Compare commits

..
28 Commits
Author SHA1 Message Date
Kilian Schuettler 65ab5eca22 update to redaction-service state 2023-07-25 16:10:57 +02:00
Kilian Schuettler 143ebee25e move and fix layout tests from redaction-service 2023-07-24 19:43:25 +02:00
Kilian Schuettler 47fd8e05d1 rename Data classes 2023-07-24 18:36:27 +02:00
Kilian Schüttler 653f280fd1 Merge branch 'TAAS-41' into 'main'
TAAS-41: TAAS Document Structure

See merge request fforesight/layout-parser!2
2023-07-24 16:09:18 +02:00
Kilian Schuettler daa68f3fa6 TAAS-41: disable experimental tests 2023-07-24 16:07:27 +02:00
Kilian Schuettler ed66043856 TAAS-41: add test files 2023-07-24 16:04:51 +02:00
Kilian Schuettler 526b1c5ad3 TAAS-41: add (inactive) experimental services 2023-07-24 15:58:06 +02:00
Kilian Schuettler 241a32cb4f TAAS-41/ RED-6725: integrate layoutparser into redactmanager 2023-07-24 15:55:31 +02:00
Timo Bejan 9c8501e76a Changes 2023-07-13 18:55:13 +03:00
Timo Bejan 3bc88bc9b7 store new document type 2023-07-13 13:01:01 +03:00
Kilian Schuettler 15a6d46f5c RED-7081: getBBox() Performance Improvement 2023-07-13 13:01:01 +03:00
Kilian Schuettler 788613c92e TAAS-41: TAAS Document Structure
* added linebreaks to ParagraphData
* moved List<String> cellText to List<ParagraphData> cellTexts
2023-07-13 13:01:01 +03:00
Kilian Schuettler 7f0aa32d1b TAAS-41: TAAS Document Structure
* added more testFiles
* hacked a workaround for CMMException
2023-07-13 13:01:01 +03:00
Kilian Schuettler f08c4ced43 TAAS-41: TAAS Document Structure
* changed TextPageBlock splitting
* changed Header and Footer Classification
* added TAAS Document Structure Prototype
2023-07-13 13:01:01 +03:00
Timo Bejan dfdeef5812 updated version 2023-07-12 23:27:49 +03:00
Timo Bejan 69a62c4dbe updated version 2023-07-12 23:22:34 +03:00
Timo Bejan e346c04d67 version fix for gitlab 2023-07-12 20:57:30 +03:00
Timo Bejan cdff0b0ece version fix for gitlab 2023-07-12 20:56:29 +03:00
Timo Bejan a9e6c1f0f8 Image moved to ff 2023-07-12 20:54:36 +03:00
Timo Bejan 4fc7bac818 Project structure cleanup / adapted to fforesight modules 2023-07-12 20:46:01 +03:00
Kilian Schuettler df9cbdc036 RED-6725: Integrate new layout parser
* ported current state from RedactManager
2023-06-15 12:52:51 +02:00
Dominique Eifländer cc1fedac41 Merge branch 'RED-5694' into 'main'
RED-5694: Upgraded to latest platform-dependency

See merge request fforesight/layout-parser!1
2023-06-13 12:33:40 +02:00
deiflaender 54555d4ce0 RED-5694: Upgraded to latest platform-dependency 2023-06-13 12:21:27 +02:00
Christoph Schabert 0ac3ee309a Update .gitlab-ci.yml file 2023-04-19 15:42:43 +02:00
Kilian Schuettler c912d27d92 RED-6009: add direct json output test 2023-04-19 12:22:19 +02:00
Christoph Schabert e60eb0710e Update layoutparser-service-image/pom.xml 2023-04-19 10:44:30 +02:00
Christoph Schabert 4d0a035cc9 Update layoutparser-service-image/pom.xml 2023-04-19 10:19:22 +02:00
Christoph Schabert fb57488ba8 Update layoutparser-service-image/pom.xml 2023-04-18 10:23:00 +02:00
307 changed files with 65020 additions and 4624 deletions
-2
View File
@@ -1,5 +1,3 @@
variables:
SONAR_PROJECT_KEY: 'fforesight_layout-parser_AYd5quv2mRkBOCG22hvF'
include:
- project: 'gitlab/gitlab'
ref: 'main'
+4 -4
View File
@@ -4,23 +4,23 @@
<modelVersion>4.0.0</modelVersion>
<parent>
<groupId>com.iqser.red</groupId>
<groupId>com.knecon.fforesight</groupId>
<artifactId>platform-docker-dependency</artifactId>
<version>1.2.0</version>
<version>0.1.0</version>
<relativePath/>
</parent>
<groupId>com.knecon.fforesight</groupId>
<artifactId>layoutparser-service-image</artifactId>
<version>1.0.0</version>
<version>0.1-SNAPSHOT</version>
<packaging>pom</packaging>
<properties>
<service.server>layoutparser-service-server</service.server>
<platform.jar>${service.server}.jar</platform.jar>
<docker.skip.push>false</docker.skip.push>
<docker.image.prefix>ff</docker.image.prefix>
<docker.image.name>${docker.image.prefix}/${service.server}</docker.image.name>
</properties>
@@ -8,21 +8,16 @@
<parent>
<groupId>com.knecon.fforesight</groupId>
<artifactId>layoutparser-service</artifactId>
<version>1.0.0</version>
<version>0.1-SNAPSHOT</version>
</parent>
<artifactId>layoutparser-service-internal-api</artifactId>
<dependencies>
<dependency>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
<version>1.18.26</version>
</dependency>
<dependency>
<groupId>com.google.guava</groupId>
<artifactId>guava</artifactId>
<version>31.1-jre</version>
<version>${guava.version}</version>
</dependency>
</dependencies>
@@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@@ -12,9 +12,10 @@ import lombok.experimental.FieldDefaults;
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class DocumentData {
PageData[] pages;
AtomicTextBlockData[] atomicTextBlocks;
AtomicPositionBlockData[] atomicPositionBlocks;
TableOfContentsData tableOfContents;
DocumentPage[] documentPages;
DocumentTextData[] documentTextData;
DocumentPositionData[] documentPositions;
DocumentStructure documentStructure;
}
@@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@@ -10,7 +10,7 @@ import lombok.experimental.FieldDefaults;
@Builder
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class PageData {
public class DocumentPage {
int number;
int height;
@@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@@ -10,7 +10,7 @@ import lombok.experimental.FieldDefaults;
@Builder
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class AtomicPositionBlockData {
public class DocumentPositionData {
Long id;
int[] stringIdxToPositionIdx;
@@ -1,11 +1,9 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.util.List;
import java.util.Map;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.NodeType;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
@@ -19,7 +17,7 @@ import lombok.experimental.FieldDefaults;
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class TableOfContentsData {
public class DocumentStructure {
EntryData root;
@@ -29,9 +27,9 @@ public class TableOfContentsData {
if (tocId.isEmpty()) {
return root;
}
EntryData entry = root.subEntries.get(tocId.get(0));
EntryData entry = root.children.get(tocId.get(0));
for (int id : tocId.subList(1, tocId.size())) {
entry = entry.subEntries.get(id);
entry = entry.children.get(id);
}
return entry;
}
@@ -39,7 +37,7 @@ public class TableOfContentsData {
public Stream<EntryData> streamAllEntries() {
return Stream.concat(Stream.of(root), root.subEntries.stream()).flatMap(TableOfContentsData::flatten);
return Stream.concat(Stream.of(root), root.children.stream()).flatMap(DocumentStructure::flatten);
}
@@ -51,7 +49,7 @@ public class TableOfContentsData {
private static Stream<EntryData> flatten(EntryData entry) {
return Stream.concat(Stream.of(entry), entry.subEntries.stream().flatMap(TableOfContentsData::flatten));
return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(DocumentStructure::flatten));
}
@@ -62,19 +60,18 @@ public class TableOfContentsData {
public static class EntryData {
NodeType type;
int[] tocId;
Long[] atomicBlocks;
Long[] pages;
int[] treeId;
Long[] atomicBlockIds;
Long[] pageNumbers;
Map<String, String> properties;
List<EntryData> subEntries;
List<EntryData> children;
@Override
public String toString() {
StringBuilder sb = new StringBuilder();
sb.append("[");
for (int i : tocId) {
for (int i : treeId) {
sb.append(i);
sb.append(",");
}
@@ -83,7 +80,7 @@ public class TableOfContentsData {
sb.append(type);
sb.append(" atbs = ");
sb.append(atomicBlocks.length);
sb.append(atomicBlockIds.length);
return sb.toString();
}
@@ -1,4 +1,6 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
@@ -10,7 +12,7 @@ import lombok.experimental.FieldDefaults;
@Builder
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class AtomicTextBlockData {
public class DocumentTextData {
Long id;
Long page;
@@ -20,4 +22,6 @@ public class AtomicTextBlockData {
int end;
int[] lineBreaks;
}
@@ -1,4 +1,6 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
package com.knecon.fforesight.service.layoutparser.internal.api.data.redaction;
import java.util.Locale;
public enum NodeType {
DOCUMENT,
@@ -9,5 +11,11 @@ public enum NodeType {
TABLE_CELL,
IMAGE,
