2;
 
 
     public Ruling(Point2D p1, Point2D p2) {
@@ -110,8 +111,8 @@ public class Ruling extends Line2D.Float {
         });
 
         for (Ruling h : horizontals) {
-            sos.add(new SortObject(SOType.HLEFT, h.getLeft() - PERPENDICULAR_PIXEL_EXPAND_AMOUNT, h));
-            sos.add(new SortObject(SOType.HRIGHT, h.getRight() + PERPENDICULAR_PIXEL_EXPAND_AMOUNT, h));
+            sos.add(new SortObject(SOType.HLEFT, h.getLeft() - PERPENDICULAR_UNIT_EXPAND_AMOUNT, h));
+            sos.add(new SortObject(SOType.HRIGHT, h.getRight() + PERPENDICULAR_UNIT_EXPAND_AMOUNT, h));
         }
 
         for (Ruling v : verticals) {
@@ -151,7 +152,7 @@ public class Ruling extends Line2D.Float {
                             if (i == null) {
                                 continue;
                             }
-                            rv.put(i, new Ruling[]{h.getKey().expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT), so.ruling.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT)});
+                            rv.put(i, new Ruling[]{h.getKey().expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT), so.ruling.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT)});
                         } catch (UnsupportedOperationException e) {
                             log.info("Some line are oblique, ignoring...");
                             continue;
@@ -267,7 +268,7 @@ public class Ruling extends Line2D.Float {
     }
 
 
-    public boolean nearlyIntersects(Ruling another, int colinearOrParallelExpandAmount) {
+    public boolean nearlyIntersects(Ruling another) {
 
         if (this.intersectsLine(another)) {
             return true;
@@ -276,9 +277,9 @@ public class Ruling extends Line2D.Float {
         boolean rv = false;
 
         if (this.perpendicularTo(another)) {
-            rv = this.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT).intersectsLine(another);
+            rv = this.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT).intersectsLine(another);
         } else {
-            rv = this.expand(colinearOrParallelExpandAmount).intersectsLine(another.expand(colinearOrParallelExpandAmount));
+            rv = this.expand(COLINEAR_OR_PARALLEL_UNIT_EXPAND_AMOUNT).intersectsLine(another.expand(COLINEAR_OR_PARALLEL_UNIT_EXPAND_AMOUNT));
         }
 
         return rv;
@@ -319,8 +320,8 @@ public class Ruling extends Line2D.Float {
 
     public Point2D intersectionPoint(Ruling other) {
 
-        Ruling this_l = this.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT);
-        Ruling other_l = other.expand(PERPENDICULAR_PIXEL_EXPAND_AMOUNT);
+        Ruling this_l = this.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT);
+        Ruling other_l = other.expand(PERPENDICULAR_UNIT_EXPAND_AMOUNT);
         Ruling horizontal, vertical;
 
         if (!this_l.intersectsLine(other_l)) {

layoutparser-service/layoutparser-service-processor/src/main/java/com/knecon/fforesight/service/layoutparser/processor/model/table/TablePageBlock.java

@@ -1,18 +1,18 @@
 package com.knecon.fforesight.service.layoutparser.processor.model.table;
 
 import java.awt.geom.Point2D;
-import java.awt.geom.Rectangle2D;
 import java.util.ArrayList;
 import java.util.Collections;
+import java.util.Comparator;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Set;
 import java.util.TreeMap;
-import java.util.stream.Collectors;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
 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.RectangleTransformations;
 
 import lombok.Getter;
 import lombok.Setter;
@@ -21,7 +21,8 @@ import lombok.extern.slf4j.Slf4j;
 @Slf4j
 public class TablePageBlock extends AbstractPageBlock {
 
-    private final TreeMap<CellPosition, Cell> cells = new TreeMap<>();
+    public static final double CELL_AREA_CONTAINED_THRESHOLD = 0.98;
+    private final TreeMap<CellPosition, Cell> cellTreeMap = new TreeMap<>();
 
     private final int rotation;
     @Getter
@@ -30,10 +31,14 @@ public class TablePageBlock extends AbstractPageBlock {
     private int unrotatedRowCount;
     private int unrotatedColCount;
     private List<List<Cell>> rows;
+    @Getter
+    @Setter
+    private List<Cell> cells;
 
 
     public TablePageBlock(List<Cell> cells, Rectangle area, int rotation) {
 
+        this.cells = cells;
         addCells(cells);
         minX = area.getLeft();
         minY = area.getBottom();
@@ -50,6 +55,7 @@ public class TablePageBlock extends AbstractPageBlock {
         return getColCount() == 0 || getRowCount() == 0;
     }
 
+
     public List<List<Cell>> getRows() {
 
         if (rows == null) {
@@ -80,14 +86,17 @@ public class TablePageBlock extends AbstractPageBlock {
 
     public int getColCount() {
 
-        return getRows().stream().mapToInt(List::size).max().orElse(0);
+        return getRows().stream()
+                .mapToInt(List::size)
+                .max()
+                .orElse(0);
 
     }
 
 
     /**
      * Detect header cells (either first row or first column):
-     * Column is marked as header if cell text is bold and row cell text is not bold.
+     * Column is marked as header if originalCell text is bold and row originalCell text is not bold.
      * Defaults to row.
      */
     private void computeHeaders() {
@@ -95,7 +104,7 @@ public class TablePageBlock extends AbstractPageBlock {
         if (rows == null) {
             rows = computeRows();
         }
-        // A bold cell is a header cell as long as every cell to the left/top is bold, too
+        // A bold originalCell is a header originalCell as long as every originalCell to the left/top is bold, too
         // we move from left to right and top to bottom
         for (int rowIndex = 0; rowIndex < rows.size(); rowIndex++) {
             List<Cell> rowCells = rows.get(rowIndex);
@@ -120,7 +129,8 @@ public class TablePageBlock extends AbstractPageBlock {
                 List<Cell> cellsToTheTop = new ArrayList<>();
                 for (int i = 0; i < rowIndex; i++) {
                     try {
-                        cellsToTheTop.add(rows.get(i).get(colIndex));
+                        cellsToTheTop.add(rows.get(i)
+                                                  .get(colIndex));
                     } catch (IndexOutOfBoundsException e) {
                         log.debug("No cell {} in row {}, ignoring.", colIndex, rowIndex);
                     }
@@ -135,7 +145,8 @@ public class TablePageBlock extends AbstractPageBlock {
                 if (lastHeaderCell != null) {
                     cell.getHeaderCells().add(lastHeaderCell);
                 }
-                if (!cell.getTextBlocks().isEmpty() && cell.getTextBlocks().get(0).getMostPopularWordStyle().equals("bold")) {
+                if (!cell.getTextBlocks().isEmpty() && cell.getTextBlocks()
+                        .get(0).getMostPopularWordStyle().equals("bold")) {
                     cell.setHeaderCell(true);
                 }
             }
@@ -151,7 +162,7 @@ public class TablePageBlock extends AbstractPageBlock {
             for (int i = 0; i < unrotatedColCount; i++) { // rows
                 List<Cell> lastRow = new ArrayList<>();
                 for (int j = unrotatedRowCount - 1; j >= 0; j--) { // cols
-                    Cell cell = cells.get(new CellPosition(j, i));
+                    Cell cell = cellTreeMap.get(new CellPosition(j, i));
                     if (cell != null) {
                         lastRow.add(cell);
                     }
@@ -162,7 +173,7 @@ public class TablePageBlock extends AbstractPageBlock {
             for (int i = unrotatedColCount - 1; i >= 0; i--) { // rows
                 List<Cell> lastRow = new ArrayList<>();
                 for (int j = 0; j < unrotatedRowCount; j++) { // cols
-                    Cell cell = cells.get(new CellPosition(j, i));
+                    Cell cell = cellTreeMap.get(new CellPosition(j, i));
                     if (cell != null) {
                         lastRow.add(cell);
                     }
@@ -173,7 +184,7 @@ public class TablePageBlock extends AbstractPageBlock {
             for (int i = 0; i < unrotatedRowCount; i++) {
                 List<Cell> lastRow = new ArrayList<>();
                 for (int j = 0; j < unrotatedColCount; j++) {
-                    Cell cell = cells.get(new CellPosition(i, j)); // JAVA_8 use getOrDefault()
+                    Cell cell = cellTreeMap.get(new CellPosition(i, j)); // JAVA_8 use getOrDefault()
                     if (cell != null) {
                         lastRow.add(cell);
                     }
@@ -187,17 +198,6 @@ public class TablePageBlock extends AbstractPageBlock {
     }
 
 
-    private void add(Cell chunk, int row, int col) {
-
-        unrotatedRowCount = Math.max(unrotatedRowCount, row + 1);
-        unrotatedColCount = Math.max(unrotatedColCount, col + 1);
-
-        CellPosition cp = new CellPosition(row, col);
-        cells.put(cp, chunk);
-
-    }
-
-
     private void addCells(List<Cell> cells) {
 
         if (cells.isEmpty()) {
@@ -206,11 +206,12 @@ public class TablePageBlock extends AbstractPageBlock {
 
         cells.removeIf(cell -> cell.getWidth() < 1.1 || cell.getHeight() < 1.1);
 
-        List<List<Cell>> rowsOfCells = calculateStructure(cells);
+        List<List<Cell>> rowsOfCellsMatrix = calculateTableStructure(cells);
 
-        for (int i = 0; i < rowsOfCells.size(); i++) {
-            for (int j = 0; j < rowsOfCells.get(i).size(); j++) {
-                add(rowsOfCells.get(i).get(j), i, j);
+        for (int i = 0; i < rowsOfCellsMatrix.size(); i++) {
+            for (int j = 0; j < rowsOfCellsMatrix.get(i).size(); j++) {
+                addCellToRowAndCol(rowsOfCellsMatrix.get(i)
+                                           .get(j), i, j);
             }
         }
 
@@ -221,29 +222,36 @@ public class TablePageBlock extends AbstractPageBlock {
      * Calculates the structure of the table. For spanning rows and columns multiple cells with the same values will be inserted.
      *
      * @param cells The found cells
-     * @return TablePageBlock Structure
+     * @return TablePageBlock Structure as a rows of cells matrix
      */
-    private List<List<Cell>> calculateStructure(List<Cell> cells) {
-
-        List<List<Cell>> matrix = new ArrayList<>();
+    private List<List<Cell>> calculateTableStructure(List<Cell> cells) {
 
         if (cells.isEmpty()) {
-            return matrix;
+            return new ArrayList<>();
         }
 
         Set<Float> uniqueX = new HashSet<>();
         Set<Float> uniqueY = new HashSet<>();
-        cells.stream().filter(c -> !c.getTextBlocks().isEmpty() || c.getHeight() > 3 && c.getWidth() > 3).forEach(c -> {
-            uniqueX.add(c.getLeft());
-            uniqueX.add(c.getRight());
-            uniqueY.add(c.getBottom());
-            uniqueY.add(c.getTop());
-        });
+        cells.stream()
+                .filter(c -> !c.getTextBlocks().isEmpty() || c.getHeight() > 3 && c.getWidth() > 3)
+                .forEach(c -> {
+                    uniqueX.add(c.getLeft());
+                    uniqueX.add(c.getRight());
+                    uniqueY.add(c.getBottom());
+                    uniqueY.add(c.getTop());
+                });
 
-        var sortedUniqueX = uniqueX.stream().sorted().toList();
-        var sortedUniqueY = uniqueY.stream().sorted().toList();
+        var sortedUniqueX = uniqueX.stream()
+                .sorted()
+                .toList();
+        var sortedUniqueY = uniqueY.stream()
+                .sorted()
+                .toList();
+
+        List<List<Cell>> rowsOfCells = new ArrayList<>();
 
         Float prevY = null;
+
         for (Float y : sortedUniqueY) {
 
             List<Cell> row = new ArrayList<>();
@@ -252,44 +260,87 @@ public class TablePageBlock extends AbstractPageBlock {
             for (Float x : sortedUniqueX) {
 
                 if (prevY != null && prevX != null) {
-                    var cell = new Cell(new Point2D.Float(prevX, prevY), new Point2D.Float(x, y));
+                    var cellFromGridStructure = new Cell(new Point2D.Float(prevX, prevY), new Point2D.Float(x, y));
 
-                    var intersectionCell = cells.stream().filter(c -> intersects(cell, c)).findFirst();
+                    if (cellFromGridStructure.hasMinimumSize()) {
 
-                    intersectionCell.ifPresent(value -> cell.getTextBlocks().addAll(value.getTextBlocks()));
-                    if (cell.hasMinimumSize()) {
-                        row.add(cell);
+                        cells.stream()
+                                .map(originalCell -> new CellWithIntersection(originalCell, RectangleTransformations.calculateIntersectedArea(cellFromGridStructure, originalCell)))
+                                .filter(cellWithIntersection -> cellWithIntersection.intersectedArea > 0)
+                                .filter(cellWithIntersection -> cellWithIntersection.originalCell.getArea() > cellWithIntersection.intersectedArea * CELL_AREA_CONTAINED_THRESHOLD)
+                                .max(Comparator.comparing(CellWithIntersection::intersectedArea))
+                                .map(CellWithIntersection::originalCell)
+                                .ifPresent(matchingCell -> cellFromGridStructure.getTextBlocks().addAll(matchingCell.getTextBlocks()));
+
+                        row.add(cellFromGridStructure);
                     }
                 }
                 prevX = x;
             }
 
-            if (prevY != null && prevX != null && !row.isEmpty()) {
-                matrix.add(row);
+            // exclude empty rows and rows where all text blocks are empty
+            if (prevY != null && prevX != null && !row.isEmpty() && !row.stream()
+                    .allMatch(cell -> cell.getTextBlocks().isEmpty())) {
+
+                rowsOfCells.add(row);
             }
             prevY = y;
         }
 
-        Collections.reverse(matrix);
+        Collections.reverse(rowsOfCells);
 