HEADER,
FOOTER
FOOTER;
public String toString() {
return this.name().charAt(0) + this.name().substring(1).toLowerCase(Locale.ROOT);
}
}
@@ -0,0 +1,21 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.Builder;
import lombok.Data;
@Data
@Builder
public class ParagraphData {
private String text;
List<Range> boldTextBoundaries;
List<Range> italicTextBoundaries;
List<Integer> linebreaks;
private String classification;
private String orientation;
private int textDirection;
}
@@ -0,0 +1,5 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
public record Range(int start, int end) {
}
@@ -0,0 +1,16 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
@Builder
@Data
@AllArgsConstructor
public class ResearchDocumentData {
String originalFile;
List<StructureObject> structureObjects;
}
@@ -0,0 +1,15 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.AllArgsConstructor;
import lombok.Data;
@Data
@AllArgsConstructor
public class RowData {
boolean header;
List<ParagraphData> cellText;
float[] bBox;
}
@@ -0,0 +1,19 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
@Data
@Builder
@AllArgsConstructor
public class StructureObject {
Integer structureObjectNumber;
int page;
int stringOffset;
float[] boundingBox;
ParagraphData paragraph;
TableData table;
}
@@ -0,0 +1,15 @@
package com.knecon.fforesight.service.layoutparser.internal.api.data.taas;
import java.util.List;
import lombok.AllArgsConstructor;
import lombok.Data;
@Data
@AllArgsConstructor
public class TableData {
List<RowData> rowData;
Integer numberOfCols;
Integer numberOfRows;
}
@@ -1,101 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SectionNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class DocumentGraph implements SemanticNode {
Set<PageNode> pages;
TableOfContents tableOfContents;
Integer numberOfPages;
TextBlock textBlock;
public TextBlock buildTextBlock() {
return streamTerminalTextBlocksInOrder().collect(new TextBlockCollector());
}
public List<SectionNode> getMainSections() {
return streamChildren().filter(node -> node instanceof SectionNode).map(node -> (SectionNode) node).collect(Collectors.toList());
}
public Stream<TextBlock> streamTerminalTextBlocksInOrder() {
return streamAllNodes().filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock);
}
public Set<EntityNode> getEntities() {
return streamAllSubNodes().map(SemanticNode::getEntities).flatMap(Set::stream).collect(Collectors.toUnmodifiableSet());
}
@Override
public List<Integer> getTocId() {
return Collections.emptyList();
}
@Override
public void setTocId(List<Integer> tocId) {
throw new UnsupportedOperationException("DocumentGraph is always the root of the Table of Contents");
}
private Stream<SemanticNode> streamAllNodes() {
return tableOfContents.streamAllEntriesInOrder().map(TableOfContents.Entry::getNode);
}
@Override
public String toString() {
return NodeType.DOCUMENT + ": " + buildTextBlock().buildSummary();
}
@Override
public Map<PageNode, Rectangle2D> getBBox() {
Map<PageNode, Rectangle2D> bBox = new HashMap<>();
for (PageNode page : pages) {
bBox.put(page, new Rectangle2D.Double(0, 0, page.getWidth(), page.getHeight()));
}
return bBox;
}
}
@@ -1,193 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph;
import static java.lang.String.format;
import java.nio.charset.StandardCharsets;
import java.util.Collections;
import java.util.LinkedList;
import java.util.List;
import java.util.stream.Stream;
import com.google.common.hash.Hashing;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@Data
public class TableOfContents {
private final Entry root;
public TableOfContents(DocumentGraph documentGraph) {
root = Entry.builder().tocId(Collections.emptyList()).type(NodeType.DOCUMENT).children(new LinkedList<>()).node(documentGraph).build();
}
public TextBlock buildTextBlock() {
return streamAllEntriesInOrder().map(Entry::getNode).filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock).collect(new TextBlockCollector());
}
public List<Integer> createNewMainEntryAndReturnId(NodeType nodeType, SemanticNode node) {
return createNewChildEntryAndReturnId(Collections.emptyList(), nodeType, node);
}
public List<Integer> createNewChildEntryAndReturnId(List<Integer> parentId, NodeType nodeType, SemanticNode node) {
if (!entryExists(parentId)) {
throw new UnsupportedOperationException(format("parentId %s does not exist!", parentId));
}
Entry parent = getEntryById(parentId);
List<Integer> newId = new LinkedList<>(parentId);
newId.add(parent.children.size());
parent.children.add(Entry.builder().tocId(newId).node(node).type(nodeType).children(new LinkedList<>()).build());
return newId;
}
private boolean entryExists(List<Integer> tocId) {
if (tocId.isEmpty()) {
return root != null;
}
Entry entry = root.children.get(tocId.get(0));
for (int id : tocId.subList(1, tocId.size())) {
if (id >= entry.children.size() || 0 > id) {
return false;
}
entry = entry.children.get(id);
}
return true;
}
public Entry getParentEntryById(List<Integer> tocId) {
return getEntryById(getParentId(tocId));
}
public boolean hasParentById(List<Integer> tocId) {
return entryExists(getParentId(tocId));
}
public Stream<SemanticNode> streamChildrenNodes(List<Integer> tocId) {
return getEntryById(tocId).children.stream().map(Entry::getNode);
}
private static List<Integer> getParentId(List<Integer> tocId) {
if (tocId.isEmpty()) {
throw new UnsupportedOperationException("Root has no parent!");
}
if (tocId.size() < 2) {
return Collections.emptyList();
}
return tocId.subList(0, tocId.size() - 1);
}
public Entry getEntryById(List<Integer> tocId) {
if (tocId.isEmpty()) {
return root;
}
Entry entry = root.children.get(tocId.get(0));
for (int id : tocId.subList(1, tocId.size())) {
entry = entry.children.get(id);
}
return entry;
}
public Stream<Entry> streamMainEntries() {
return root.children.stream();
}
public Stream<Entry> streamAllEntriesInOrder() {
return Stream.of(root).flatMap(TableOfContents::flatten);
}
public Stream<Entry> streamAllSubEntriesInOrder(List<Integer> parentId) {
return getEntryById(parentId).getChildren().stream().flatMap(TableOfContents::flatten);
}
@Override
public String toString() {
return String.join("\n", streamAllEntriesInOrder().map(Entry::toString).toList());
}
public String toString(List<Integer> id) {
return String.join("\n", streamAllSubEntriesInOrder(id).map(Entry::toString).toList());
}
private static Stream<Entry> flatten(Entry entry) {
return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(TableOfContents::flatten));
}
@Builder
@Getter
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE, makeFinal = true)
public static class Entry {
List<Integer> tocId;
NodeType type;
SemanticNode node;
List<Entry> children;
@Override
public String toString() {
return node.toString();
}
@Override
public int hashCode() {
return Hashing.murmur3_32_fixed().hashString(toString(), StandardCharsets.UTF_8).hashCode();
}
@Override
public boolean equals(Object o) {
return o instanceof Entry && o.hashCode() == this.hashCode();
}
}
}
@@ -1,76 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.entity;
import java.util.Collections;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
public interface EntityNode {
/**
* This represents the text, which is contained within the boundary of the Entity.
*
* @return String
*/
String getValue();
/**
* The Boundary primarily defines the Entity, all other values may be inferred from it.
*
* @return Boundary, uniquely identifying this Entity
*/
Boundary getBoundary();
/**
* The deepest fully containing node represents the node which is the deepest node in the document tree structure,
* whose boundary also fully contains the boundary of this entity.
*
* @return the deepest fully containing node
*/
SemanticNode getDeepestFullyContainingNode();
/**
* The intersecting nodes represent all nodes, whose boundary intersects the boundary of this entity.
*
* @return all intersecting Nodes
*/
List<SemanticNode> getIntersectingNodes();
void setDeepestFullyContainingNode(SemanticNode semanticNode);
void addIntersectingNode(SemanticNode semanticNode);
void setIntersectingNodes(List<SemanticNode> semanticNodes);
/**
* @return all pages this entity intersects.
*/
Set<PageNode> getPages();
void setPages(Set<PageNode> pages);
/**
* removes all occurrences of this node in the graph and resets all graph specific fields.
*/
default void removeFromGraph() {
getIntersectingNodes().forEach(node -> node.getEntities().remove(this));
getPages().forEach(page -> page.getEntities().remove(this));
setPages(Collections.emptySet());
setDeepestFullyContainingNode(null);
setIntersectingNodes(Collections.emptyList());
}
}
@@ -1,45 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.entity;
import java.awt.geom.Rectangle2D;
import java.nio.charset.StandardCharsets;
import java.util.List;
import com.google.common.hash.Hashing;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import lombok.AccessLevel;
import lombok.Builder;
import lombok.Data;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@FieldDefaults(level = AccessLevel.PRIVATE)
public class EntityPosition {
PageNode pageNode;
List<Rectangle2D> rectanglePerLine;
public String getId() {
return String.valueOf(hashCode());
}
@Override
public int hashCode() {
StringBuilder sb = new StringBuilder();
sb.append(pageNode.getNumber());
rectanglePerLine.forEach(r -> sb.append(r.getX()).append(r.getY()).append(r.getWidth()).append(r.getHeight()));
return Hashing.murmur3_128().hashString(sb.toString(), StandardCharsets.UTF_8).hashCode();
}
@Override
public boolean equals(Object o) {
return o instanceof EntityPosition && o.hashCode() == this.hashCode();
}
}
@@ -1,53 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class FooterNode implements SemanticNode {
List<Integer> tocId;
TextBlock terminalTextBlock;
@Builder.Default
boolean terminal = true;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public TextBlock buildTextBlock() {
return terminalTextBlock;
}
@Override
public String toString() {
return tocId + ": " + NodeType.FOOTER + ": " + terminalTextBlock.buildSummary();
}
}
@@ -1,53 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class HeaderNode implements SemanticNode {
List<Integer> tocId;
TextBlock terminalTextBlock;
@Builder.Default
boolean terminal = true;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public TextBlock buildTextBlock() {
return terminalTextBlock;
}
@Override
public String toString() {
return tocId + ": " + NodeType.HEADER + ": " + terminalTextBlock.buildSummary();
}
}
@@ -1,60 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class HeadlineNode implements SemanticNode {
List<Integer> tocId;
TextBlock terminalTextBlock;
@Builder.Default
boolean terminal = true;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public TextBlock buildTextBlock() {
return terminalTextBlock;
}
@Override
public String toString() {
return tocId + ": " + NodeType.HEADLINE + ": " + terminalTextBlock.buildSummary();
}
@Override
public SemanticNode getHeadline() {
return this;
}
}
@@ -1,87 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
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.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class ImageNode implements SemanticNode {
List<Integer> tocId;
ImageType imageType;
boolean transparency;
Rectangle2D position;
boolean redaction;
boolean ignored;
@Builder.Default
String redactionReason = "";
@Builder.Default
String legalBasis = "";
@Builder.Default
int matchedRule = -1;
@EqualsAndHashCode.Exclude
PageNode page;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public TextBlock buildTextBlock() {
return streamAllSubNodes().filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock).collect(new TextBlockCollector());
}
@Override
public Set<PageNode> getPages() {
return Collections.singleton(page);
}
@Override
public String toString() {
return tocId + ": " + NodeType.IMAGE + ": " + imageType.toString() + " " + position;
}
@Override
public Map<PageNode, Rectangle2D> getBBox() {
Map<PageNode, Rectangle2D> bBoxPerPage = new HashMap<>();
bBoxPerPage.put(page, position);
return bBoxPerPage;
}
}
@@ -1,9 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
public enum ImageType {
LOGO,
FORMULA,
SIGNATURE,
OTHER,
OCR
}
@@ -1,71 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.EqualsAndHashCode;
import lombok.Getter;
import lombok.Setter;
import lombok.experimental.FieldDefaults;
@Getter
@Setter
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class PageNode {
Integer number;
Integer height;
Integer width;
Integer rotation;
@EqualsAndHashCode.Exclude
List<SemanticNode> mainBody;
@EqualsAndHashCode.Exclude
HeaderNode header;
@EqualsAndHashCode.Exclude
FooterNode footer;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Builder.Default
@EqualsAndHashCode.Exclude
Set<ImageNode> images = new HashSet<>();
public TextBlock getMainBodyTextBlock() {
return mainBody.stream().filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock).collect(new TextBlockCollector());
}
@Override
public String toString() {
return String.valueOf(number);
}
@Override
public int hashCode() {
return number;
}
@Override
public boolean equals(Object o) {
return o instanceof PageNode && o.hashCode() == this.hashCode();
}
}
@@ -1,51 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class ParagraphNode implements SemanticNode {
List<Integer> tocId;
TextBlock terminalTextBlock;
@Builder.Default
boolean terminal = true;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public TextBlock buildTextBlock() {
return terminalTextBlock;
}
@Override
public String toString() {
return tocId + ": " + NodeType.PARAGRAPH + ": " + terminalTextBlock.buildSummary();
}
}
@@ -1,63 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.NoSuchElementException;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class SectionNode implements SemanticNode {
List<Integer> tocId;
TextBlock textBlock;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public TextBlock buildTextBlock() {
if (textBlock == null) {
textBlock = streamAllSubNodes().filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock).collect(new TextBlockCollector());
}
return textBlock;
}
@Override
public String toString() {
return tocId.toString() + ": " + NodeType.SECTION + ": " + buildTextBlock().buildSummary();
}
public HeadlineNode getHeadline() {
return streamChildren().filter(node -> node instanceof HeadlineNode)
.map(node -> (HeadlineNode) node)
.findFirst()
.orElseThrow(() -> new NoSuchElementException("ClassificationSection has no Headline!"));
}
}
@@ -1,275 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.awt.geom.Rectangle2D;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.services.RectangleTransformations;
public interface SemanticNode {
/**
* Searches all Nodes located underneath this Node in the TableOfContents and concatenates their AtomicTextBlocks into a single TextBlockEntity.
* So, for a ClassificationSection all TextBlocks of Subsections, Paragraphs, and Tables are concatenated into a single TextBlockEntity
* If the Node is Terminal, the TerminalTextBlock will be returned instead.
*
* @return ClassificationTextBlock containing all AtomicTextBlocks that are located under this Node.
*/
TextBlock buildTextBlock();
/**
* Any Node maintains its own Set of Entities.
* This Set contains all Entities whose boundary intersects the boundary of this node.
*
* @return Set of all Entities associated with this Node
*/
Set<EntityNode> getEntities();
/**
* Each AtomicTextBlock is assigned a page, so to get the pages this node appears on, it collects the PageNodes from each AtomicTextBlock belonging to this node's ClassificationTextBlock.
*
* @return Set of PageNodes this node appears on.
*/
default Set<PageNode> getPages() {
return buildTextBlock().getPages();
}
/**
* @return the TableOfContents of the ClassificationDocument this node belongs to
*/
TableOfContents getTableOfContents();
/**
* The id is a List of Integers uniquely identifying this node in the TableOfContents.
*
* @return the TableOfContents ID
*/
List<Integer> getTocId();
/**
* This should only be used during graph construction.
*
* @param tocId List of Integers
*/
void setTocId(List<Integer> tocId);
/**
* Traverses the Tree up, until it hits a HeadlineNode or hits a SectionNode which will then return the first HeadlineNode from its children.
* Throws NotFoundException if no Headline is found this way
*
* @return First HeadlineNode found
*/
default SemanticNode getHeadline() {
return getParent().getHeadline();
}
/**
* @return boolean indicating wether this Node has a Parent in the TableOfContents
*/
default boolean hasParent() {
return getTableOfContents().hasParentById(getTocId());
}
/**
* @return The SemanticNode representing the Parent in the TableOfContents
* throws NotFoundException, when no parent is present
*/
default SemanticNode getParent() {
return getTableOfContents().getParentEntryById(getTocId()).getNode();
}
/**
* Terminal means a SemanticNode has direct access to a ClassificationTextBlock, by default this is false and must be overridden.
* Currently only Sections, Images, and Tables are not terminal.
* A TableCell might be Terminal depending on its area compared to the page.
*
* @return boolean, indicating if a Node has direct access to a ClassificationTextBlock
*/
default boolean isTerminal() {
return false;
}
/**
* Terminal means a SemanticNode has direct access to a ClassificationTextBlock, by default this is false and must be overridden.
* Currently only Sections and Tables are not terminal.