-        return matrix;
-    }
-
-
-
-    public boolean intersects(Cell cell1, Cell cell2) {
-        if (cell1.getHeight() <= 0 || cell2.getHeight() <= 0) {
-            return false;
+        // now cells are removed which are part of a column without any text blocks
+        // this is done by first computing the inverse matrix which contains call columns of cells
+        // then the column indices that have to be removed are determined
+        List<List<Cell>> columnsOfCells = new ArrayList<>();
+        int maxRowLength = rowsOfCells.stream()
+                .map(List::size)
+                .max(java.util.Comparator.naturalOrder())
+                .orElse(0);
+        for (int i = 0; i < maxRowLength; i++) {
+            columnsOfCells.add(new ArrayList<>());
         }
-        double x0 = cell1.getX() + 2;
-        double y0 = cell1.getY() + 2;
-        return (cell2.x + cell2.width > x0 &&
-                cell2.y + cell2.height > y0 &&
-                cell2.x < x0 + cell1.getWidth() -2 &&
-                cell2.y < y0 + cell1.getHeight() -2);
+
+        for (List<Cell> row : rowsOfCells) {
+            for (int j = 0; j < row.size(); j++) {
+                columnsOfCells.get(j).add(row.get(j));
+            }
+        }
+
+        List<Integer> columnIndicesToRemove = new ArrayList<>();
+        int columnIndex = 0;
+        for (List<Cell> col : columnsOfCells) {
+            if (col.stream()
+                    .allMatch(cell -> cell.getTextBlocks().isEmpty())) {
+                columnIndicesToRemove.add(columnIndex);
+            }
+            columnIndex++;
+        }
+        columnIndicesToRemove.sort(Collections.reverseOrder());
+
+        // update all rows so that the values of the empty columns get removed
+        var rowsOfCellsBefore = new ArrayList<>(rowsOfCells);
+        rowsOfCells = new ArrayList<>();
+        for (List<Cell> row : rowsOfCellsBefore) {
+            var updatedRow = new ArrayList<>(row);
+            columnIndicesToRemove.forEach(idxToRemove -> updatedRow.remove(updatedRow.get(idxToRemove)));
+            rowsOfCells.add(updatedRow);
+        }
+
+        return rowsOfCells;
     }
 
 
+    private void addCellToRowAndCol(Cell cell, int row, int col) {
+
+        unrotatedRowCount = Math.max(unrotatedRowCount, row + 1);
+        unrotatedColCount = Math.max(unrotatedColCount, col + 1);
+
+        CellPosition cp = new CellPosition(row, col);
+        cellTreeMap.put(cp, cell);
+
+    }
+
 
     @Override
     public String getText() {
@@ -314,7 +365,7 @@ public class TablePageBlock extends AbstractPageBlock {
                             if (!first) {
                                 sb.append("\n");
                             }
-                            sb.append('\"').append(textBlock.getText().replaceAll("\"", "\\\"")).append('\"');
+                            sb.append('\"').append(textBlock.getText().replaceAll("\"", "\"")).append('\"');
                             first = false;
                         }
                     }
@@ -328,8 +379,6 @@ public class TablePageBlock extends AbstractPageBlock {
     }
 
 
-
-
     public String getTextAsHtml() {
 
         StringBuilder sb = new StringBuilder();
@@ -363,4 +412,9 @@ public class TablePageBlock extends AbstractPageBlock {
         return sb.toString();
     }
 
+
+    record CellWithIntersection(Cell originalCell, double intersectedArea) {
+
+    }
+
 }

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/RulingCleaningService.java

@@ -1,21 +1,21 @@
 package com.knecon.fforesight.service.layoutparser.processor.services;
 
-import java.awt.geom.Line2D;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.X_FIRST_RULING_COMPARATOR;
+
 import java.awt.geom.Point2D;
 import java.util.ArrayList;
-import java.util.Collection;
-import java.util.Collections;
-import java.util.Comparator;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.stream.Collectors;
 
 import org.springframework.stereotype.Service;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.table.CleanRulings;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Rectangle;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
-import com.knecon.fforesight.service.layoutparser.processor.utils.DoubleComparisons;
+import com.knecon.fforesight.service.layoutparser.processor.utils.UnionFind;
 
 import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
@@ -25,25 +25,145 @@ import lombok.extern.slf4j.Slf4j;
 @RequiredArgsConstructor
 public class RulingCleaningService {
 
-    private static final float THRESHOLD = 6;
+    private static final float THRESHOLD_X_VERTICAL = 1;
+    private static final float THRESHOLD_Y_VERTICAL = 2;
+    private static final float THRESHOLD_X_HORIZONTAL = 2;
+    private static final float THRESHOLD_Y_HORIZONTAL = 3;
 
 
     public CleanRulings getCleanRulings(List<TableCells> tableCells, List<Ruling> rulings) {
 
+        Rulings verticalAndHorizontalRulingLines;
+
         if (!rulings.isEmpty()) {
-            snapPoints(rulings);
+            verticalAndHorizontalRulingLines = extractVerticalAndHorizontalRulingLines(rulings);
+        } else {
+            verticalAndHorizontalRulingLines = getRulingsFromParsedCells(tableCells);
         }
 
+        verticalAndHorizontalRulingLines.verticalLines.sort(X_FIRST_RULING_COMPARATOR);
+        verticalAndHorizontalRulingLines.horizontalLines.sort(X_FIRST_RULING_COMPARATOR);
+        verticalAndHorizontalRulingLines = cleanRulings(verticalAndHorizontalRulingLines);
+
+        return CleanRulings.builder().vertical(verticalAndHorizontalRulingLines.verticalLines()).horizontal(verticalAndHorizontalRulingLines.horizontalLines()).build();
+    }
+
+
+    private Rulings cleanRulings(Rulings rulings) {
+
+        List<List<Rectangle>> groupedOverlappingVerticalRectangles = groupOverlappingRectangles(rulings.verticalLines.stream()
+                                                                                                        .map(RulingCleaningService::getOverlapRectangle)
+                                                                                                        .distinct()
+                                                                                                        .toList());
+        List<Ruling> cleanedVerticalRulings = groupedOverlappingVerticalRectangles.stream()
+                .map(rectList -> getXCenteredRuling(Rectangle.boundingBoxOf(rectList)))
+                .toList();
+
+        List<List<Rectangle>> groupedOverlappingHorizontalRectangles = groupOverlappingRectangles(rulings.horizontalLines.stream()
+                                                                                                          .map(RulingCleaningService::getOverlapRectangle)
+                                                                                                          .distinct()
+                                                                                                          .toList());
+
+        List<Ruling> cleanedHorizontalRulings = groupedOverlappingHorizontalRectangles.stream()
+                .map(rectList -> getYCenteredRuling(Rectangle.boundingBoxOf(rectList)))
+                .collect(Collectors.toList());
+
+        return new Rulings(cleanedVerticalRulings, cleanedHorizontalRulings);
+    }
+
+
+    private List<List<Rectangle>> groupOverlappingRectangles(List<Rectangle> rectangles) {
+
+        UnionFind<Rectangle> unionFind = new UnionFind<>();
+        for (int i = 0; i < rectangles.size(); i++) {
+            for (int j = i + 1; j < rectangles.size(); j++) {
+                Rectangle rectangle1 = rectangles.get(i);
+                Rectangle rectangle2 = rectangles.get(j);
+
+                // we can stop early when we are too far off because of x-y-sorting
+                if(rectangle1.getRight() < rectangle2.getLeft() && rectangle1.getBottom() < rectangle2.getTop()) {
+                    break;
+                }
+
+                if (rectangle1.intersects(rectangle2)) {
+                    unionFind.union(rectangle1, rectangle2);
+                }
+            }
+        }
+
+        Map<Rectangle, List<Rectangle>> groups = new HashMap<>();
+        for (Rectangle rectangle : rectangles) {
+            Rectangle root = unionFind.find(rectangle);
+            groups.computeIfAbsent(root, k -> new ArrayList<>()).add(rectangle);
+        }
+        return new ArrayList<>(groups.values());
+    }
+
+
+    private static Rectangle getOverlapRectangle(Ruling ruling) {
+
+        float top;
+        float left;
+        float w;
+        float h;
+
+        if (ruling.x1 < ruling.x2) {
+            left = ruling.x1;
+            w = ruling.x2 - ruling.x1;
+        } else {
+            left = ruling.x2;
+            w = ruling.x1 - ruling.x2;
+        }
+        if (ruling.y1 < ruling.y2) {
+            top = ruling.y1;
+            h = ruling.y2 - ruling.y1;
+        } else {
+            top = ruling.y2;
+            h = ruling.y1 - ruling.y2;
+        }
+
+        if (ruling.horizontal()) {
+            return new Rectangle(top - THRESHOLD_Y_HORIZONTAL, left - THRESHOLD_X_HORIZONTAL, w + 2 * THRESHOLD_X_HORIZONTAL, h + 2 * THRESHOLD_Y_HORIZONTAL);
+        } else {
+            return new Rectangle(top - THRESHOLD_Y_VERTICAL, left - THRESHOLD_X_VERTICAL, w + 2 * THRESHOLD_X_VERTICAL, h + 2 * THRESHOLD_Y_VERTICAL);
+        }
+    }
+
+
+    public static Ruling getXCenteredRuling(Rectangle rectangle) {
+
+        float x = (float) rectangle.getCenterX();
+        float y1 = rectangle.getTop();
+        float y2 = rectangle.getBottom();
+
+        Point2D point1 = new Point2D.Float(x, y1 + THRESHOLD_Y_VERTICAL);
+        Point2D point2 = new Point2D.Float(x, y2 - THRESHOLD_Y_VERTICAL);
+
+        return new Ruling(point1, point2);
+    }
+
+
+    public static Ruling getYCenteredRuling(Rectangle rectangle) {
+
+        float x1 = rectangle.getLeft();
+        float x2 = rectangle.getRight();
+        float y = (float) rectangle.getCenterY();
+
+        Point2D point1 = new Point2D.Float(x1 + THRESHOLD_X_HORIZONTAL, y);
+        Point2D point2 = new Point2D.Float(x2 - THRESHOLD_X_HORIZONTAL, y);
+
+        return new Ruling(point1, point2);
+    }
+
+
+    private Rulings extractVerticalAndHorizontalRulingLines(List<Ruling> rulings) {
+
         List<Ruling> vrs = new ArrayList<>();
         for (Ruling vr : rulings) {
             if (vr.vertical()) {
                 vrs.add(vr);
             }
         }
-        if (vrs.isEmpty()) {
-            vrs.addAll(extractVerticalRulings(tableCells));
-        }
-        List<Ruling> verticalRulingLines = collapseOrientedRulings(vrs);
 
         List<Ruling> hrs = new ArrayList<>();
         for (Ruling hr : rulings) {
@@ -51,98 +171,26 @@ public class RulingCleaningService {
                 hrs.add(hr);
             }
         }
-        if (hrs.isEmpty()) {
-            hrs.addAll(extractHorizontalRulings(tableCells));
-        }
-        List<Ruling> horizontalRulingLines = collapseOrientedRulings(hrs);
-
-        return CleanRulings.builder().vertical(verticalRulingLines).horizontal(horizontalRulingLines).build();
+        return new Rulings(vrs, hrs);
     }
 
 
-    public void snapPoints(List<? extends Line2D.Float> rulings) {
+    private Rulings getRulingsFromParsedCells(List<TableCells> tableCells) {
 
-        // collect points and keep a Line -> p1,p2 map
-        Map<Line2D.Float, Point2D[]> linesToPoints = new HashMap<>();
-        List<Point2D> points = new ArrayList<>();
-        for (Line2D.Float r : rulings) {
-            Point2D p1 = r.getP1();
-            Point2D p2 = r.getP2();
-            linesToPoints.put(r, new Point2D[]{p1, p2});
-            points.add(p1);
-            points.add(p2);
-        }
-
-        // snap by X
-        points.sort(Comparator.comparingDouble(Point2D::getX));
-
-        List<List<Point2D>> groupedPoints = new ArrayList<>();
-        groupedPoints.add(new ArrayList<>(Collections.singletonList(points.get(0))));
-
-        for (Point2D p : points.subList(1, points.size() - 1)) {
-            List<Point2D> last = groupedPoints.get(groupedPoints.size() - 1);
-            if (Math.abs(p.getX() - last.get(0).getX()) < THRESHOLD) {
-                groupedPoints.get(groupedPoints.size() - 1).add(p);
-            } else {
-                groupedPoints.add(new ArrayList<>(Collections.singletonList(p)));
-            }
-        }
-
-        for (List<Point2D> group : groupedPoints) {
-            float avgLoc = 0;
-            for (Point2D p : group) {
-                avgLoc += p.getX();
-            }
-            avgLoc /= group.size();
-            for (Point2D p : group) {
-                p.setLocation(avgLoc, p.getY());
-            }
-        }
-        // ---
-
-        // snap by Y
-        points.sort(Comparator.comparingDouble(Point2D::getY));
-
-        groupedPoints = new ArrayList<>();
-        groupedPoints.add(new ArrayList<>(Collections.singletonList(points.get(0))));
-
-        for (Point2D p : points.subList(1, points.size() - 1)) {
-            List<Point2D> last = groupedPoints.get(groupedPoints.size() - 1);
-            if (Math.abs(p.getY() - last.get(0).getY()) < THRESHOLD) {
-                groupedPoints.get(groupedPoints.size() - 1).add(p);
-            } else {
-                groupedPoints.add(new ArrayList<>(Collections.singletonList(p)));
-            }
-        }
-
-        for (List<Point2D> group : groupedPoints) {
-            float avgLoc = 0;
-            for (Point2D p : group) {
-                avgLoc += p.getY();
-            }
-            avgLoc /= group.size();
-            for (Point2D p : group) {
-                p.setLocation(p.getX(), avgLoc);
-            }
-        }
-        // ---
-
-        // finally, modify lines
-        for (Map.Entry<Line2D.Float, Point2D[]> ltp : linesToPoints.entrySet()) {
-            Point2D[] p = ltp.getValue();
-            ltp.getKey().setLine(p[0], p[1]);
-        }
+        List<Ruling> vrs = extractVerticalRulingsFromParsedCells(tableCells);
+        List<Ruling> hrs = extractHorizontalRulingsFromParsedCells(tableCells);
+        return new Rulings(vrs, hrs);
     }
 
 
-    private Collection<? extends Ruling> extractVerticalRulings(List<TableCells> cvParsedTableCells) {
+    private List<Ruling> extractVerticalRulingsFromParsedCells(List<TableCells> tableCells) {
 
         List<Ruling> vrs = new ArrayList<>();
 
-        if (cvParsedTableCells != null) {
-            for (TableCells cvParsedTableCell : cvParsedTableCells) {
-                Ruling leftLine = createRuling(cvParsedTableCell.getX0(), cvParsedTableCell.getX0(), cvParsedTableCell.getY0(), cvParsedTableCell.getY1());
-                Ruling rightLine = createRuling(cvParsedTableCell.getX1(), cvParsedTableCell.getX1(), cvParsedTableCell.getY0(), cvParsedTableCell.getY1());
+        if (tableCells != null) {
+            for (TableCells tableCell : tableCells) {
+                Ruling leftLine = createRuling(tableCell.getX0(), tableCell.getX0(), tableCell.getY0(), tableCell.getY1());
+                Ruling rightLine = createRuling(tableCell.getX1(), tableCell.getX1(), tableCell.getY0(), tableCell.getY1());
                 vrs.add(leftLine);
                 vrs.add(rightLine);
             }
@@ -151,19 +199,18 @@ public class RulingCleaningService {
     }
 
 
-    private Collection<? extends Ruling> extractHorizontalRulings(List<TableCells> cvParsedTableCells) {
+    private List<Ruling> extractHorizontalRulingsFromParsedCells(List<TableCells> tableCells) {
 
         List<Ruling> hrs = new ArrayList<>();
 
-        if (cvParsedTableCells != null) {
-            for (TableCells cvParsedTableCell : cvParsedTableCells) {
-                Ruling topLine = createRuling(cvParsedTableCell.getX0(), cvParsedTableCell.getX1(), cvParsedTableCell.getY1(), cvParsedTableCell.getY1());
-                Ruling baseLine = createRuling(cvParsedTableCell.getX0(), cvParsedTableCell.getX1(), cvParsedTableCell.getY0(), cvParsedTableCell.getY0());
+        if (tableCells != null) {
+            for (TableCells tableCell : tableCells) {
+                Ruling topLine = createRuling(tableCell.getX0(), tableCell.getX1(), tableCell.getY1(), tableCell.getY1());
+                Ruling baseLine = createRuling(tableCell.getX0(), tableCell.getX1(), tableCell.getY0(), tableCell.getY0());
                 hrs.add(topLine);
                 hrs.add(baseLine);
             }
         }
-
         return hrs;
     }
 
@@ -189,46 +236,8 @@ public class RulingCleaningService {
     }
 
 
-    private List<Ruling> collapseOrientedRulings(List<Ruling> lines) {
+    private record Rulings(List<Ruling> verticalLines, List<Ruling> horizontalLines) {
 
-        int COLINEAR_OR_PARALLEL_PIXEL_EXPAND_AMOUNT = 1;
-        return collapseOrientedRulings(lines, COLINEAR_OR_PARALLEL_PIXEL_EXPAND_AMOUNT);
-    }
-
-
-    private List<Ruling> collapseOrientedRulings(List<Ruling> lines, int expandAmount) {
-
-        ArrayList<Ruling> rv = new ArrayList<>();
-        lines.sort((a, b) -> {
-            final float diff = a.getPosition() - b.getPosition();
-            return Float.compare(diff == 0 ? a.getStart() - b.getStart() : diff, 0f);
-        });
-
-        for (Ruling next_line : lines) {
-            Ruling last = rv.isEmpty() ? null : rv.get(rv.size() - 1);
-            // if current line colinear with next, and are "close enough": expand current line
-            if (last != null && DoubleComparisons.feq(next_line.getPosition(), last.getPosition()) && last.nearlyIntersects(next_line, expandAmount)) {
-                final float lastStart = last.getStart();
-                final float lastEnd = last.getEnd();
-
-                final boolean lastFlipped = lastStart > lastEnd;
-                final boolean nextFlipped = next_line.getStart() > next_line.getEnd();
-
-                boolean differentDirections = nextFlipped != lastFlipped;
-                float nextS = differentDirections ? next_line.getEnd() : next_line.getStart();
-                float nextE = differentDirections ? next_line.getStart() : next_line.getEnd();
-
-                final float newStart = lastFlipped ? Math.max(nextS, lastStart) : Math.min(nextS, lastStart);
-                final float newEnd = lastFlipped ? Math.min(nextE, lastEnd) : Math.max(nextE, lastEnd);
-                last.setStartEnd(newStart, newEnd);
-                assert !last.oblique();
-            } else if (next_line.length() == 0) {
-                continue;
-            } else {
-                rv.add(next_line);
-            }
-        }
-        return rv;
     }
 
 }

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/TableExtractionService.java

@@ -1,14 +1,13 @@
 package com.knecon.fforesight.service.layoutparser.processor.services;
 
-import java.awt.geom.Point2D;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.CELL_SIZE_COMPARATOR;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.RECTANGLE_SIZE_COMPARATOR;
+
 import java.util.ArrayList;
-import java.util.Comparator;
-import java.util.HashMap;
 import java.util.HashSet;
-import java.util.Iterator;
 import java.util.List;
 import java.util.Map;
-import java.util.Set;
+import java.util.stream.Collectors;
 
 import org.springframework.stereotype.Service;
 
@@ -21,59 +20,15 @@ import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
 import com.knecon.fforesight.service.layoutparser.processor.utils.DoubleComparisons;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangularIntersectionFinder;
+import com.knecon.fforesight.service.layoutparser.processor.utils.SpreadsheetFinder;
 
 @Service
 public class TableExtractionService {
 
-    private static final Comparator<Point2D> X_FIRST_POINT_COMPARATOR = (arg0, arg1) -> {
-
-        int rv = 0;
-        float arg0X = DoubleComparisons.round(arg0.getX(), 2);
-        float arg0Y = DoubleComparisons.round(arg0.getY(), 2);
-        float arg1X = DoubleComparisons.round(arg1.getX(), 2);
-        float arg1Y = DoubleComparisons.round(arg1.getY(), 2);
-
-        if (arg0X > arg1X) {
-            rv = 1;
-        } else if (arg0X < arg1X) {
-            rv = -1;
-        } else if (arg0Y > arg1Y) {
-            rv = 1;
-        } else if (arg0Y < arg1Y) {
-            rv = -1;
-        }
-        return rv;
-    };
-    private static final Comparator<Point2D> POINT_COMPARATOR = (arg0, arg1) -> {
-
-        int rv = 0;
-        float arg0X = DoubleComparisons.round(arg0.getX(), 2);
-        float arg0Y = DoubleComparisons.round(arg0.getY(), 2);
-        float arg1X = DoubleComparisons.round(arg1.getX(), 2);
-        float arg1Y = DoubleComparisons.round(arg1.getY(), 2);
-
-        if (arg0Y > arg1Y) {
-            rv = 1;
-        } else if (arg0Y < arg1Y) {
-            rv = -1;
-        } else if (arg0X > arg1X) {
-            rv = 1;
-        } else if (arg0X < arg1X) {
-            rv = -1;
-        }
-        return rv;
-    };
-
-
-    public boolean contains(Cell cell, double x, double y, double w, double h) {
-
-        if (cell.isEmpty() || w <= 0 || h <= 0) {
-            return false;
-        }
-        double x0 = cell.getX();
-        double y0 = cell.getY();
-        return (x >= x0 - 2 && y >= y0 - 2 && (x + w) <= x0 + cell.getWidth() + 2 && (y + h) <= y0 + cell.getHeight() + 2);
-    }
+    private static final int MAX_TABLE_CONTAINED_CELLS_WITH_TEXT = 1;
+    private static final int TEXT_BLOCK_CONTAINMENT_TOLERANCE = 2;
+    private static final double TABLE_UNIFORMITY_THRESHOLD = 0.7;
 
 
     /**
@@ -89,22 +44,18 @@ public class TableExtractionService {
      * @param cleanRulings The lines used to build the table.
      * @param page         Page object that contains textblocks and statistics.
      */
+
     public void extractTables(CleanRulings cleanRulings, ClassificationPage page) {
 
         List<Cell> cells = findCells(cleanRulings.getHorizontal(), cleanRulings.getVertical());
-
-        List<TextPageBlock> toBeRemoved = new ArrayList<>();
+        // sort cells by size (height * width) ascending so that textBlocks are always assigned to the smallest cells that contain them
+        cells.sort(CELL_SIZE_COMPARATOR);
 
         for (AbstractPageBlock abstractPageBlock : page.getTextBlocks()) {
             TextPageBlock textBlock = (TextPageBlock) abstractPageBlock;
             for (Cell cell : cells) {
-                if (cell.hasMinimumSize() && contains(cell,
-                        textBlock.getPdfMinX(),
-                        textBlock.getPdfMinY(),
-                        textBlock.getPdfMaxX() - textBlock.getPdfMinX(),
-                        textBlock.getPdfMaxY() - textBlock.getPdfMinY())) {
+                if (cell.hasMinimumSize() && doesCellContainTextBlock(cell, textBlock)) {
                     cell.addTextBlock(textBlock);
-                    toBeRemoved.add(textBlock);
                     break;
                 }
             }
@@ -113,245 +64,94 @@ public class TableExtractionService {
         cells = new ArrayList<>(new HashSet<>(cells));
         DoubleComparisons.sort(cells, Rectangle.ILL_DEFINED_ORDER);
 
-        List<Rectangle> spreadsheetAreas = findSpreadsheetsFromCells(cells);
+        List<Rectangle> spreadsheetAreas = SpreadsheetFinder.findSpreadsheetsFromCells(cells);
+        // sort spreadsheetAreas by size (height * width) ascending so that cells are placed in the smallest tables first
+        // this way no cell duplication occurs when tables are contained in other tables and only the most inner table contains the cells
+        spreadsheetAreas.sort(RECTANGLE_SIZE_COMPARATOR);
 
         List<TablePageBlock> tables = new ArrayList<>();
         for (Rectangle area : spreadsheetAreas) {
 
-            List<Cell> overlappingCells = new ArrayList<>();
+            List<Cell> containedCells = new ArrayList<>();
             for (Cell c : cells) {
-                if (c.hasMinimumSize() && c.intersects(area)) {
-                    overlappingCells.add(c);
+                if (c.hasMinimumSize() && area.contains(c)) {
+                    containedCells.add(c);
                 }
             }
-            tables.add(new TablePageBlock(overlappingCells, area, page.getRotation()));
+
+            var containedCellsWithText = containedCells.stream()
+                    .filter(cell -> !cell.getTextBlocks().isEmpty())
+                    .toList();
+
+            // verify if table would contain fewer cells with text than the threshold allows
+            if (containedCellsWithText.size() >= MAX_TABLE_CONTAINED_CELLS_WITH_TEXT && checkIfTableCellsAreUniform(containedCells)) {
+                tables.add(new TablePageBlock(containedCells, area, page.getRotation()));
+                cells.removeAll(containedCells);
+            }
         }
 
         for (TablePageBlock table : tables) {
             int position = -1;
 
-            Iterator<AbstractPageBlock> itty = page.getTextBlocks().iterator();
-            while (itty.hasNext()) {
-                AbstractPageBlock textBlock = itty.next();
-                if (textBlock instanceof TextPageBlock ? table.containsBlock((TextPageBlock) textBlock) : table.contains(textBlock) && position == -1) {
-                    position = page.getTextBlocks().indexOf(textBlock);
+            for (AbstractPageBlock pageBlock : page.getTextBlocks()) {
+                if (pageBlock instanceof TextPageBlock ? table.containsBlock((TextPageBlock) pageBlock) : table.contains(pageBlock) && position == -1) {
+                    position = page.getTextBlocks().indexOf(pageBlock);
                 }
             }
             if (position != -1) {
                 page.getTextBlocks().add(position, table);
+
+                var toBeRemoved = table.getCells()
+                        .stream()
+                        .map(Cell::getTextBlocks)
+                        .flatMap(List::stream)
+                        .toList();
+                // remove text blocks from the page that were also added with the table (from its contained cells)
+                page.getTextBlocks().removeAll(toBeRemoved);
             }
         }
-
-        page.getTextBlocks().removeAll(toBeRemoved);
     }
 
 
-    public List<Cell> findCells(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+    private boolean checkIfTableCellsAreUniform(List<Cell> containedCells) {
 
-        // Fix for 211.pdf
-        for (Ruling r : horizontalRulingLines) {
-            if (r.getX2() < r.getX1()) {
-                double a = r.getX2();
-                r.x2 = (float) r.getX1();
-                r.x1 = (float) a;
-            }
+        if(containedCells.size() <= 2) {
+            return true;
         }
 
-        List<Cell> cellsFound = new ArrayList<>();
-        Map<Point2D, Ruling[]> intersectionPoints = Ruling.findIntersections(horizontalRulingLines, verticalRulingLines);
-        List<Point2D> intersectionPointsList = new ArrayList<>(intersectionPoints.keySet());
-        intersectionPointsList.sort(POINT_COMPARATOR);
+        Map<Long, List<Long>> cellsGroupedByRoundedWidth = containedCells.stream()
+                .map(Rectangle::getWidth)
+                .map(size -> Math.round(size / 10.0) * 10)
+                .collect(Collectors.groupingBy(Long::longValue));
 
-        for (int i = 0; i < intersectionPointsList.size(); i++) {
-            Point2D topLeft = intersectionPointsList.get(i);
-            Ruling[] hv = intersectionPoints.get(topLeft);
-
-            // CrossingPointsDirectlyBelow( topLeft );
-            List<Point2D> xPoints = new ArrayList<>();
-            // CrossingPointsDirectlyToTheRight( topLeft );
-            List<Point2D> yPoints = new ArrayList<>();
-
-            for (Point2D p : intersectionPointsList.subList(i, intersectionPointsList.size())) {
-                if (p.getX() == topLeft.getX() && p.getY() > topLeft.getY()) {
-                    xPoints.add(p);
-                }
-                if (p.getY() == topLeft.getY() && p.getX() > topLeft.getX()) {
-                    yPoints.add(p);
-                }
-            }
-            outer:
-            for (Point2D xPoint : xPoints) {
-                // is there a vertical edge b/w topLeft and xPoint?
-                if (!hv[1].equals(intersectionPoints.get(xPoint)[1])) {
-                    continue;
-                }
-                for (Point2D yPoint : yPoints) {
-                    // is there an horizontal edge b/w topLeft and yPoint ?
-                    if (!hv[0].equals(intersectionPoints.get(yPoint)[0])) {
-                        continue;
-                    }
-                    Point2D btmRight = new Point2D.Float((float) yPoint.getX(), (float) xPoint.getY());
-                    if (intersectionPoints.containsKey(btmRight) && intersectionPoints.get(btmRight)[0].equals(intersectionPoints.get(xPoint)[0]) && intersectionPoints.get(btmRight)[1].equals(
-                            intersectionPoints.get(yPoint)[1])) {
-                        cellsFound.add(new Cell(topLeft, btmRight));
-                        break outer;
-                    }
-                }
-            }
-        }
-
-        // TODO create cells for vertical ruling lines with aligned endpoints at the top/bottom of a grid
-        // that aren't connected with an horizontal ruler?
-        // see: https://github.com/jazzido/tabula-extractor/issues/78#issuecomment-41481207
-
-        return cellsFound;
+        return (double) cellsGroupedByRoundedWidth.size() / containedCells.size() <= TABLE_UNIFORMITY_THRESHOLD;
     }
 