*
* @return AtomicTextBlock
*/
default TextBlock getTerminalTextBlock() {
throw new UnsupportedOperationException("Only terminal Nodes have access to TerminalTextBlocks!");
}
default void setTerminalTextBlock(TextBlock textBlock) {
throw new UnsupportedOperationException();
}
/**
* Each AtomicTextBlock has an index on its page, this returns the number of the first AtomicTextBlock underneath this node.
* If this node does not have any AtomicTexBlocks underneath it, e.g. an empty TableCell. It returns -1.
*
* @return Integer representing the number on the page
*/
default Integer getNumberOnPage() {
TextBlock textBlock = buildTextBlock();
if (textBlock.getAtomicTextBlocks().size() > 0) {
return buildTextBlock().getAtomicTextBlocks().get(0).getNumberOnPage();
} else {
return -1;
}
}
/**
* @return true, if this node's ClassificationTextBlock is not empty
*/
default boolean hasText() {
return buildTextBlock().length() > 0;
}
/**
* @param string A String which the ClassificationTextBlock might contain
* @return true, if this node's ClassificationTextBlock contains the string
*/
default boolean containsString(String string) {
return buildTextBlock().getSearchText().contains(string);
}
/**
* @param strings A List of Strings which the ClassificationTextBlock might contain
* @return true, if this node's ClassificationTextBlock contains any of the strings
*/
default boolean containsAnyString(List<String> strings) {
return strings.stream().anyMatch(this::containsString);
}
/**
* This function is used during insertion of EntityNodes into the graph, it checks if the boundary of the EntityNode intersects or even contains the EntityNode.
* It sets the fields accordingly and recursively calls this function on all its children.
*
* @param entityNode EntityNode, which is being inserted into the graph
*/
default void addThisToEntityIfIntersects(EntityNode entityNode) {
TextBlock textBlock = buildTextBlock();
if (textBlock.getBoundary().intersects(entityNode.getBoundary())) {
if (textBlock.containsBoundary(entityNode.getBoundary())) {
entityNode.setDeepestFullyContainingNode(this);
}
entityNode.addIntersectingNode(this);
streamChildren().forEach(node -> node.addThisToEntityIfIntersects(entityNode));
}
}
/**
* Streams all children located directly underneath this node in the TableOfContents.
*
* @return Stream of all children
*/
default Stream<SemanticNode> streamChildren() {
return getTableOfContents().streamChildrenNodes(getTocId());
}
/**
* recursively streams all SemanticNodes located underneath this node in the TableOfContents in order.
*
* @return Stream of all SubNodes
*/
default Stream<SemanticNode> streamAllSubNodes() {
return getTableOfContents().streamAllSubEntriesInOrder(getTocId()).map(TableOfContents.Entry::getNode);
}
/**
* @return Boundary of this Node's ClassificationTextBlock
*/
default Boundary getBoundary() {
return buildTextBlock().getBoundary();
}
/**
* If this Node is Terminal it will calculate the boundingBox of its TerminalTextBlock, otherwise it will calculate the Union of the BoundingBoxes of all its Children.
* If called on the ClassificationDocument, it will return the cropbox of each page
*
* @return Rectangle2D fully encapsulating this Node for each page.
*/
default Map<PageNode, Rectangle2D> getBBox() {
Map<PageNode, Rectangle2D> bBoxPerPage = new HashMap<>();
if (isTerminal()) {
return getBBoxFromTerminalTextBlock(bBoxPerPage);
}
return getBBoxFromChildren(bBoxPerPage);
}
/**
* TODO this does not yet work for sections spanning multiple columns.
*
* @param bBoxPerPage initial empty BoundingBox
* @return The union of the BoundingBoxes of all children
*/
private Map<PageNode, Rectangle2D> getBBoxFromChildren(Map<PageNode, Rectangle2D> bBoxPerPage) {
return streamChildren().map(SemanticNode::getBBox).reduce((map1, map2) -> {
map1.forEach((page, rectangle) -> map2.merge(page, rectangle, (rect1, rect2) -> rect1.createUnion(rect2).getBounds2D()));
return map2;
}).orElse(bBoxPerPage);
}
/**
* @param bBoxPerPage initial empty BoundingBox
* @return The union of all BoundingBoxes of the ClassificationTextBlock of this node
*/
private Map<PageNode, Rectangle2D> getBBoxFromTerminalTextBlock(Map<PageNode, Rectangle2D> bBoxPerPage) {
Map<PageNode, List<AtomicTextBlock>> atomicTextBlockPerPage = buildTextBlock().getAtomicTextBlocks().stream().collect(Collectors.groupingBy(AtomicTextBlock::getPage));
atomicTextBlockPerPage.forEach((page, atbs) -> bBoxPerPage.put(page, RectangleTransformations.bBoxUnionAtomicTextBlock(atbs)));
return bBoxPerPage;
}
}
@@ -1,92 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.awt.geom.Rectangle2D;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class TableCellNode implements SemanticNode {
List<Integer> tocId;
int row;
int col;
boolean header;
Rectangle2D bBox;
@Builder.Default
boolean terminal = true;
TextBlock terminalTextBlock;
TextBlock textBlock;
@EqualsAndHashCode.Exclude
TableOfContents tableOfContents;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
@Override
public Map<PageNode, Rectangle2D> getBBox() {
Map<PageNode, Rectangle2D> bBoxPerPage = new HashMap<>();
getPages().forEach(page -> bBoxPerPage.put(page, bBox));
return bBoxPerPage;
}
@Override
public TextBlock buildTextBlock() {
if (terminal) {
return terminalTextBlock;
}
if (textBlock == null) {
textBlock = streamAllSubNodes().filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock).collect(new TextBlockCollector());
}
return textBlock;
}
@Override
public String toString() {
return tocId + ": " + NodeType.TABLE_CELL + ": " + buildTextBlock().buildSummary();
}
public boolean hasHeader(String headerString) {
return getHeaders().anyMatch(header -> header.buildTextBlock().getSearchText().strip().equals(headerString));
}
private Stream<TableCellNode> getHeaders() {
TableNode tableNode = (TableNode) getParent();
return tableNode.streamHeadersForCell(row, col);
}
}
@@ -1,73 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class TableNode implements SemanticNode {
List<Integer> tocId;
TableOfContents tableOfContents;
Integer numberOfRows;
Integer numberOfCols;
TextBlock textBlock;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<EntityNode> entities = new HashSet<>();
public Stream<TableCellNode> streamTableCells() {
return streamChildren().map(node -> (TableCellNode) node);
}
public Stream<TableCellNode> streamHeaders() {
return streamTableCells().filter(TableCellNode::isHeader);
}
public Stream<TableCellNode> streamHeadersForCell(int row, int col) {
return streamHeaders().filter(cell -> cell.getRow() == row || cell.getCol() == col);
}
@Override
public TextBlock buildTextBlock() {
if (textBlock == null) {
textBlock = streamAllSubNodes().filter(SemanticNode::isTerminal).map(SemanticNode::getTerminalTextBlock).collect(new TextBlockCollector());
}
return textBlock;
}
@Override
public String toString() {
return tocId.toString() + ": " + NodeType.TABLE + ": " + buildTextBlock().buildSummary();
}
}
@@ -1,131 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock;
import java.awt.geom.Rectangle2D;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityPosition;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.services.RectangleTransformations;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class AtomicTextBlock implements TextBlock {
Long id;
Integer numberOnPage;
PageNode page;
//string coordinates
Boundary boundary;
String searchText;
List<Integer> lineBreaks;
//position coordinates
List<Integer> stringIdxToPositionIdx;
List<Rectangle2D> positions;
@EqualsAndHashCode.Exclude
SemanticNode parent;
@Override
public int numberOfLines() {
return lineBreaks.size() + 1;
}
public CharSequence getLine(int lineNumber) {
if (lineNumber >= numberOfLines() || lineNumber < 0) {
throw new IndexOutOfBoundsException(String.format("line %d out of range for AtomicTextBlock with %d lines", lineNumber, numberOfLines()));
}
if (lineNumber == 0) {
return subSequence(boundary.start(), lineBreaks.get(0) + boundary.start());
} else if (lineNumber == numberOfLines() - 1) {
return subSequence(lineBreaks.get(lineBreaks.size() - 1) + boundary.start(), boundary.end());
}
return subSequence(lineBreaks.get(lineNumber - 1) + boundary.start(), lineBreaks.get(lineNumber) + boundary.start());
}
@Override
public List<AtomicTextBlock> getAtomicTextBlocks() {
return List.of(this);
}
@Override
public int getNextLinebreak(int fromIndex) {
return lineBreaks.stream()//
.filter(linebreak -> linebreak > fromIndex - boundary.start()) //
.findFirst() //
.orElse(searchText.length()) + boundary.start();
}
@Override
public int getPreviousLinebreak(int fromIndex) {
return lineBreaks.stream()//
.filter(linebreak -> linebreak <= fromIndex - boundary.start())//
.reduce((a, b) -> b)//
.orElse(0) + boundary.start();
}
@Override
public Rectangle2D getPosition(int stringIdx) {
return positions.get(stringIdxToPositionIdx.get(stringIdx - boundary.start()));
}
@Override
public List<Rectangle2D> getPositions(Boundary stringBoundary) {
if (!containsBoundary(stringBoundary)) {
throw new IndexOutOfBoundsException(String.format("%s is out of bounds for %s", stringBoundary, this.boundary));
}
if (stringBoundary.end() == this.boundary.end()) {
return positions.subList(stringIdxToPositionIdx.get(stringBoundary.start() - this.boundary.start()), positions.size());
}
return positions.subList(stringIdxToPositionIdx.get(stringBoundary.start() - this.boundary.start()),
stringIdxToPositionIdx.get(stringBoundary.end() - this.boundary.start()));
}
public List<EntityPosition> getEntityPositionsPerPage(Boundary stringBoundary) {
List<Rectangle2D> positionsPerLine = stringBoundary.split(getLineBreaks().stream().map(lb -> lb + boundary.start()).filter(stringBoundary::contains).toList())
.stream()
.map(this::getPositions)
.map(RectangleTransformations::rectangleUnion)
.toList();
return List.of(EntityPosition.builder().rectanglePerLine(positionsPerLine).pageNode(page).build());
}
@Override
public String toString() {
return searchText;
}
}
@@ -1,146 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.mapper;
import java.awt.geom.Rectangle2D;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.TableOfContentsData;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.DocumentGraph;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ImageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableCellNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import lombok.experimental.UtilityClass;
@UtilityClass
public class DocumentDataMapper {
public DocumentData toDocumentData(DocumentGraph documentGraph) {
List<AtomicTextBlockData> atomicTextBlockData = documentGraph.streamTerminalTextBlocksInOrder()
.flatMap(textBlock -> textBlock.getAtomicTextBlocks().stream())
.distinct()
.map(DocumentDataMapper::toAtomicTextBlockData)
.toList();
List<AtomicPositionBlockData> atomicPositionBlockData = documentGraph.streamTerminalTextBlocksInOrder()
.flatMap(textBlock -> textBlock.getAtomicTextBlocks().stream())
.distinct()
.map(DocumentDataMapper::toAtomicPositionBlockData)
.toList();
List<PageData> pageData = documentGraph.getPages().stream().map(DocumentDataMapper::toPageData).toList();
TableOfContentsData tableOfContentsData = toTableOfContentsData(documentGraph.getTableOfContents());
return DocumentData.builder()
.atomicTextBlocks(atomicTextBlockData.toArray(new AtomicTextBlockData[0]))
.atomicPositionBlocks(atomicPositionBlockData.toArray(new AtomicPositionBlockData[0]))
.pages(pageData.toArray(new PageData[0]))
.tableOfContents(tableOfContentsData)
.build();
}
private TableOfContentsData toTableOfContentsData(TableOfContents tableOfContents) {
return new TableOfContentsData(toEntryData(tableOfContents.getRoot()));
}
private TableOfContentsData.EntryData toEntryData(TableOfContents.Entry entry) {
Long[] atomicTextBlocks;
if (entry.getNode().isTerminal()) {
atomicTextBlocks = toAtomicTextBlockIds(entry.getNode().getTerminalTextBlock());
} else {
atomicTextBlocks = new Long[]{};
}
Map<String, String> properties = switch (entry.getType()) {
case TABLE -> PropertiesMapper.buildTableProperties((TableNode) entry.getNode());
case TABLE_CELL -> PropertiesMapper.buildTableCellProperties((TableCellNode) entry.getNode());
case IMAGE -> PropertiesMapper.buildImageProperties((ImageNode) entry.getNode());
default -> new HashMap<>();
};
return TableOfContentsData.EntryData.builder()
.tocId(toPrimitiveIntArray(entry.getTocId()))
.subEntries(entry.getChildren().stream().map(DocumentDataMapper::toEntryData).toList())
.type(entry.getType())
.atomicBlocks(atomicTextBlocks)
.pages(entry.getNode().getPages().stream().map(PageNode::getNumber).map(Integer::longValue).toArray(Long[]::new))
.properties(properties)
.build();
}
private Long[] toAtomicTextBlockIds(TextBlock textBlock) {
return textBlock.getAtomicTextBlocks().stream().map(AtomicTextBlock::getId).toArray(Long[]::new);
}
private PageData toPageData(PageNode p) {
return PageData.builder().rotation(p.getRotation()).height(p.getHeight()).width(p.getWidth()).number(p.getNumber()).build();
}
private AtomicTextBlockData toAtomicTextBlockData(AtomicTextBlock atomicTextBlock) {
return AtomicTextBlockData.builder()
.id(atomicTextBlock.getId())
.page(atomicTextBlock.getPage().getNumber().longValue())
.searchText(atomicTextBlock.getSearchText())
.numberOnPage(atomicTextBlock.getNumberOnPage())
.start(atomicTextBlock.getBoundary().start())
.end(atomicTextBlock.getBoundary().end())
.lineBreaks(toPrimitiveIntArray(atomicTextBlock.getLineBreaks()))
.build();
}
private AtomicPositionBlockData toAtomicPositionBlockData(AtomicTextBlock atomicTextBlock) {
return AtomicPositionBlockData.builder()
.id(atomicTextBlock.getId())
.positions(toPrimitiveFloatMatrix(atomicTextBlock.getPositions()))
.stringIdxToPositionIdx(toPrimitiveIntArray(atomicTextBlock.getStringIdxToPositionIdx()))
.build();
}
private float[][] toPrimitiveFloatMatrix(List<Rectangle2D> positions) {
float[][] positionMatrix = new float[positions.size()][];
for (int i = 0; i < positions.size(); i++) {
float[] singlePositions = new float[4];
singlePositions[0] = (float) positions.get(i).getMinX();
singlePositions[1] = (float) positions.get(i).getMinY();
singlePositions[2] = (float) positions.get(i).getWidth();
singlePositions[3] = (float) positions.get(i).getHeight();
positionMatrix[i] = singlePositions;
}
return positionMatrix;
}
private int[] toPrimitiveIntArray(List<Integer> list) {
int[] array = new int[list.size()];
for (int i = 0; i < list.size(); i++) {
array[i] = list.get(i);
}
return array;
}
}
@@ -1,229 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.mapper;
import static com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.NodeType.FOOTER;
import static com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.NodeType.HEADER;
import java.awt.geom.Rectangle2D;
import java.util.Arrays;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import java.util.NoSuchElementException;
import com.google.common.primitives.Ints;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.TableOfContentsData;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.DocumentGraph;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.FooterNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.HeaderNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.HeadlineNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ImageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ParagraphNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SectionNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableCellNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlockCollector;
import lombok.experimental.UtilityClass;
@UtilityClass
public class DocumentGraphMapper {
public DocumentGraph toDocumentGraph(DocumentData documentData) {
DocumentGraph documentGraph = new DocumentGraph();
Context context = new Context(documentData,
new TableOfContents(documentGraph),
new LinkedList<>(),
new LinkedList<>(),
Arrays.stream(documentData.getAtomicTextBlocks()).toList(),
Arrays.stream(documentData.getAtomicPositionBlocks()).toList());
context.pages.addAll(Arrays.stream(documentData.getPages()).map(DocumentGraphMapper::buildPage).toList());
context.tableOfContents.getRoot().getChildren().addAll(buildEntries(documentData.getTableOfContents().getRoot().getSubEntries(), context));
documentGraph.setTableOfContents(context.tableOfContents);
documentGraph.setPages(new HashSet<>(context.pages));
documentGraph.setNumberOfPages(documentData.getPages().length);
documentGraph.setTextBlock(documentGraph.buildTextBlock());
return documentGraph;
}
private List<TableOfContents.Entry> buildEntries(List<TableOfContentsData.EntryData> entries,
Context context) {
List<TableOfContents.Entry> newEntries = new LinkedList<>();
for (TableOfContentsData.EntryData entryData : entries) {
boolean terminal = isTerminal(entryData);
List<PageNode> pages = Arrays.stream(entryData.getPages()).map(pageNumber -> getPage(pageNumber, context)).toList();
SemanticNode node = switch (entryData.getType()) {
case SECTION -> buildSection(context);
case PARAGRAPH -> buildParagraph(context, terminal);
case HEADLINE -> buildHeadline(context, terminal);
case HEADER -> buildHeader(context, terminal);
case FOOTER -> buildFooter(context, terminal);
case TABLE -> buildTable(context, entryData.getProperties());