 
-    private List<Rectangle> findSpreadsheetsFromCells(List<? extends Rectangle> cells) {
-        // via: http://stackoverflow.com/questions/13746284/merging-multiple-adjacent-rectangles-into-one-polygon
-        List<Rectangle> rectangles = new ArrayList<>();
-        Set<Point2D> pointSet = new HashSet<>();
-        Map<Point2D, Point2D> edgesH = new HashMap<>();
-        Map<Point2D, Point2D> edgesV = new HashMap<>();
-        int i = 0;
+    private boolean doesCellContainTextBlock(Cell cell, TextPageBlock textBlock) {
 
-        for (Rectangle cell : cells) {
-            for (Point2D pt : cell.getPoints()) {
-                if (pointSet.contains(pt)) { // shared vertex, remove it
-                    pointSet.remove(pt);
-                } else {
-                    pointSet.add(pt);
-                }
-            }
+        double x = textBlock.getPdfMinX();
+        double y = textBlock.getPdfMinY();
+        double w = textBlock.getPdfMaxX() - textBlock.getPdfMinX();
+        double h = textBlock.getPdfMaxY() - textBlock.getPdfMinY();
+        if (cell.isEmpty() || w <= 0 || h <= 0) {
+            return false;
         }
-
-        // X first sort
-        List<Point2D> pointsSortX = new ArrayList<>(pointSet);
-        pointsSortX.sort(X_FIRST_POINT_COMPARATOR);
-        // Y first sort
-        List<Point2D> pointsSortY = new ArrayList<>(pointSet);
-        pointsSortY.sort(POINT_COMPARATOR);
-
-        while (i < pointSet.size()) {
-            float currY = (float) pointsSortY.get(i).getY();
-            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortY.get(i).getY(), currY)) {
-                edgesH.put(pointsSortY.get(i), pointsSortY.get(i + 1));
-                edgesH.put(pointsSortY.get(i + 1), pointsSortY.get(i));
-                i += 2;
-            }
-        }
-
-        i = 0;
-        while (i < pointSet.size()) {
-            float currX = (float) pointsSortX.get(i).getX();
-            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortX.get(i).getX(), currX)) {
-                edgesV.put(pointsSortX.get(i), pointsSortX.get(i + 1));
-                edgesV.put(pointsSortX.get(i + 1), pointsSortX.get(i));
-                i += 2;
-            }
-        }
-
-        // Get all the polygons
-        List<List<PolygonVertex>> polygons = new ArrayList<>();
-        Point2D nextVertex;
-        while (!edgesH.isEmpty()) {
-            ArrayList<PolygonVertex> polygon = new ArrayList<>();
-            Point2D first = edgesH.keySet().iterator().next();
-            polygon.add(new PolygonVertex(first, Direction.HORIZONTAL));
-            edgesH.remove(first);
-
-            while (true) {
-                PolygonVertex curr = polygon.get(polygon.size() - 1);
-                PolygonVertex lastAddedVertex;
-                if (curr.direction == Direction.HORIZONTAL) {
-                    nextVertex = edgesV.get(curr.point);
-                    edgesV.remove(curr.point);
-                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.VERTICAL);
-                } else {
-                    nextVertex = edgesH.get(curr.point);
-                    edgesH.remove(curr.point);
-                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.HORIZONTAL);
-                }
-                polygon.add(lastAddedVertex);
-
-                if (lastAddedVertex.equals(polygon.get(0))) {
-                    // closed polygon
-                    polygon.remove(polygon.size() - 1);
-                    break;
-                }
-            }
-
-            for (PolygonVertex vertex : polygon) {
-                edgesH.remove(vertex.point);
-                edgesV.remove(vertex.point);
-            }
-            polygons.add(polygon);
-        }
-
-        // calculate grid-aligned minimum area rectangles for each found polygon
-        for (List<PolygonVertex> poly : polygons) {
-            float top = Float.MAX_VALUE;
-            float left = Float.MAX_VALUE;
-            float bottom = Float.MIN_VALUE;
-            float right = Float.MIN_VALUE;
-            for (PolygonVertex pt : poly) {
-                top = (float) Math.min(top, pt.point.getY());
-                left = (float) Math.min(left, pt.point.getX());
-                bottom = (float) Math.max(bottom, pt.point.getY());
-                right = (float) Math.max(right, pt.point.getX());
-            }
-            rectangles.add(new Rectangle(top, left, right - left, bottom - top));
-        }
-
-        return rectangles;
+        double x0 = cell.getX();
+        double y0 = cell.getY();
+        return (x >= x0 - TEXT_BLOCK_CONTAINMENT_TOLERANCE
+                && y >= y0 - TEXT_BLOCK_CONTAINMENT_TOLERANCE
+                && (x + w) <= x0 + cell.getWidth() + 2 * TEXT_BLOCK_CONTAINMENT_TOLERANCE
+                && (y + h) <= y0 + cell.getHeight() + 2 * TEXT_BLOCK_CONTAINMENT_TOLERANCE);
     }
 
 
-    private enum Direction {
-        HORIZONTAL,
-        VERTICAL
-    }
-
-    static class PolygonVertex {
-
-        Point2D point;
-        Direction direction;
-
-
-        PolygonVertex(Point2D point, Direction direction) {
-
-            this.direction = direction;
-            this.point = point;
-        }
-
-
-        @Override
-        public boolean equals(Object other) {
-
-            if (this == other) {
-                return true;
-            }
-            if (!(other instanceof PolygonVertex)) {
-                return false;
-            }
-            return this.point.equals(((PolygonVertex) other).point);
-        }
-
-
-        @Override
-        public int hashCode() {
-
-            return this.point.hashCode();
-        }
-
-
-        @Override
-        public String toString() {
-
-            return String.format("%s[point=%s,direction=%s]", this.getClass().getName(), this.point.toString(), this.direction.toString());
-        }
+    public static List<Cell> findCells(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
 
+        return RectangularIntersectionFinder.find(horizontalRulingLines, verticalRulingLines)
+                .stream()
+                .map(Cell::new)
+                .collect(Collectors.toList());
     }
 
 }

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> tocId = context.getDocumentTree().createNewChildEntryAndReturnId(section, imageNode);
-        imageNode.setTreeId(tocId);
+        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()));
@@ -188,14 +191,33 @@ public class LayoutGridService {
     @SneakyThrows
     private void addPlacedText(Page page, Rectangle2D textBBox, String s, LayoutGrid layoutGrid) {
 
-        Point2D.Float upperLeftCorner = switch (page.getRotation()) {
-            case 90 -> new Point2D.Float((float) (textBBox.getMinX()), (float) textBBox.getMinY());
-            case 180 -> new Point2D.Float((float) (textBBox.getMaxX()), (float) textBBox.getMinY());
-            case 270 -> new Point2D.Float((float) (textBBox.getMaxX()), (float) textBBox.getMaxY());
-            default -> new Point2D.Float((float) (textBBox.getMinX()), (float) textBBox.getMaxY());
-        };
+        // translates text, such that its right edge is a bit to the left of the drawn box
+        float translationAmount = ((FONT.getStringWidth(s) / 1000) * FONT_SIZE + (2 * LINE_WIDTH) + 4);
+
+        Point2D upperLeftCorner;
+        Point2D translationVector;
+        switch (page.getRotation()) {
+            case 90 -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMinY());
+                translationVector = new Point2D.Double(FONT_SIZE, -translationAmount);
+            }
+            case 180 -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMinY());
+                translationVector = new Point2D.Double(translationAmount, FONT_SIZE);
+            }
+            case 270 -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMaxX(), textBBox.getMaxY());
+                translationVector = new Point2D.Double(-FONT_SIZE, translationAmount);
+            }
+            default -> {
+                upperLeftCorner = new Point2D.Double(textBBox.getMinX(), textBBox.getMaxY());
+                translationVector = new Point2D.Double(-translationAmount, -FONT_SIZE);
+            }
+        }
+
+        upperLeftCorner = add(upperLeftCorner, translationVector);
+
         var placedTexts = layoutGrid.getVisualizationsPerPages().get(page.getNumber() - 1).getPlacedTexts();
-        upperLeftCorner.setLocation(upperLeftCorner.getX() - ((FONT.getStringWidth(s) / 1000) * FONT_SIZE + (2 * LINE_WIDTH) + 4), upperLeftCorner.getY() - FONT_SIZE);
         placedTexts.add(PlacedText.textFacingUp(s, upperLeftCorner, FONT_SIZE, Color.BLACK, FONT));
     }
 
@@ -317,4 +339,10 @@ public class LayoutGridService {
                         .add(new ColoredRectangle(textBBox, color, LINE_WIDTH)));
     }
 
+
+    private Point2D add(Point2D a, Point2D b) {
+
+        return new Point2D.Double(a.getX() + b.getX(), a.getY() + b.getY());
+    }
+
 }

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

@@ -0,0 +1,28 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import java.awt.Color;
+
+import lombok.AccessLevel;
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Getter;
+import lombok.experimental.FieldDefaults;
+
+@Builder
+@AllArgsConstructor
+@Getter
+@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
+public class DrawingOptions {
+
+    boolean stroke;
+    @Builder.Default
+    Color strokeColor = Color.BLACK;
+    @Builder.Default
+    float strokeWidth = 1f;
+
+    boolean fill;
+    @Builder.Default
+    Color fillColor = Color.BLACK;
+
+}
+

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

@@ -0,0 +1,88 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import java.awt.geom.Point2D;
+import java.util.Comparator;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Cell;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Rectangle;
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
+
+public class GeometricComparators {
+
+    private static final int COMPARATOR_ROUNDING = 2;
+
+    public static final Comparator<Point2D> X_FIRST_POINT_COMPARATOR = (point1, point2) -> {
+
+        int rv = 0;
+        float point1X = DoubleComparisons.round(point1.getX(), COMPARATOR_ROUNDING);
+        float point1Y = DoubleComparisons.round(point1.getY(), COMPARATOR_ROUNDING);
+        float point2X = DoubleComparisons.round(point2.getX(), COMPARATOR_ROUNDING);
+        float point2Y = DoubleComparisons.round(point2.getY(), COMPARATOR_ROUNDING);
+
+        if (point1X > point2X) {
+            rv = 1;
+        } else if (point1X < point2X) {
+            rv = -1;
+        } else if (point1Y > point2Y) {
+            rv = 1;
+        } else if (point1Y < point2Y) {
+            rv = -1;
+        }
+        return rv;
+    };
+
+    public static final Comparator<Point2D> Y_FIRST_POINT_COMPARATOR = (point1, point2) -> {
+
+        int rv = 0;
+        float point1X = DoubleComparisons.round(point1.getX(), COMPARATOR_ROUNDING);
+        float point1Y = DoubleComparisons.round(point1.getY(), COMPARATOR_ROUNDING);
+        float point2X = DoubleComparisons.round(point2.getX(), COMPARATOR_ROUNDING);
+        float point2Y = DoubleComparisons.round(point2.getY(), COMPARATOR_ROUNDING);
+
+        if (point1Y > point2Y) {
+            rv = 1;
+        } else if (point1Y < point2Y) {
+            rv = -1;
+        } else if (point1X > point2X) {
+            rv = 1;
+        } else if (point1X < point2X) {
+            rv = -1;
+        }
+        return rv;
+    };
+
+    public static final Comparator<Cell> CELL_SIZE_COMPARATOR = (cell1, cell2) -> {
+
+        Double cell1Size = cell1.getHeight() * cell1.getWidth();
+        Double cell2Size = cell2.getHeight() * cell2.getWidth();
+        return cell1Size.compareTo(cell2Size);
+    };
+
+    public static final Comparator<Rectangle> RECTANGLE_SIZE_COMPARATOR = (rect1, rect2) -> {
+
+        Double rect1Size = rect1.getHeight() * rect1.getWidth();
+        Double rect2Size = rect2.getHeight() * rect2.getWidth();
+        return rect1Size.compareTo(rect2Size);
+    };
+
+    public static final Comparator<Ruling> X_FIRST_RULING_COMPARATOR = (ruling1, ruling2) -> {
+
+        int rv = 0;
+        float point1X = DoubleComparisons.round(Math.min(ruling1.getLeft(), ruling1.getRight()), COMPARATOR_ROUNDING);
+        float point1Y = DoubleComparisons.round(Math.min(ruling1.getTop(), ruling1.getBottom()), COMPARATOR_ROUNDING);
+        float point2X = DoubleComparisons.round(Math.min(ruling2.getLeft(), ruling2.getRight()), COMPARATOR_ROUNDING);
+        float point2Y = DoubleComparisons.round(Math.min(ruling2.getTop(), ruling2.getBottom()), COMPARATOR_ROUNDING);
+
+        if (point1X > point2X) {
+            rv = 1;
+        } else if (point1X < point2X) {
+            rv = -1;
+        } else if (point1Y > point2Y) {
+            rv = 1;
+        } else if (point1Y < point2Y) {
+            rv = -1;
+        }
+        return rv;
+    };
+
+}

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

@@ -21,11 +21,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Pa
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 
-import lombok.AccessLevel;
-import lombok.Builder;
-import lombok.Getter;
 import lombok.SneakyThrows;
-import lombok.experimental.FieldDefaults;
 import lombok.experimental.UtilityClass;
 import lombok.extern.slf4j.Slf4j;
 