case TABLE_CELL -> buildTableCell(context, entryData.getProperties(), terminal);
case IMAGE -> buildImage(context, entryData.getProperties());
default -> throw new UnsupportedOperationException("Not yet implemented for type " + entryData.getType());
};
if (node.isTerminal()) {
TextBlock textBlock = toTextBlock(entryData.getAtomicBlocks(), context, node);
node.setTerminalTextBlock(textBlock);
}
List<Integer> tocId = Arrays.stream(entryData.getTocId()).boxed().toList();
node.setTocId(tocId);
if (entryData.getType() == HEADER) {
pages.forEach(page -> page.setHeader((HeaderNode) node));
} else if (entryData.getType() == FOOTER) {
pages.forEach(page -> page.setFooter((FooterNode) node));
} else {
pages.forEach(page -> page.getMainBody().add(node));
}
newEntries.add(TableOfContents.Entry.builder().tocId(tocId).type(entryData.getType()).children(buildEntries(entryData.getSubEntries(), context)).node(node).build());
}
return newEntries;
}
private HeadlineNode buildHeadline(Context context, boolean terminal) {
return HeadlineNode.builder().terminal(terminal).tableOfContents(context.tableOfContents()).build();
}
private static boolean isTerminal(TableOfContentsData.EntryData entryData) {
return entryData.getAtomicBlocks().length > 0;
}
private ImageNode buildImage(Context context, Map<String, String> properties) {
var builder = ImageNode.builder();
PropertiesMapper.parseImageProperties(properties, builder);
return builder.tableOfContents(context.tableOfContents()).build();
}
private TableCellNode buildTableCell(Context context, Map<String, String> properties, boolean terminal) {
TableCellNode.TableCellNodeBuilder builder = TableCellNode.builder();
PropertiesMapper.parseTableCellProperties(properties, builder);
return builder.terminal(terminal).tableOfContents(context.tableOfContents()).build();
}
private TableNode buildTable(Context context, Map<String, String> properties) {
TableNode.TableNodeBuilder builder = TableNode.builder();
PropertiesMapper.parseTableProperties(properties, builder);
return TableNode.builder().tableOfContents(context.tableOfContents()).build();
}
private FooterNode buildFooter(Context context, boolean terminal) {
return FooterNode.builder().terminal(terminal).tableOfContents(context.tableOfContents()).build();
}
private HeaderNode buildHeader(Context context, boolean terminal) {
return HeaderNode.builder().terminal(terminal).tableOfContents(context.tableOfContents()).build();
}
private SectionNode buildSection(Context context) {
return SectionNode.builder().tableOfContents(context.tableOfContents()).build();
}
private ParagraphNode buildParagraph(Context context, boolean terminal) {
return ParagraphNode.builder().terminal(terminal).tableOfContents(context.tableOfContents()).build();
}
private TextBlock toTextBlock(Long[] atomicTextBlockIds, Context context, SemanticNode parent) {
return Arrays.stream(atomicTextBlockIds)
.map(atomicTextBlockId -> toAtomicTextBlock(context.atomicTextBlockData.get(Math.toIntExact(atomicTextBlockId)),
context.atomicPositionBlockData.get(Math.toIntExact(atomicTextBlockId)),
parent,
context))
.collect(new TextBlockCollector());
}
private PageNode buildPage(PageData p) {
return PageNode.builder().rotation(p.getRotation()).height(p.getHeight()).width(p.getWidth()).number(p.getNumber()).mainBody(new LinkedList<>()).build();
}
private AtomicTextBlock toAtomicTextBlock(AtomicTextBlockData atomicTextBlockData,
AtomicPositionBlockData atomicPositionBlockData,
SemanticNode parent,
Context context) {
return AtomicTextBlock.builder()
.id(atomicTextBlockData.getId())
.numberOnPage(atomicTextBlockData.getNumberOnPage())
.page(getPage(atomicTextBlockData.getPage(), context))
.boundary(new Boundary(atomicTextBlockData.getStart(), atomicTextBlockData.getEnd()))
.searchText(atomicTextBlockData.getSearchText())
.lineBreaks(Ints.asList(atomicTextBlockData.getLineBreaks()))
.stringIdxToPositionIdx(Ints.asList(atomicPositionBlockData.getStringIdxToPositionIdx()))
.positions(toRectangle2DList(atomicPositionBlockData.getPositions()))
.parent(parent)
.build();
}
private static List<Rectangle2D> toRectangle2DList(float[][] positions) {
return Arrays.stream(positions).map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3])).toList();
}
private PageNode getPage(Long pageIndex, Context context) {
return context.pages.stream()
.filter(page -> page.getNumber() == Math.toIntExact(pageIndex))
.findFirst()
.orElseThrow(() -> new NoSuchElementException(String.format("ClassificationPage with number %d not found", pageIndex)));
}
record Context(
DocumentData layoutParsingModel,
TableOfContents tableOfContents,
List<PageNode> pages,
List<SectionNode> sections,
List<AtomicTextBlockData> atomicTextBlockData,
List<AtomicPositionBlockData> atomicPositionBlockData) {
}
}
@@ -1,101 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.mapper;
import java.awt.geom.Rectangle2D;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ImageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ImageType;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableCellNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableNode;
import com.knecon.fforesight.service.layoutparser.internal.api.services.RectangleTransformations;
public class PropertiesMapper {
public static Map<String, String> buildImageProperties(ImageNode image) {
Map<String, String> properties = new HashMap<>();
properties.put("imageType", image.getImageType().toString());
properties.put("transparency", String.valueOf(image.isTransparency()));
properties.put("position", RectangleTransformations.toString(image.getPosition()));
return properties;
}
public static Map<String, String> buildTableCellProperties(TableCellNode tableCell) {
Map<String, String> properties = new HashMap<>();
properties.put("row", String.valueOf(tableCell.getRow()));
properties.put("col", String.valueOf(tableCell.getCol()));
properties.put("header", String.valueOf(tableCell.isHeader()));
if (tableCell.getPages().size() > 1 || tableCell.getBBox().keySet().size() > 1) {
throw new IllegalArgumentException("TableCell can only occur on a single page!");
}
String bBoxString = RectangleTransformations.toString(tableCell.getBBox().get(tableCell.getPages().stream().findFirst().get()));
properties.put("bBox", bBoxString);
return properties;
}
public static Map<String, String> buildTableProperties(TableNode table) {
Map<String, String> properties = new HashMap<>();
properties.put("numberOfRows", String.valueOf(table.getNumberOfRows()));
properties.put("numberOfCols", String.valueOf(table.getNumberOfCols()));
return properties;
}
public static void parseImageProperties(Map<String, String> properties, ImageNode.ImageNodeBuilder builder) {
builder.imageType(parseImageType(properties.get("imageType")));
builder.transparency(Boolean.parseBoolean(properties.get("transparency")));
builder.position(parseRectangle2D(properties.get("position")));
}
public static void parseTableCellProperties(Map<String, String> properties, TableCellNode.TableCellNodeBuilder builder) {
builder.row(Integer.parseInt(properties.get("row")));
builder.col(Integer.parseInt(properties.get("col")));
builder.header(Boolean.parseBoolean(properties.get("header")));
builder.bBox(parseRectangle2D(properties.get("bBox")));
}
public static void parseTableProperties(Map<String, String> properties, TableNode.TableNodeBuilder builder) {
builder.numberOfRows(Integer.parseInt(properties.get("numberOfRows")));
builder.numberOfCols(Integer.parseInt(properties.get("numberOfCols")));
}
private static ImageType parseImageType(String imageType) {
return switch (imageType) {
case "LOGO" -> ImageType.LOGO;
case "FORMULA" -> ImageType.FORMULA;
case "SIGNATURE" -> ImageType.SIGNATURE;
case "OCR" -> ImageType.OCR;
default -> ImageType.OTHER;
};
}
public static String toString(Rectangle2D rectangle2D) {
return String.format("%f,%f,%f,%f", rectangle2D.getX(), rectangle2D.getY(), rectangle2D.getWidth(), rectangle2D.getHeight());
}
public static Rectangle2D parseRectangle2D(String bBox) {
List<Float> floats = Arrays.stream(bBox.split(",")).map(Float::parseFloat).toList();
return new Rectangle2D.Float(floats.get(0), floats.get(1), floats.get(2), floats.get(3));
}
}
@@ -2,7 +2,7 @@ package com.knecon.fforesight.service.layoutparser.internal.api.queue;
public class LayoutParsingQueueNames {
public static final String LAYOUT_PARSING_REQUEST_QUEUE = "LAYOUTPARSING_REQUEST_QUEUE";
public static final String LAYOUT_PARSING_DLQ = "LAYOUTPARSING_DLQ";
public static final String LAYOUT_PARSING_FINISHED_EVENT_QUEUE = "LAYOUTPARSING_FINISHED_EVENT_QUEUE";
public static final String LAYOUT_PARSING_REQUEST_QUEUE = "layout_parsing_request_queue";
public static final String LAYOUT_PARSING_DLQ = "layout_parsing_dead_letter_queue";
public static final String LAYOUT_PARSING_FINISHED_EVENT_QUEUE = "layout_parsing_response_queue";
}
@@ -8,13 +8,16 @@ import lombok.Builder;
@Builder
public record LayoutParsingRequest(
LayoutParsingType layoutParsingType,
Map<String, String> identifier,
String originFileStorageId,
Optional<String> tablesFileStorageId,
Optional<String> imagesFileStorageId,
String structureFileStorageId,
String researchDocumentStorageId,
String textBlockFileStorageId,
String positionBlockFileStorageId,
String pageFileStorageId) {
String pageFileStorageId,
String sectionGridStorageId) {
}
@@ -0,0 +1,7 @@
package com.knecon.fforesight.service.layoutparser.internal.api.queue;
public enum LayoutParsingType {
REDACT_MANAGER,
TAAS,
DOCUMINE
}
@@ -1,10 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.services;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
public interface EntityEnrichmentService {
void enrichEntity(EntityNode entity, TextBlock textBlock);
}
@@ -1,56 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.services;
import java.util.Collections;
import java.util.NoSuchElementException;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import lombok.RequiredArgsConstructor;
@RequiredArgsConstructor
public class EntityInsertionService {
private final EntityEnrichmentService entityEnrichmentService;
public void addEntityToGraph(EntityNode entity, TableOfContents tableOfContents) {
try {
SemanticNode containingNode = tableOfContents.streamChildrenNodes(Collections.emptyList())
.filter(node -> node.buildTextBlock().containsBoundary(entity.getBoundary()))
.findFirst()
.orElseThrow(() -> new NoSuchElementException("No containing Node found!"));
containingNode.addThisToEntityIfIntersects(entity);
TextBlock textBlock = entity.getDeepestFullyContainingNode().buildTextBlock();
entityEnrichmentService.enrichEntity(entity, textBlock);
addToPages(entity);
addToNodeEntitySets(entity);
} catch (NoSuchElementException e) {
entity.removeFromGraph();
}
}
private void addToPages(EntityNode entity) {
Set<PageNode> pages = entity.getDeepestFullyContainingNode().getPages();
entity.getPages().addAll(pages);
pages.forEach(page -> page.getEntities().add(entity));
}
private void addToNodeEntitySets(EntityNode entity) {
entity.getIntersectingNodes().forEach(node -> node.getEntities().add(entity));
}
}
@@ -1,95 +0,0 @@
package com.knecon.fforesight.service.layoutparser.internal.api.services;
import static java.lang.String.format;
import java.awt.geom.Area;
import java.awt.geom.Rectangle2D;
import java.util.Arrays;
import java.util.List;
import java.util.Set;
import java.util.function.BiConsumer;
import java.util.function.BinaryOperator;
import java.util.function.Function;
import java.util.function.Supplier;
import java.util.stream.Collector;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.AtomicTextBlock;
import lombok.experimental.UtilityClass;
@UtilityClass
public class RectangleTransformations {
public static Rectangle2D pad(Rectangle2D rectangle2D, int deltaX, int deltaY) {
return new Rectangle2D.Double(rectangle2D.getMinX() - deltaX, rectangle2D.getMinY() - deltaY, rectangle2D.getWidth() + 2 * deltaX, rectangle2D.getHeight() + 2 * deltaY);
}
public static Rectangle2D bBoxUnionAtomicTextBlock(List<AtomicTextBlock> atomicTextBlocks) {
return atomicTextBlocks.stream().flatMap(atomicTextBlock -> atomicTextBlock.getPositions().stream()).collect(new Rectangle2DUnion());
}
public static Rectangle2D rectangleUnion(List<Rectangle2D> rectangle2DList) {
return rectangle2DList.stream().collect(new Rectangle2DUnion());
}
public static String toString(Rectangle2D rectangle2D) {
return format("%f,%f,%f,%f", rectangle2D.getX(), rectangle2D.getY(), rectangle2D.getWidth(), rectangle2D.getHeight());
}
public static Rectangle2D parseRectangle2D(String bBox) {
List<Float> floats = Arrays.stream(bBox.split(",")).map(Float::parseFloat).toList();
return new Rectangle2D.Float(floats.get(0), floats.get(1), floats.get(2), floats.get(3));
}
private static class Rectangle2DUnion implements Collector<Rectangle2D, Area, Rectangle2D> {
@Override
public Supplier<Area> supplier() {
return Area::new;
}
@Override
public BiConsumer<Area, Rectangle2D> accumulator() {
return (area, rectangle2D) -> area.add(new Area(rectangle2D));
}
@Override
public BinaryOperator<Area> combiner() {
return (area1, area2) -> {
area1.add(area2);
return area1;
};
}
@Override
public Function<Area, Rectangle2D> finisher() {
return Area::getBounds2D;
}
@Override
public Set<Characteristics> characteristics() {
return Set.of(Characteristics.CONCURRENT, Characteristics.UNORDERED);
}
}
}
@@ -0,0 +1,16 @@
<Configuration>
<Appenders>
<Console name="CONSOLE" target="SYSTEM_OUT">
<PatternLayout pattern="%d{HH:mm:ss.SSS} [%t] %-5level %logger{36} - %msg%n"/>
</Console>
</Appenders>
<Loggers>
<Root level="warn">
<AppenderRef ref="CONSOLE"/>
</Root>
<Logger name="com.iqser" level="info"/>
</Loggers>
</Configuration>
@@ -6,7 +6,7 @@
<parent>
<groupId>com.knecon.fforesight</groupId>
<artifactId>layoutparser-service</artifactId>
<version>1.0.0</version>
<version>0.1-SNAPSHOT</version>
</parent>
<artifactId>layoutparser-service-processor</artifactId>
@@ -14,9 +14,14 @@
<dependencies>
<dependency>
<groupId>com.iqser.red.service</groupId>
<artifactId>persistence-service-internal-api-v1</artifactId>
<artifactId>persistence-service-shared-api-v1</artifactId>
<version>2.36.0</version>
</dependency>
<dependency>
<groupId>com.knecon.fforesight</groupId>
<artifactId>tenant-commons</artifactId>
<version>${tennat-commons.version}</version>
</dependency>
<dependency>
<groupId>com.knecon.fforesight</groupId>
<artifactId>layoutparser-service-internal-api</artifactId>
@@ -25,17 +30,7 @@
<dependency>
<groupId>com.iqser.red.commons</groupId>
<artifactId>storage-commons</artifactId>
<version>1.13.0</version>
</dependency>
<dependency>
<groupId>com.iqser.red.commons</groupId>
<artifactId>spring-commons</artifactId>
<version>6.2.0</version>
</dependency>
<dependency>
<groupId>com.dslplatform</groupId>
<artifactId>dsl-json-java8</artifactId>
<version>1.10.0</version>
<version>${storage-commons.version}</version>
</dependency>
<dependency>
<groupId>org.apache.pdfbox</groupId>
@@ -47,11 +42,6 @@
<artifactId>pdfbox-tools</artifactId>
<version>${pdfbox.version}</version>
</dependency>
<dependency>
<groupId>com.google.guava</groupId>
<artifactId>guava</artifactId>
<version>31.1-jre</version>
</dependency>
<dependency>
<groupId>com.fasterxml.jackson.module</groupId>
<artifactId>jackson-module-afterburner</artifactId>
@@ -62,68 +52,20 @@
<artifactId>jackson-datatype-jsr310</artifactId>
<version>${jackson.version}</version>
</dependency>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-security</artifactId>
<version>${spring.version}</version>
</dependency>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-web</artifactId>
<version>${spring.version}</version>
</dependency>
<dependency>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
<optional>true</optional>
</dependency>
<dependency>
<groupId>org.springframework.cloud</groupId>
<artifactId>spring-cloud-starter-openfeign</artifactId>
<version>4.0.2</version>
</dependency>
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-amqp</artifactId>
<version>${spring.version}</version>
</dependency>
<dependency>
<groupId>org.junit.jupiter</groupId>
<artifactId>junit-jupiter</artifactId>
<version>RELEASE</version>
<scope>test</scope>
</dependency>
</dependencies>
<repositories>
<repository>
<id>spring-milestones</id>
<name>Spring Milestones</name>
<url>https://repo.spring.io/milestone</url>
<snapshots>
<enabled>false</enabled>
</snapshots>
</repository>
<repository>
<id>spring-snapshots</id>
<name>Spring Snapshots</name>
<url>https://repo.spring.io/snapshot</url>
<releases>
<enabled>false</enabled>
</releases>
</repository>
</repositories>
<pluginRepositories>
<pluginRepository>
<id>spring-milestones</id>
<name>Spring Milestones</name>
<url>https://repo.spring.io/milestone</url>
<snapshots>
<enabled>false</enabled>
</snapshots>
</pluginRepository>
<pluginRepository>
<id>spring-snapshots</id>
<name>Spring Snapshots</name>
<url>https://repo.spring.io/snapshot</url>
<releases>
<enabled>false</enabled>
</releases>
</pluginRepository>
</pluginRepositories>
</project>
@@ -0,0 +1,141 @@
package com.knecon.fforesight.service.layoutparser.processor;
import static java.lang.String.format;
import java.io.IOException;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
import com.knecon.fforesight.service.layoutparser.processor.adapter.CvTableParsingAdapter;