@@ -41,20 +37,20 @@ public class PdfVisualisationUtility {
 
     public void drawNode(PDDocument document, DocumentTree.Entry entry) {
 
-        Options options = buildStandardOptionsForNodes(entry);
+        DrawingOptions options = buildStandardOptionsForNodes(entry);
 
         drawBBoxAndLabelAndNumberOnPage(document, entry, options);
 
     }
 
 
-    public void drawTextBlock(PDDocument document, TextBlock textBlock, Options options) {
+    public void drawTextBlock(PDDocument document, TextBlock textBlock, DrawingOptions options) {
 
         textBlock.getAtomicTextBlocks().forEach(atb -> drawAtomicTextBlock(document, atb, options));
     }
 
 
-    public void drawAtomicTextBlock(PDDocument document, AtomicTextBlock atomicTextBlock, Options options) {
+    public void drawAtomicTextBlock(PDDocument document, AtomicTextBlock atomicTextBlock, DrawingOptions options) {
 
         drawRectangle2DList(document, atomicTextBlock.getPage().getNumber(), atomicTextBlock.getPositions().stream().toList(), options);
 
@@ -62,7 +58,7 @@ public class PdfVisualisationUtility {
 
 
     @SneakyThrows
-    public void drawText(String string, PDDocument document, Point2D location, Integer pageNumber, Options options) {
+    public void drawText(String string, PDDocument document, Point2D location, Integer pageNumber, DrawingOptions options) {
 
         var pdPage = document.getPage(pageNumber - 1);
         var contentStream = new PDPageContentStream(document, pdPage, PDPageContentStream.AppendMode.APPEND, true);
@@ -80,14 +76,14 @@ public class PdfVisualisationUtility {
 
 
     @SneakyThrows
-    public void drawRectangle2DList(PDDocument document, int pageNumber, List<Rectangle2D> rectCollection, Options options) {
+    public void drawRectangle2DList(PDDocument document, int pageNumber, List<Rectangle2D> rectCollection, DrawingOptions options) {
 
         var pdPage = document.getPage(pageNumber - 1);
         drawRectangle2DList(document, rectCollection, options, pdPage);
     }
 
 
-    private void drawRectangle2DList(PDDocument document, List<Rectangle2D> rectCollection, Options options, PDPage pdPage) throws IOException {
+    private void drawRectangle2DList(PDDocument document, List<Rectangle2D> rectCollection, DrawingOptions options, PDPage pdPage) throws IOException {
 
         var contentStream = new PDPageContentStream(document, pdPage, PDPageContentStream.AppendMode.APPEND, true);
 
@@ -110,9 +106,9 @@ public class PdfVisualisationUtility {
     }
 
 
-    private Options buildStandardOptionsForNodes(DocumentTree.Entry entry) {
+    private DrawingOptions buildStandardOptionsForNodes(DocumentTree.Entry entry) {
 
-        return Options.builder().stroke(true).strokeColor(switch (entry.getType()) {
+        return DrawingOptions.builder().stroke(true).strokeColor(switch (entry.getType()) {
             case DOCUMENT -> Color.LIGHT_GRAY;
             case HEADER, FOOTER -> Color.GREEN;
             case PARAGRAPH -> Color.BLUE;
@@ -125,7 +121,7 @@ public class PdfVisualisationUtility {
     }
 
 
-    private void drawBBoxAndLabelAndNumberOnPage(PDDocument document, DocumentTree.Entry entry, Options options) {
+    private void drawBBoxAndLabelAndNumberOnPage(PDDocument document, DocumentTree.Entry entry, DrawingOptions options) {
 
         Map<Page, Rectangle2D> rectanglesPerPage = entry.getNode().getBBox();
         rectanglesPerPage.forEach((page, rectangle2D) -> {
@@ -152,7 +148,7 @@ public class PdfVisualisationUtility {
 
 
     @SneakyThrows
-    public static void drawLine2DList(PDDocument pdDocument, int pageNumber, List<? extends Line2D> line2DS, Options options) {
+    public static void drawLine2DList(PDDocument pdDocument, int pageNumber, List<? extends Line2D> line2DS, DrawingOptions options) {
 
         var pdPage = pdDocument.getPage(pageNumber - 1);
         var contentStream = new PDPageContentStream(pdDocument, pdPage, PDPageContentStream.AppendMode.APPEND, true);
@@ -176,21 +172,4 @@ public class PdfVisualisationUtility {
         contentStream.close();
     }
 
-
-    @Builder
-    @Getter
-    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
-    public static class Options {
-
-        boolean fill;
-        boolean stroke;
-        @Builder.Default
-        Color strokeColor = Color.BLACK;
-        @Builder.Default
-        float strokeWidth = 1f;
-        @Builder.Default
-        Color fillColor = Color.BLACK;
-
-    }
-
 }

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

@@ -2,8 +2,11 @@ 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;
@@ -19,6 +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;
@@ -37,15 +42,28 @@ public class RectangleTransformations {
     }
 
 
+    public static double calculateIntersectedArea(Rectangle2D r1, Rectangle2D r2) {
+
+        Area a1 = new Area(r1);
+        Area a2 = new Area(r2);
+        a1.intersect(a2);
+        Rectangle2D intersection = a1.getBounds2D();
+        return intersection.getWidth() * intersection.getHeight();
+    }
+
+
     public static Rectangle2D bBoxUnionAtomicTextBlock(List<AtomicTextBlock> atomicTextBlocks) {
 
         return atomicTextBlocks.stream().flatMap(atomicTextBlock -> atomicTextBlock.getPositions().stream()).collect(new Rectangle2DBBoxCollector());
     }
+
+
     public static Collector<Rectangle2D, Rectangle2DBBoxCollector.BBox, Rectangle2D> collectBBox() {
 
         return new Rectangle2DBBoxCollector();
     }
 
+
     public static PDRectangle toPDRectangleBBox(List<Rectangle> rectangles) {
 
         Rectangle2D rectangle2D = RectangleTransformations.rectangleBBox(rectangles);
@@ -70,6 +88,7 @@ public class RectangleTransformations {
         return format("%f,%f,%f,%f", rectangle2D.getX(), rectangle2D.getY(), rectangle2D.getWidth(), rectangle2D.getHeight());
     }
 
+
     public static Rectangle2D rectangleBBox(List<Rectangle> rectangles) {
 
         return rectangles.stream().map(RectangleTransformations::toRectangle2D).collect(new Rectangle2DBBoxCollector());
@@ -84,6 +103,7 @@ public class RectangleTransformations {
                 -redactionLogRectangle.getHeight());
     }
 
+
     public static Rectangle2D toRectangle2D(PDRectangle rectangle) {
 
         return new Rectangle2D.Double(rectangle.getLowerLeftX(), rectangle.getLowerLeftY(), rectangle.getWidth(), rectangle.getHeight());
@@ -133,7 +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/RectangularIntersectionFinder.java

@@ -0,0 +1,77 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.Y_FIRST_POINT_COMPARATOR;
+
+import java.awt.geom.Point2D;
+import java.awt.geom.Rectangle2D;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
+
+public class RectangularIntersectionFinder {
+
+    public static List<Rectangle2D> find(List<Ruling> horizontalRulingLines, List<Ruling> verticalRulingLines) {
+
+        // Fix for 211.pdf
+        for (Ruling r : horizontalRulingLines) {
+            if (r.getX2() < r.getX1()) {
+                double a = r.getX2();
+                r.x2 = (float) r.getX1();
+                r.x1 = (float) a;
+            }
+        }
+
+        List<Rectangle2D> foundRectangles = new ArrayList<>();
+        Map<Point2D, Ruling[]> intersectionPoints = Ruling.findIntersections(horizontalRulingLines, verticalRulingLines);
+        List<Point2D> intersectionPointsList = new ArrayList<>(intersectionPoints.keySet());
+        intersectionPointsList.sort(Y_FIRST_POINT_COMPARATOR);
+
+        for (int i = 0; i < intersectionPointsList.size(); i++) {
+            Point2D topLeft = intersectionPointsList.get(i);
+            Ruling[] hv = intersectionPoints.get(topLeft);
+
+            // CrossingPointsDirectlyBelow( topLeft );
+            List<Point2D> xPoints = new ArrayList<>();
+            // CrossingPointsDirectlyToTheRight( topLeft );
+            List<Point2D> yPoints = new ArrayList<>();
+
+            for (Point2D p : intersectionPointsList.subList(i, intersectionPointsList.size())) {
+                if (p.getX() == topLeft.getX() && p.getY() > topLeft.getY()) {
+                    xPoints.add(p);
+                }
+                if (p.getY() == topLeft.getY() && p.getX() > topLeft.getX()) {
+                    yPoints.add(p);
+                }
+            }
+            outer:
+            for (Point2D xPoint : xPoints) {
+                // is there a vertical edge b/w topLeft and xPoint?
+                if (!hv[1].equals(intersectionPoints.get(xPoint)[1])) {
+                    continue;
+                }
+                for (Point2D yPoint : yPoints) {
+                    // is there a horizontal edge b/w topLeft and yPoint ?
+                    if (!hv[0].equals(intersectionPoints.get(yPoint)[0])) {
+                        continue;
+                    }
+                    Point2D btmRight = new Point2D.Float((float) yPoint.getX(), (float) xPoint.getY());
+                    if (intersectionPoints.containsKey(btmRight)
+                        && intersectionPoints.get(btmRight)[0].equals(intersectionPoints.get(xPoint)[0])
+                        && intersectionPoints.get(btmRight)[1].equals(intersectionPoints.get(yPoint)[1])) {
+                        foundRectangles.add(new Rectangle2D.Double(topLeft.getX(), topLeft.getY(), btmRight.getX() - topLeft.getX(), btmRight.getY() - topLeft.getY()));
+                        break outer;
+                    }
+                }
+            }
+        }
+
+        // TODO create cells for vertical ruling lines with aligned endpoints at the top/bottom of a grid
+        // that aren't connected with an horizontal ruler?
+        // see: https://github.com/jazzido/tabula-extractor/issues/78#issuecomment-41481207
+
+        return foundRectangles;
+    }
+
+}

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

@@ -0,0 +1,172 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.X_FIRST_POINT_COMPARATOR;
+import static com.knecon.fforesight.service.layoutparser.processor.utils.GeometricComparators.Y_FIRST_POINT_COMPARATOR;
+
+import java.awt.geom.Point2D;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+
+import com.knecon.fforesight.service.layoutparser.processor.model.table.Rectangle;
+
+public class SpreadsheetFinder {
+
+    private static final int MAX_OUTER_POINT_TOLERANCE = 10;
+    private static final float AREA_TOLERANCE = 0.001f;
+
+
+    public static List<Rectangle> findSpreadsheetsFromCells(List<? extends Rectangle> cells) {
+        // via: http://stackoverflow.com/questions/13746284/merging-multiple-adjacent-rectangles-into-one-polygon
+        List<Rectangle> rectangles = new ArrayList<>();
+        Set<Point2D> pointSet = new HashSet<>();
+        Map<Point2D, Point2D> edgesH = new HashMap<>();
+        Map<Point2D, Point2D> edgesV = new HashMap<>();
+
+        for (Rectangle cell : cells) {
+            for (Point2D pt : cell.getPoints()) {
+                if (pointSet.contains(pt)) { // shared vertex, remove it
+                    pointSet.remove(pt);
+                } else {
+                    pointSet.add(pt);
+                }
+            }
+        }
+
+        // X first sort
+        List<Point2D> pointsSortX = new ArrayList<>(pointSet);
+        pointsSortX.sort(X_FIRST_POINT_COMPARATOR);
+        // Y first sort
+        List<Point2D> pointsSortY = new ArrayList<>(pointSet);
+        pointsSortY.sort(Y_FIRST_POINT_COMPARATOR);
+
+        int i = 0;
+        while (i < pointSet.size()) {
+            float currY = (float) pointsSortY.get(i).getY();
+            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortY.get(i).getY(), currY)) {
+                edgesH.put(pointsSortY.get(i), pointsSortY.get(i + 1));
+                edgesH.put(pointsSortY.get(i + 1), pointsSortY.get(i));
+                i += 2;
+            }
+        }
+
+        i = 0;
+        while (i < pointSet.size()) {
+            float currX = (float) pointsSortX.get(i).getX();
+            while (i < pointSet.size() && DoubleComparisons.feq(pointsSortX.get(i).getX(), currX)) {
+                edgesV.put(pointsSortX.get(i), pointsSortX.get(i + 1));
+                edgesV.put(pointsSortX.get(i + 1), pointsSortX.get(i));
+                i += 2;
+            }
+        }
+
+        // Get all the polygons
+        List<List<PolygonVertex>> polygons = new ArrayList<>();
+        Point2D nextVertex;
+        while (!edgesH.isEmpty()) {
+            ArrayList<PolygonVertex> polygon = new ArrayList<>();
+            Point2D first = edgesH.keySet()
+                    .iterator().next();
+            polygon.add(new PolygonVertex(first, Direction.HORIZONTAL));
+            edgesH.remove(first);
+
+            while (true) {
+                PolygonVertex curr = polygon.get(polygon.size() - 1);
+                PolygonVertex lastAddedVertex;
+                if (curr.direction == Direction.HORIZONTAL) {
+                    nextVertex = edgesV.get(curr.point);
+                    edgesV.remove(curr.point);
+                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.VERTICAL);
+                } else {
+                    nextVertex = edgesH.get(curr.point);
+                    edgesH.remove(curr.point);
+                    lastAddedVertex = new PolygonVertex(nextVertex, Direction.HORIZONTAL);
+                }
+                polygon.add(lastAddedVertex);
+
+                if (lastAddedVertex.equals(polygon.get(0))) {
+                    // closed polygon
+                    polygon.remove(polygon.size() - 1);
+                    break;
+                }
+            }
+
+            for (PolygonVertex vertex : polygon) {
+                edgesH.remove(vertex.point);
+                edgesV.remove(vertex.point);
+            }
+            polygons.add(polygon);
+        }
+
+        // calculate grid-aligned minimum area rectangles for each found polygon
+        for (List<PolygonVertex> poly : polygons) {
+            float top = Float.MAX_VALUE;
+            float left = Float.MAX_VALUE;
+            float bottom = Float.MIN_VALUE;
+            float right = Float.MIN_VALUE;
+            for (PolygonVertex pt : poly) {
+                top = (float) Math.min(top, pt.point.getY());
+                left = (float) Math.min(left, pt.point.getX());
+                bottom = (float) Math.max(bottom, pt.point.getY());
+                right = (float) Math.max(right, pt.point.getX());
+            }
+
+            // do not add polygons with too many outer points as they are unlikely to be tables
+            if (poly.size() <= MAX_OUTER_POINT_TOLERANCE) {
+                rectangles.add(new Rectangle(top - AREA_TOLERANCE, left - AREA_TOLERANCE, right - left + 2 * AREA_TOLERANCE, bottom - top + 2 * AREA_TOLERANCE));
+            }
+        }
+        return rectangles;
+    }
+
+
+    private enum Direction {
+        HORIZONTAL,
+        VERTICAL
+    }
+
+    static class PolygonVertex {
+
+        Point2D point;
+        Direction direction;
+
+
+        PolygonVertex(Point2D point, Direction direction) {
+
+            this.direction = direction;
+            this.point = point;
+        }
+
+
+        @Override
+        public boolean equals(Object other) {
+
+            if (this == other) {
+                return true;
+            }
+            if (!(other instanceof PolygonVertex)) {
+                return false;
+            }
+            return this.point.equals(((PolygonVertex) other).point);
+        }
+
+
+        @Override
+        public int hashCode() {
+
+            return this.point.hashCode();
+        }
+
+
+        @Override
+        public String toString() {
+
+            return String.format("%s[point=%s,direction=%s]", this.getClass().getName(), this.point.toString(), this.direction.toString());
+        }
+
+    }
+
+}

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

@@ -30,8 +30,6 @@ public class TableMergingUtility {
             if (consecutiveTable.getColCount() == originalTablePageBlock.getColCount() && !hasTableHeader(consecutiveTable) && outerBoundaryAlignsX(originalTablePageBlock,
                     consecutiveTable)) {
                 consecutiveTablesWithSameColCountAndHeaders.add(consecutiveTable);
-            } else {
-                break;
             }
         }
         return Stream.concat(Stream.of(originalTablePageBlock), consecutiveTablesWithSameColCountAndHeaders.stream()).toList();

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

@@ -0,0 +1,44 @@
+package com.knecon.fforesight.service.layoutparser.processor.utils;
+
+import java.util.HashMap;
+import java.util.Map;
+
+// simple implementation of a disjoint-set data structure
+// https://en.wikipedia.org/wiki/Disjoint-set_data_structure
+public class UnionFind<T> {
+
+    Map<T, T> parents = new HashMap<>();
+    Map<T, Integer> numberOfObjects = new HashMap<>();
+
+
+    public T find(T node) {
+
+        if (!parents.containsKey(node)) {
+            parents.put(node, node);
+            numberOfObjects.put(node, 1);
+        }
+        if (!node.equals(parents.get(node))) {
+            parents.put(node, find(parents.get(node)));
+        }
+        return parents.get(node);
+    }
+
+
+    public void union(T node1, T node2) {
+
+        T root1 = find(node1);
+        T root2 = find(node2);
+
+        if (!root1.equals(root2)) {
+            if (numberOfObjects.getOrDefault(root1, 1) < numberOfObjects.getOrDefault(root2, 1)) {
+                parents.put(root1, root2);
+                numberOfObjects.put(root2, numberOfObjects.get(root2) + numberOfObjects.get(root1));
+            } else {
+                parents.put(root2, root1);
+                numberOfObjects.put(root1, numberOfObjects.get(root1) + numberOfObjects.get(root2));
+            }
+        }
+    }
+
+}
+

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

@@ -29,6 +29,7 @@ import com.knecon.fforesight.service.layoutparser.processor.python_api.model.ima
 import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.TaasDocumentDataMapper;
+import com.knecon.fforesight.service.layoutparser.processor.utils.DrawingOptions;
 import com.knecon.fforesight.service.layoutparser.server.utils.AbstractTest;
 import com.knecon.fforesight.service.layoutparser.server.utils.visualizations.PdfDraw;
 
@@ -111,7 +112,7 @@ public class BdrJsonBuildTest extends AbstractTest {
 
         try (PDDocument pdDocument = Loader.loadPDF(file); var outputStream = new FileOutputStream(resultingFileName)) {
             PdfDraw.drawDocumentGraph(pdDocument, document);
-            PdfDraw.drawTextBlock(pdDocument, textBlock, PdfDraw.Options.builder().stroke(true).strokeWidth(0.1f).strokeColor(Color.YELLOW).build());
+            PdfDraw.drawTextBlock(pdDocument, textBlock, DrawingOptions.builder().stroke(true).strokeWidth(0.1f).strokeColor(Color.YELLOW).build());
             pdDocument.save(outputStream);
         }
     }

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/graph/DocumentGraphVisualizationTest.java

@@ -13,6 +13,7 @@ import org.springframework.core.io.ClassPathResource;
 
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Document;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
+import com.knecon.fforesight.service.layoutparser.processor.utils.DrawingOptions;
 import com.knecon.fforesight.service.layoutparser.server.utils.BuildDocumentTest;
 import com.knecon.fforesight.service.layoutparser.server.utils.visualizations.PdfDraw;
 
@@ -70,7 +71,7 @@ public class DocumentGraphVisualizationTest extends BuildDocumentTest {
         try (PDDocument pdDocument = Loader.loadPDF(fileResource.getFile())) {
             log.info("drawing document");
             PdfDraw.drawDocumentGraph(pdDocument, documentGraph);
-            PdfDraw.drawTextBlock(pdDocument, textBlock, PdfDraw.Options.builder().stroke(true).strokeWidth(0.1f).strokeColor(Color.YELLOW).build());
+            PdfDraw.drawTextBlock(pdDocument, textBlock, DrawingOptions.builder().stroke(true).strokeWidth(0.1f).strokeColor(Color.YELLOW).build());
             log.info("saving document");
             pdDocument.save(tmpFile);
             log.info("saved document");

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

@@ -25,7 +25,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
     @SneakyThrows
     public void testViewerDocument() {
 
-        String fileName = "files/S-Metolachlor_RAR_01_Volume_1_2018-09-06.pdf";
+        String fileName = "files/SinglePages/T5 VV-640252-Page16.pdf";
         String tmpFileName = "/tmp/" + Path.of(fileName).getFileName() + "_VIEWER.pdf";
 
         var documentFile = new ClassPathResource(fileName).getFile();
@@ -35,9 +35,10 @@ public class ViewerDocumentTest extends BuildDocumentTest {
         Document document = buildGraph(fileName, LayoutParsingType.REDACT_MANAGER);
         long start = System.currentTimeMillis();
         layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true);
-        System.out.printf("Total time: %.2fs%n",  ((float) (System.currentTimeMillis() - start)) / 1000);
+        System.out.printf("Total time: %.2fs%n", ((float) (System.currentTimeMillis() - start)) / 1000);
     }
 
+
     @Test
     @Disabled
     @SneakyThrows
@@ -51,7 +52,11 @@ public class ViewerDocumentTest extends BuildDocumentTest {
         var tableResponse = mapper.readValue(new ClassPathResource(tableFileName).getInputStream(), TableServiceResponse.class);
         var documentFile = new ClassPathResource(fileName).getFile();
 
-        var classificationDocument = layoutParsingPipeline.parseLayout(LayoutParsingType.DOCUMINE, documentFile, new ImageServiceResponse(), tableResponse, Path.of(fileName).getFileName().toFile().toString());
+        var classificationDocument = layoutParsingPipeline.parseLayout(LayoutParsingType.DOCUMINE,
+                                                                       documentFile,
+                                                                       new ImageServiceResponse(),
+                                                                       tableResponse,
+                                                                       Path.of(fileName).getFileName().toFile().toString());
         ViewerDocumentService viewerDocumentService = new ViewerDocumentService(null);
         LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
         Document document = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
@@ -60,3 +65,4 @@ public class ViewerDocumentTest extends BuildDocumentTest {
     }
 
 }
+

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;
@@ -25,16 +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.python_api.adapter.CvTableParsingAdapter;
-import com.knecon.fforesight.service.layoutparser.processor.python_api.adapter.ImageServiceResponseAdapter;
+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;
@@ -50,12 +55,6 @@ public class PdfSegmentationServiceTest extends AbstractTest {
     @Autowired
     private RedactManagerClassificationService redactManagerClassificationService;
 
-    @Autowired
-    private CvTableParsingAdapter cvTableParsingAdapter;
-
-    @Autowired
-    private ImageServiceResponseAdapter imageServiceResponseAdapter;
-
     @Autowired
     private SectionsBuilderService sectionsBuilderService;
 
@@ -64,10 +63,10 @@ public class PdfSegmentationServiceTest extends AbstractTest {
     public ClassificationDocument buildClassificationDocument(File originDocument, TableServiceResponse tableServiceResponse) {
 
         ClassificationDocument classificationDocument = layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
-                originDocument,
-                new ImageServiceResponse(),
-                tableServiceResponse,
-                "document");
+                                                                                          originDocument,
+                                                                                          new ImageServiceResponse(),
+                                                                                          tableServiceResponse,
+                                                                                          "document");
 
         redactManagerClassificationService.classifyDocument(classificationDocument);
 
@@ -87,11 +86,11 @@ public class PdfSegmentationServiceTest extends AbstractTest {
     @Test
     public void tablesToHtmlDebugger() throws IOException {
 
-        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/A20622A izRMS (CZ) fRR Part B9_Page185.pdf");
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T5 VV-640252-Page16.pdf");
 
         ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
 
-        toHtml(document, "/tmp/A20622A izRMS (CZ) fRR Part B9_Page185.html");
+        toHtml(document, "/tmp/T5.html");
 
     }
 
@@ -109,6 +108,7 @@ public class PdfSegmentationServiceTest extends AbstractTest {
     }
 
 
+    @Disabled
     @Test
     public void testScanRotationBorderIsIgnored() throws IOException {
 
@@ -117,8 +117,16 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         var tableServiceResponse = objectMapper.readValue(cvTablesResource.getInputStream(), TableServiceResponse.class);
 
         ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile(), tableServiceResponse);
-        assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList())).isNotEmpty();
-        var tables = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).toList();
+        assertThat(document.getSections()
+                           .stream()
+                           .flatMap(paragraph -> paragraph.getTables()
+                                   .stream())
+                           .collect(Collectors.toList())).isNotEmpty();
+        var tables = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList();
 
         // Quality of the table parsing is not good, because the file is rotated at scanning.
         // We only asset that the table border is not the page border.
@@ -140,12 +148,12 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         imageServiceResponse.getData()
                 .forEach(imageMetadata -> images.computeIfAbsent(imageMetadata.getPosition().getPageNumber(), x -> new ArrayList<>())
                         .add(new ClassifiedImage(new Rectangle2D.Double(imageMetadata.getPosition().getX1(),
-                                imageMetadata.getPosition().getY1(),
-                                imageMetadata.getGeometry().getWidth(),
-                                imageMetadata.getGeometry().getHeight()),
-                                ImageType.valueOf(imageMetadata.getClassification().getLabel().toUpperCase(Locale.ROOT)),
-                                imageMetadata.isAlpha(),
-                                imageMetadata.getPosition().getPageNumber())));
+                                                                        imageMetadata.getPosition().getY1(),
+                                                                        imageMetadata.getGeometry().getWidth(),
+                                                                        imageMetadata.getGeometry().getHeight()),
+                                                 ImageType.valueOf(imageMetadata.getClassification().getLabel().toUpperCase(Locale.ROOT)),
+                                                 imageMetadata.isAlpha(),
+                                                 imageMetadata.getPosition().getPageNumber())));
 
         System.out.println("object");
     }
@@ -157,11 +165,22 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         ClassPathResource pdfFileResource = new ClassPathResource("files/Minimal Examples/Spanning Cells.pdf");
 
         ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
-        assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList())).isNotEmpty();
-        TablePageBlock table = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(0);
+        assertThat(document.getSections()
+                           .stream()
+                           .flatMap(paragraph -> paragraph.getTables()
+                                   .stream())
+                           .collect(Collectors.toList())).isNotEmpty();
+        TablePageBlock table = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(0);
         assertThat(table.getColCount()).isEqualTo(6);
         assertThat(table.getRowCount()).isEqualTo(13);
-        assertThat(table.getRows().stream().mapToInt(List::size).sum()).isEqualTo(6 * 13);
+        assertThat(table.getRows()
+                           .stream()
+                           .mapToInt(List::size).sum()).isEqualTo(6 * 13);
     }
 
 
@@ -171,15 +190,37 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         ClassPathResource pdfFileResource = new ClassPathResource("files/Minimal Examples/Merge Table.pdf");
 
         ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
-        assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList())).isNotEmpty();
-        TablePageBlock firstTable = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(0);
+        assertThat(document.getSections()
+                           .stream()
+                           .flatMap(paragraph -> paragraph.getTables()
+                                   .stream())
+                           .collect(Collectors.toList())).isNotEmpty();
+        TablePageBlock firstTable = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(0);
         assertThat(firstTable.getColCount()).isEqualTo(8);
         assertThat(firstTable.getRowCount()).isEqualTo(1);
-        TablePageBlock secondTable = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(1);
+        TablePageBlock secondTable = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(1);
         assertThat(secondTable.getColCount()).isEqualTo(8);
         assertThat(secondTable.getRowCount()).isEqualTo(2);
-        List<List<Cell>> firstTableHeaderCells = firstTable.getRows().get(0).stream().map(Collections::singletonList).collect(Collectors.toList());
-        assertThat(secondTable.getRows().stream().allMatch(row -> row.stream().map(Cell::getHeaderCells).toList().equals(firstTableHeaderCells))).isTrue();
+        List<List<Cell>> firstTableHeaderCells = firstTable.getRows()
+                .get(0)
+                .stream()
+                .map(Collections::singletonList)
+                .collect(Collectors.toList());
+        assertThat(secondTable.getRows()
+                           .stream()
+                           .allMatch(row -> row.stream()
+                                   .map(Cell::getHeaderCells)
+                                   .toList().equals(firstTableHeaderCells))).isTrue();
     }
 
 
@@ -189,15 +230,37 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         ClassPathResource pdfFileResource = new ClassPathResource("files/Minimal Examples/Merge Multi Page Table.pdf");
 
         ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
-        assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList())).isNotEmpty();
-        TablePageBlock firstTable = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(0);
+        assertThat(document.getSections()
+                           .stream()
+                           .flatMap(paragraph -> paragraph.getTables()
+                                   .stream())
+                           .collect(Collectors.toList())).isNotEmpty();
+        TablePageBlock firstTable = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(0);
         assertThat(firstTable.getColCount()).isEqualTo(9);
         assertThat(firstTable.getRowCount()).isEqualTo(5);
-        TablePageBlock secondTable = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(1);
+        TablePageBlock secondTable = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(1);
         assertThat(secondTable.getColCount()).isEqualTo(9);
         assertThat(secondTable.getRowCount()).isEqualTo(6);
-        List<List<Cell>> firstTableHeaderCells = firstTable.getRows().get(firstTable.getRowCount() - 1).stream().map(Cell::getHeaderCells).collect(Collectors.toList());
-        assertThat(secondTable.getRows().stream().allMatch(row -> row.stream().map(Cell::getHeaderCells).toList().equals(firstTableHeaderCells))).isTrue();
+        List<List<Cell>> firstTableHeaderCells = firstTable.getRows()
+                .get(firstTable.getRowCount() - 1)
+                .stream()
+                .map(Cell::getHeaderCells)
+                .collect(Collectors.toList());
+        assertThat(secondTable.getRows()
+                           .stream()
+                           .allMatch(row -> row.stream()
+                                   .map(Cell::getHeaderCells)
+                                   .toList().equals(firstTableHeaderCells))).isTrue();
     }
 
 
@@ -207,19 +270,41 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         ClassPathResource pdfFileResource = new ClassPathResource("files/Minimal Examples/Rotated Table Headers.pdf");
 
         ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
-        assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList())).isNotEmpty();
-        TablePageBlock firstTable = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(0);
+        assertThat(document.getSections()
+                           .stream()
+                           .flatMap(paragraph -> paragraph.getTables()
+                                   .stream())
+                           .collect(Collectors.toList())).isNotEmpty();
+        TablePageBlock firstTable = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(0);
         assertThat(firstTable.getColCount()).isEqualTo(8);
         assertThat(firstTable.getRowCount()).isEqualTo(1);
-        TablePageBlock secondTable = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList()).get(1);
+        TablePageBlock secondTable = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(1);
         assertThat(secondTable.getColCount()).isEqualTo(8);
         assertThat(secondTable.getRowCount()).isEqualTo(6);
-        List<List<Cell>> firstTableHeaderCells = firstTable.getRows().get(0).stream().map(Collections::singletonList).collect(Collectors.toList());
-        assertThat(secondTable.getRows().stream().allMatch(row -> row.stream().map(Cell::getHeaderCells).toList().equals(firstTableHeaderCells))).isTrue();
+        List<List<Cell>> firstTableHeaderCells = firstTable.getRows()
+                .get(0)
+                .stream()
+                .map(Collections::singletonList)
+                .collect(Collectors.toList());
+        assertThat(secondTable.getRows()
+                           .stream()
+                           .allMatch(row -> row.stream()
+                                   .map(Cell::getHeaderCells)
+                                   .toList().equals(firstTableHeaderCells))).isTrue();
     }
 
 
-    @Test // Non-sense test
+    @Test
     public void testDoc56Page170() throws IOException {
 
         ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/56 Fludioxonil_RAR_12_Volume_3CA_B-7_2018-02-21_Page170.pdf");
@@ -230,8 +315,8 @@ public class PdfSegmentationServiceTest extends AbstractTest {
 
         validateTable(document, 0, 1, 1, 0, 0);
         validateTable(document, 1, 2, 2, 0, 0);
-        validateTable(document, 2, 6, 20, 0, 0);
-        validateTable(document, 3, 7, 31, 0, 0);
+        validateTable(document, 2, 4, 19, 12, 0);
+        validateTable(document, 3, 2, 12, 0, 0);
 
     }
 
@@ -265,29 +350,30 @@ public class PdfSegmentationServiceTest extends AbstractTest {
         validateTable(document, 0, 8, 8, 0, 0);
 
         List<List<String>> values = Arrays.asList(Arrays.asList("Annex point Reference within DAR/RAR",
-                        "Author, date",
-                        "Study title",
-                        "Analytical method Author, date, No.",
-                        "Technique, LOQ of the method, validated working range",
-                        "Method meets analytical validation criteria",
-                        "Remarks (in case validation criteria are not met)",
-                        "Acceptability of the method"),
-                Arrays.asList("Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
-                        "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies"),
-                Arrays.asList("CA 7.1.2.1.1 DAR (2009)",
-                        "Evans P.G. 2001 TMJ4569B, VV-323245",
-                        "Azoxystrobin Laboratory Degradation Study in Three Soil Types, Sampled from Holland and the United Kingdom",
-                        "Method: RAM 269 Johnson R.I., Tummon O.J., Earl M. 1995 RJ1864B, VV-377731 Johnson R.I., Tummon O.J., Earl M. 1998 RAM 269/02, VV-124072 Johnson R.I., Tummon O.J., Earl M. 2000 RAM 269/03, VV-123986 Validation: Robinson N.J. 2001 TMJ4617B, VV-895845",
-                        "LC-MS/MS LOQ: 0.01 mg/kg (R401553 (SYN50165 7), R402173 (SYN501114 )) or 0.02 mg/kg (azoxystrobin, R230310, R234886) Working range: 0.02-1.0 or 0.01-0.5 mg/kg (depending on analyte) Other supporting quantificati on methods: HPLC-UV GC-MSD",
-                        "Y",
-                        "N/A",
-                        "Y"));
+                                                                "Author, date",
+                                                                "Study title",
+                                                                "Analytical method Author, date, No.",
+                                                                "Technique, LOQ of the method, validated working range",
+                                                                "Method meets analytical validation criteria",
+                                                                "Remarks (in case validation criteria are not met)",
+                                                                "Acceptability of the method"),
+                                                  Arrays.asList(
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies",
+                                                          "Part (a) Methods in soil, water, sediment, air and any additional matrices used in support of environmental fate studies"),
+                                                  Arrays.asList("CA 7.1.2.1.1 DAR (2009)",
+                                                                "Evans P.G. 2001 TMJ4569B, VV-323245",
+                                                                "Azoxystrobin Laboratory Degradation Study in Three Soil Types, Sampled from Holland and the United Kingdom",
+                                                                "Method: RAM 269 Johnson R.I., Tummon O.J., Earl M. 1995 RJ1864B, VV-377731 Johnson R.I., Tummon O.J., Earl M. 1998 RAM 269/02, VV-124072 Johnson R.I., Tummon O.J., Earl M. 2000 RAM 269/03, VV-123986 Validation: Robinson N.J. 2001 TMJ4617B, VV-895845",
+                                                                "LC-MS/MS LOQ: 0.01 mg/kg (R401553 (SYN50165 7), R402173 (SYN501114 )) or 0.02 mg/kg (azoxystrobin, R230310, R234886) Working range: 0.02-1.0 or 0.01-0.5 mg/kg (depending on analyte) Other supporting quantificati on methods: HPLC-UV GC-MSD",
+                                                                "Y",
+                                                                "N/A",
+                                                                "Y"));
 
         validateTable(document, 0, values);
 
@@ -579,10 +665,156 @@ public class PdfSegmentationServiceTest extends AbstractTest {
     }
 
 
+    @Test
+    public void testT0() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T0 TableWithMergedCells.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 1);
+
+        validateTable(document, 0, 6, 8, 0, 0);
+    }
+
+
+    @Test
+    public void testT1() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T1 MultipleNestedTable.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 4);
+
+        validateTable(document, 0, 3, 3, 0, 0);
+        validateTable(document, 1, 3, 6, 2, 0);
+        validateTable(document, 2, 3, 3, 1, 0);
+        validateTable(document, 3, 3, 3, 0, 0);
+
+    }
+
+
+    @Test
+    public void testT2() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T2 MultipleTables.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 6);
+
+        validateTable(document, 0, 5, 5, 0, 0);
+        validateTable(document, 1, 5, 6, 0, 0);
+        validateTable(document, 2, 5, 5, 0, 0);
+        validateTable(document, 3, 5, 5, 0, 0);
+        validateTable(document, 4, 5, 5, 0, 0);
+        validateTable(document, 5, 5, 5, 0, 0);
+
+    }
+
+
+    @Test
+    public void testT3() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T3 S-Meto_Page29.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 1);
+
+        validateTable(document, 0, 6, 5, 0, 0);
+
+    }
+
+
+    @Test
+    public void testT4() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T4 138 IDD0000261736_Page16.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 1);
+
+        validateTable(document, 0, 5, 8, 1, 0);
+
+    }
+
+
+    @Test
+    public void testT5() throws IOException {
+
+        ClassPathResource pdfFileResource = new ClassPathResource("files/SinglePages/T5 VV-640252-Page16.pdf");
+
+        ClassificationDocument document = buildClassificationDocument(pdfFileResource.getFile());
+
+        validateTableSize(document, 5);
+        validateTable(document, 0, 1, 1, 0, 0);
+        validateTable(document, 1, 1, 1, 0, 0);
+        validateTable(document, 2, 1, 1, 0, 0);
+        validateTable(document, 3, 1, 1, 0, 0);
+        validateTable(document, 4, 1, 1, 0, 0);
+
+    }
+
+
+    @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) {
 
-        var tables = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).toList();
+        var tables = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList();
         StringBuilder sb = new StringBuilder();
 
         int currentPage = 1;
@@ -603,9 +835,19 @@ public class PdfSegmentationServiceTest extends AbstractTest {
 
     private void validateTable(ClassificationDocument document, int tableIndex, int colCount, int rowCount, int emptyCellsCountCorrect, int emptyCellsCountIncorrect) {
 
-        TablePageBlock table = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).toList().get(tableIndex);
+        TablePageBlock table = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(tableIndex);
         List<List<Cell>> rows = table.getRows();
-        int emptyCellsFoundFound = rows.stream().flatMap(List::stream).toList().stream().filter(f -> f.toString().equals("")).toList().size();
+        int emptyCellsFoundFound = rows.stream()
+                .flatMap(List::stream)
+                .toList()
+                .stream()
+                .filter(f -> f.toString().isEmpty())
+                .toList().size();
 
         for (List<Cell> row : table.getRows()) {
             row.forEach(r -> System.out.println(r.toString()));
@@ -620,11 +862,20 @@ public class PdfSegmentationServiceTest extends AbstractTest {
 
     private void validateTable(ClassificationDocument document, int tableIndex, List<List<String>> values) {
 
-        TablePageBlock table = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).toList().get(tableIndex);
+        TablePageBlock table = document.getSections()
+                .stream()
+                .flatMap(paragraph -> paragraph.getTables()
+                        .stream())
+                .toList()
+                .get(tableIndex);
         List<List<Cell>> rows = table.getRows();
 
-        List<Cell> rowsFlattened = rows.stream().flatMap(List::stream).toList();
-        List<String> valuesFlattened = values.stream().flatMap(List::stream).toList();
+        List<Cell> rowsFlattened = rows.stream()
+                .flatMap(List::stream)
+                .toList();
+        List<String> valuesFlattened = values.stream()
+                .flatMap(List::stream)
+                .toList();
 
         for (int i = 0; i < valuesFlattened.size(); i++) {
             Cell cell = rowsFlattened.get(i);
@@ -637,7 +888,11 @@ public class PdfSegmentationServiceTest extends AbstractTest {
 
     private void validateTableSize(ClassificationDocument document, int tableSize) {
 
-        assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).toList().size()).isEqualTo(tableSize);
+        assertThat(document.getSections()
+                           .stream()
+                           .flatMap(paragraph -> paragraph.getTables()
+                                   .stream())
+                           .toList().size()).isEqualTo(tableSize);
 
     }
 

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

@@ -1,13 +1,17 @@
 package com.knecon.fforesight.service.layoutparser.server.services;
 
+import java.awt.Color;
+import java.awt.geom.Rectangle2D;
 import java.nio.file.Files;
 import java.nio.file.Path;
 import java.util.Collections;
 import java.util.LinkedList;
 import java.util.List;
+import java.util.stream.Collectors;
 
 import org.apache.pdfbox.Loader;
 import org.apache.pdfbox.pdmodel.PDDocument;
+import org.junit.jupiter.api.Disabled;
 import org.junit.jupiter.api.Test;
 import org.springframework.core.io.ClassPathResource;
 
@@ -26,29 +30,50 @@ import com.knecon.fforesight.service.layoutparser.processor.services.RulingClean
 import com.knecon.fforesight.service.layoutparser.processor.services.factory.DocumentGraphFactory;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.DocumentDataMapper;
 import com.knecon.fforesight.service.layoutparser.processor.services.mapper.PropertiesMapper;
-import com.knecon.fforesight.service.layoutparser.processor.services.visualization.LayoutGridService;
+import com.knecon.fforesight.service.layoutparser.processor.utils.DrawingOptions;
+import com.knecon.fforesight.service.layoutparser.processor.utils.RectangularIntersectionFinder;
 import com.knecon.fforesight.service.layoutparser.server.utils.BuildDocumentTest;
 import com.knecon.fforesight.service.layoutparser.server.utils.visualizations.PdfDraw;
-import com.knecon.fforesight.service.viewerdoc.service.ViewerDocumentService;
 
 import lombok.SneakyThrows;
 
 public class RulingCleaningServiceTest extends BuildDocumentTest {
 
     @Test
-//    @Disabled
+    @Disabled
+    @SneakyThrows
+    public void textRectanglesFromRulingsExtraction() {
+
+        String fileName = "files/SinglePages/T5 VV-640252-Page16.pdf";
+        String lineFileName = "/tmp/" + Path.of(fileName).getFileName().toString() + "_CELLS.pdf";
+        List<PageContents> pageContents = PageContentExtractor.getSortedPageContents(fileName);
+        RulingCleaningService rulingCleaningService = new RulingCleaningService();
+        List<List<Rectangle2D>> rectanglesPerPage = new LinkedList<>();
+        for (PageContents pageContent : pageContents) {
+            CleanRulings cleanRulings = rulingCleaningService.getCleanRulings(Collections.emptyList(), pageContent.getRulings());
+            List<Rectangle2D> rects = RectangularIntersectionFinder.find(cleanRulings.getHorizontal(), cleanRulings.getVertical());
+            rectanglesPerPage.add(rects);
+        }
+
+        PdfDraw.drawRectanglesPerPage(fileName, rectanglesPerPage, lineFileName, DrawingOptions.builder().stroke(true).strokeColor(Color.RED).build());
+    }
+
+
+    @Test
+    @Disabled
     @SneakyThrows
     public void textRulingExtraction() {
 
-        String fileName = "files/211.pdf";
+        String fileName = "files/SinglePages/19 Chlorothalonil RAR 08 Volume 3CA B 6b metabolites Oct 2017_Page35.pdf";
         String lineFileName = "/tmp/" + Path.of(fileName).getFileName().toString() + "_LINES.pdf";
         List<PageContents> pageContents = PageContentExtractor.getSortedPageContents(fileName);
         RulingCleaningService rulingCleaningService = new RulingCleaningService();
-        PdfDraw.drawLinesPerPage(fileName, pageContents.stream().map(PageContents::getRulings).toList(), lineFileName);
         List<CleanRulings> cleanRulingsPerPage = new LinkedList<>();
         for (PageContents pageContent : pageContents) {
             cleanRulingsPerPage.add(rulingCleaningService.getCleanRulings(Collections.emptyList(), pageContent.getRulings()));
         }
+        var cleanRulings = cleanRulingsPerPage.stream().map(CleanRulings::getVertical).collect(Collectors.toList());
+        PdfDraw.drawLinesPerPage(fileName,  cleanRulings, lineFileName);
 
     }
 
@@ -57,9 +82,6 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
     @SneakyThrows
     public void testTableExtraction() {
 
-        ViewerDocumentService viewerDocumentService = new ViewerDocumentService(null);
-        LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
-
         ClassPathResource resource = new ClassPathResource("files");
         List<String> pdfFileNames = Files.walk(resource.getFile().toPath())
                 .filter(path -> path.getFileName().toString().endsWith(".pdf"))
@@ -67,8 +89,8 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
                 .map(Path::toString)
                 .toList();
 
-        for (int i = 0; i < pdfFileNames.size(); i++) {
-            writeJsons(Path.of(pdfFileNames.get(i)));
+        for (String pdfFileName : pdfFileNames) {
+            writeJsons(Path.of(pdfFileName));
         }
     }
 
@@ -88,13 +110,13 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
                 filename.toFile().toString()));
         DocumentData documentDataBefore = DocumentDataMapper.toDocumentData(documentGraphBefore);
         DocumentData documentDataAfter = DocumentDataMapper.toDocumentData(documentGraphAfter);
-        if (!compareStructures(documentDataBefore.getDocumentStructure(), documentDataAfter.getDocumentStructure(), filename.getFileName().toString())) {
-            String tmpFileNameBefore = "C:/Users/YANNIK~1/AppData/Local/Temp/before." + filename.getFileName().toString();
+        if (!compareStructures(documentDataBefore.getDocumentStructure(), documentDataAfter.getDocumentStructure())) {
+            String tmpFileNameBefore = "/tmp/before." + filename.getFileName().toString();
             try (PDDocument pdDocument = Loader.loadPDF(filename.toFile())) {
                 PdfDraw.drawDocumentGraph(pdDocument, documentGraphBefore);
                 pdDocument.save(tmpFileNameBefore);
             }
-            String tmpFileNameAfter = "C:/Users/YANNIK~1/AppData/Local/Temp/after." + filename.getFileName().toString();
+            String tmpFileNameAfter = "/tmp/after." + filename.getFileName().toString();
             try (PDDocument pdDocument = Loader.loadPDF(filename.toFile())) {
                 PdfDraw.drawDocumentGraph(pdDocument, documentGraphAfter);
                 pdDocument.save(tmpFileNameAfter);
@@ -105,9 +127,9 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
 
 
     @SneakyThrows
-    private boolean compareStructures(DocumentStructure structure1, DocumentStructure structure2, String pdfName) {
+    private boolean compareStructures(DocumentStructure structure1, DocumentStructure structure2) {
 
-        List listStructure1 = structure1.streamAllEntries()
+        List<Table> listStructure1 = structure1.streamAllEntries()
                 .filter(entryData -> entryData.getType().equals(NodeType.TABLE))
                 .map(DocumentStructure.EntryData::getProperties)
                 .map(properties -> {
@@ -117,7 +139,7 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
                 })
                 .toList();
 
-        List listStructure2 = structure2.streamAllEntries()
+        List<Table> listStructure2 = structure2.streamAllEntries()
                 .filter(entryData -> entryData.getType().equals(NodeType.TABLE))
                 .map(DocumentStructure.EntryData::getProperties)
                 .map(properties -> {
@@ -128,8 +150,8 @@ public class RulingCleaningServiceTest extends BuildDocumentTest {
                 .toList();
 
         for (int i = 0; i < listStructure1.size(); i++) {
-            Table tableNode1 = (Table) listStructure1.get(i);
-            Table tableNode2 = (Table) listStructure2.get(i);
+            Table tableNode1 = listStructure1.get(i);
+            Table tableNode2 = listStructure2.get(i);
             if (tableNode1.getNumberOfRows() != tableNode2.getNumberOfRows() || tableNode1.getNumberOfCols() != tableNode2.getNumberOfCols()) {
                 return false;
             }

layoutparser-service/layoutparser-service-server/src/test/java/com/knecon/fforesight/service/layoutparser/server/utils/visualizations/PdfDraw.java

@@ -24,20 +24,31 @@ import com.knecon.fforesight.service.layoutparser.processor.model.graph.nodes.Pa
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.AtomicTextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.graph.textblock.TextBlock;
 import com.knecon.fforesight.service.layoutparser.processor.model.table.Ruling;
+import com.knecon.fforesight.service.layoutparser.processor.utils.DrawingOptions;
 import com.knecon.fforesight.service.layoutparser.processor.utils.PdfVisualisationUtility;
 import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
 
-import lombok.AccessLevel;
-import lombok.AllArgsConstructor;
-import lombok.Builder;
-import lombok.Getter;
 import lombok.SneakyThrows;
-import lombok.experimental.FieldDefaults;
 import lombok.experimental.UtilityClass;
 
 @UtilityClass
 public class PdfDraw {
 
+    public static void drawRectanglesPerPage(String filename, List<List<Rectangle2D>> rectanglesPerPage, String tmpFileName, DrawingOptions options) throws IOException {
+
+        ClassPathResource pdfResource = new ClassPathResource(filename);
+        try (PDDocument pdDocument = Loader.loadPDF(pdfResource.getFile()); var out = new FileOutputStream(tmpFileName)) {
+            for (int pageNumber = 1; pageNumber < pdDocument.getNumberOfPages() + 1; pageNumber++) {
+                PdfVisualisationUtility.drawRectangle2DList(pdDocument,
+                        pageNumber,
+                        rectanglesPerPage.get(pageNumber - 1),
+                        options);
+            }
+            pdDocument.save(out);
+        }
+
+    }
+
     public static void drawRectanglesPerPage(String filename, List<List<Rectangle2D>> rectanglesPerPage, String tmpFileName) throws IOException {
 
         ClassPathResource pdfResource = new ClassPathResource(filename);
@@ -46,7 +57,7 @@ public class PdfDraw {
                 PdfVisualisationUtility.drawRectangle2DList(pdDocument,
                         pageNumber,
                         rectanglesPerPage.get(pageNumber - 1),
-                        PdfVisualisationUtility.Options.builder().stroke(true).build());
+                        DrawingOptions.builder().stroke(true).build());
             }
             pdDocument.save(out);
         }
@@ -62,13 +73,13 @@ public class PdfDraw {
                 var rectanglesOnPage = rectanglesPerPage.get(pageNumber - 1);
                 for (int lineNumber = 0; lineNumber < rectanglesOnPage.size(); lineNumber++) {
                     var rectanglesInLine = rectanglesOnPage.get(lineNumber);
-                    PdfVisualisationUtility.drawRectangle2DList(pdDocument, pageNumber, rectanglesInLine, PdfVisualisationUtility.Options.builder().stroke(true).build());
+                    PdfVisualisationUtility.drawRectangle2DList(pdDocument, pageNumber, rectanglesInLine, DrawingOptions.builder().stroke(true).build());
                     double y = Math.min(rectanglesInLine.get(0).getMinY(), rectanglesInLine.get(0).getMaxY());
                     PdfVisualisationUtility.drawText(String.format("%d", lineNumber),
                             pdDocument,
                             new Point2D.Double(rectanglesInLine.get(0).getX() - (5 + (5 * countNumberOfDigits(lineNumber))), y + 2),
                             pageNumber,
-                            PdfVisualisationUtility.Options.builder().stroke(true).build());
+                            DrawingOptions.builder().stroke(true).build());
                 }
             }
             pdDocument.save(out);
@@ -99,20 +110,20 @@ public class PdfDraw {
 
     public static void drawNode(PDDocument document, DocumentTree.Entry entry) {
 
-        Options options = buildStandardOptionsForNodes(entry);
+        DrawingOptions options = buildStandardOptionsForNodes(entry);
 
         drawBBoxAndLabelAndNumberOnPage(document, entry, options);
 
     }
 
 
-    public static void drawTextBlock(PDDocument document, TextBlock textBlock, Options options) {
+    public static void drawTextBlock(PDDocument document, TextBlock textBlock, DrawingOptions options) {
 
         textBlock.getAtomicTextBlocks().forEach(atb -> drawAtomicTextBlock(document, atb, options));
     }
 
 
-    public static void drawAtomicTextBlock(PDDocument document, AtomicTextBlock atomicTextBlock, Options options) {
+    public static void drawAtomicTextBlock(PDDocument document, AtomicTextBlock atomicTextBlock, DrawingOptions options) {
 
         drawRectangle2DList(document, atomicTextBlock.getPage().getNumber(), atomicTextBlock.getPositions().stream().toList(), options);
 
@@ -120,7 +131,7 @@ public class PdfDraw {
 
 
     @SneakyThrows
-    private static void drawText(String string, PDDocument document, Point2D location, Integer pageNumber, Options options, boolean rotate) {
+    private static void drawText(String string, PDDocument document, Point2D location, Integer pageNumber, DrawingOptions options, boolean rotate) {
 
         var pdPage = document.getPage(pageNumber - 1);
         var contentStream = new PDPageContentStream(document, pdPage, PDPageContentStream.AppendMode.APPEND, true);
@@ -142,14 +153,14 @@ public class PdfDraw {
 
 
     @SneakyThrows
-    public static void drawRectangle2DList(PDDocument document, int pageNumber, List<Rectangle2D> rectCollection, Options options) {
+    public static void drawRectangle2DList(PDDocument document, int pageNumber, List<Rectangle2D> rectCollection, DrawingOptions options) {
 
         var pdPage = document.getPage(pageNumber - 1);
         drawRectangle2DList(document, rectCollection, options, pdPage);
     }
 
 
-    private static void drawRectangle2DList(PDDocument document, List<Rectangle2D> rectCollection, Options options, PDPage pdPage) throws IOException {
+    private static void drawRectangle2DList(PDDocument document, List<Rectangle2D> rectCollection, DrawingOptions options, PDPage pdPage) throws IOException {
 
         var contentStream = new PDPageContentStream(document, pdPage, PDPageContentStream.AppendMode.APPEND, true);
 
@@ -181,12 +192,12 @@ public class PdfDraw {
 //                PdfVisualisationUtility.drawLine2DList(pdDocument,
 //                        pageNumber,
 //                        list.get(pageNumber - 1),
-//                        PdfVisualisationUtility.Options.builder().stroke(true).build());
+//                        PdfVisualisationUtility.DrawingOptions.builder().stroke(true).build());
                 PdfVisualisationUtility.drawRectangle2DList(pdDocument,
                         pageNumber,
                         rectanglesPerPage.get(pageNumber - 1),
-                        PdfVisualisationUtility.Options.builder().stroke(true).build());
-                PdfVisualisationUtility.drawRectangle2DList(pdDocument, pageNumber, list.get(pageNumber - 1), PdfVisualisationUtility.Options.builder().stroke(true).build());
+                        DrawingOptions.builder().stroke(true).build());
+                PdfVisualisationUtility.drawRectangle2DList(pdDocument, pageNumber, list.get(pageNumber - 1), DrawingOptions.builder().stroke(true).build());
             }
             pdDocument.save(out);
         }
@@ -202,35 +213,18 @@ public class PdfDraw {
                 PdfVisualisationUtility.drawLine2DList(pdDocument,
                         pageNumber,
                         linesPerPage.get(pageNumber - 1),
-                        PdfVisualisationUtility.Options.builder().strokeColor(Color.RED).stroke(true).build());
+                        DrawingOptions.builder().strokeColor(Color.RED).stroke(true).build());
             }
             pdDocument.save(out);
         }
     }
 
 
-    @Builder
-    @AllArgsConstructor
-    @Getter
-    @FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
-    public static class Options {
-
-        boolean stroke;
-        @Builder.Default
-        Color strokeColor = Color.BLACK;
-        @Builder.Default
-        float strokeWidth = 1f;
-
-        boolean fill;
-        @Builder.Default
-        Color fillColor = Color.BLACK;
-
-    }
 
 
-    private static Options buildStandardOptionsForNodes(DocumentTree.Entry entry) {
+    private static DrawingOptions buildStandardOptionsForNodes(DocumentTree.Entry entry) {
 
-        return Options.builder().stroke(true).strokeColor(switch (entry.getType()) {
+        return DrawingOptions.builder().stroke(true).strokeColor(switch (entry.getType()) {
             case DOCUMENT -> Color.LIGHT_GRAY;
             case HEADER, FOOTER -> Color.GREEN;
             case PARAGRAPH -> Color.BLUE;
@@ -243,7 +237,7 @@ public class PdfDraw {
     }
 
 
-    private static void drawBBoxAndLabelAndNumberOnPage(PDDocument document, DocumentTree.Entry entry, Options options) {
+    private static void drawBBoxAndLabelAndNumberOnPage(PDDocument document, DocumentTree.Entry entry, DrawingOptions options) {
 
         Map<Page, Rectangle2D> rectanglesPerPage = entry.getNode().getBBox();
         for (Page page : rectanglesPerPage.keySet()) {

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);
             }