import com.knecon.fforesight.service.layoutparser.processor.adapter.ImageServiceResponseAdapter;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.image.ImageServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.factory.DocumentGraphFactory;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.mapper.redaction.DocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.mapper.taas.TaasDocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
import com.knecon.fforesight.service.layoutparser.processor.services.PdfParsingService;
import com.knecon.fforesight.service.layoutparser.processor.services.SectionGridCreatorService;
import com.knecon.fforesight.service.layoutparser.processor.services.SectionsBuilderService;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.DocuMineClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.RedactManagerClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.services.classification.TaasClassificationService;
import lombok.RequiredArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@Service
@RequiredArgsConstructor
public class LayoutParsingPipeline {
private final ImageServiceResponseAdapter imageServiceResponseAdapter;
private final CvTableParsingAdapter cvTableParsingAdapter;
private final LayoutParsingStorageService layoutParsingStorageService;
private final PdfParsingService pdfParsingService;
private final SectionsBuilderService sectionsBuilderService;
private final SectionGridCreatorService sectionGridCreatorService;
private final TaasClassificationService taasClassificationService;
private final RedactManagerClassificationService redactManagerClassificationService;
private final DocuMineClassificationService docuMineClassificationService;
public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
long start = System.currentTimeMillis();
PDDocument originDocument = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
if (layoutParsingRequest.imagesFileStorageId().isPresent()) {
imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.pageFileStorageId());
}
TableServiceResponse tableServiceResponse = new TableServiceResponse();
if (layoutParsingRequest.tablesFileStorageId().isPresent()) {
tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.pageFileStorageId());
}
Document documentGraph = parseLayout(layoutParsingRequest.layoutParsingType(), originDocument, imageServiceResponse, tableServiceResponse);
int numberOfPages = originDocument.getNumberOfPages();
originDocument.close();
layoutParsingStorageService.storeSectionGrid(layoutParsingRequest, sectionGridCreatorService.createSectionGrid(documentGraph));
layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
if (layoutParsingRequest.layoutParsingType().equals(LayoutParsingType.TAAS)) {
var researchDocumentData = TaasDocumentDataMapper.fromDocument(documentGraph);
layoutParsingStorageService.storeResearchDocumentData(layoutParsingRequest, researchDocumentData);
}
return LayoutParsingFinishedEvent.builder()
.identifier(layoutParsingRequest.identifier())
.numberOfPages(numberOfPages)
.duration(System.currentTimeMillis() - start)
.message(format("Layout parsing is finished and files have been saved with Ids:\n Structure: %s\nText: %s\nPositions: %s\nPageData: %s",
layoutParsingRequest.structureFileStorageId(),
layoutParsingRequest.textBlockFileStorageId(),
layoutParsingRequest.positionBlockFileStorageId(),
layoutParsingRequest.pageFileStorageId()))
.build();
}
public Document parseLayout(LayoutParsingType layoutParsingType,
PDDocument originDocument,
ImageServiceResponse imageServiceResponse,
TableServiceResponse tableServiceResponse) {
ClassificationDocument classificationDocument = pdfParsingService.parseDocument(layoutParsingType,
originDocument,
cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse),
imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse));
switch (layoutParsingType) {
case TAAS -> taasClassificationService.classifyDocument(classificationDocument);
case DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
case REDACT_MANAGER -> redactManagerClassificationService.classifyDocument(classificationDocument);
}
sectionsBuilderService.buildSections(classificationDocument);
return DocumentGraphFactory.buildDocumentGraph(classificationDocument);
}
public Document parseLayoutWithTimer(LayoutParsingType layoutParsingType,
PDDocument originDocument,
ImageServiceResponse imageServiceResponse,
TableServiceResponse tableServiceResponse) {
long start = System.currentTimeMillis();
ClassificationDocument classificationDocument = pdfParsingService.parseDocument(layoutParsingType, originDocument,
cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse),
imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse));
System.out.printf("parsed %d ms", System.currentTimeMillis() - start);
start = System.currentTimeMillis();
switch (layoutParsingType) {
case TAAS -> taasClassificationService.classifyDocument(classificationDocument);
case DOCUMINE -> docuMineClassificationService.classifyDocument(classificationDocument);
case REDACT_MANAGER -> redactManagerClassificationService.classifyDocument(classificationDocument);
}
System.out.printf(", classified %d ms", System.currentTimeMillis() - start);
start = System.currentTimeMillis();
sectionsBuilderService.buildSections(classificationDocument);
System.out.printf(", sections built %d ms", System.currentTimeMillis() - start);
start = System.currentTimeMillis();
Document document = DocumentGraphFactory.buildDocumentGraph(classificationDocument);
System.out.printf(", graph constructed %d ms", System.currentTimeMillis() - start);
return document;
}
}
@@ -1,88 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor;
import static java.lang.String.format;
import java.io.IOException;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.DocumentGraph;
import com.knecon.fforesight.service.layoutparser.internal.api.mapper.DocumentDataMapper;
import com.knecon.fforesight.service.layoutparser.processor.adapter.CvTableParsingAdapter;
import com.knecon.fforesight.service.layoutparser.processor.adapter.ImageServiceResponseAdapter;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.image.ImageServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.ClassificationDocument;
import com.knecon.fforesight.service.layoutparser.processor.classification.service.ClassificationService;
import com.knecon.fforesight.service.layoutparser.processor.classification.service.PdfParsingService;
import com.knecon.fforesight.service.layoutparser.processor.classification.service.SectionsBuilderService;
import com.knecon.fforesight.service.layoutparser.processor.factory.DocumentGraphFactory;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingFinishedEvent;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import lombok.RequiredArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@Service
@RequiredArgsConstructor
public class LayoutParsingService {
private final ImageServiceResponseAdapter imageServiceResponseAdapter;
private final CvTableParsingAdapter cvTableParsingAdapter;
private final LayoutParsingStorageService layoutParsingStorageService;
private final PdfParsingService pdfParsingService;
private final ClassificationService classificationService;
private final SectionsBuilderService sectionsBuilderService;
private final DocumentGraphFactory documentGraphFactory;
public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
long start = System.currentTimeMillis();
PDDocument originDocument = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
if (layoutParsingRequest.imagesFileStorageId().isPresent()) {
imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.pageFileStorageId());
}
TableServiceResponse tableServiceResponse = new TableServiceResponse();
if (layoutParsingRequest.tablesFileStorageId().isPresent()) {
tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.pageFileStorageId());
}
DocumentGraph documentGraph = parseLayout(originDocument, imageServiceResponse, tableServiceResponse);
int numberOfPages = originDocument.getNumberOfPages();
originDocument.close();
layoutParsingStorageService.storeDocumentData(layoutParsingRequest, DocumentDataMapper.toDocumentData(documentGraph));
return LayoutParsingFinishedEvent.builder()
.identifier(layoutParsingRequest.identifier())
.numberOfPages(numberOfPages)
.duration(System.currentTimeMillis() - start)
.message(format("Layout parsing is finished and files have been saved with Ids:\n Structure: %s\nText: %s\nPositions: %s\nPageData: %s",
layoutParsingRequest.structureFileStorageId(),
layoutParsingRequest.textBlockFileStorageId(),
layoutParsingRequest.positionBlockFileStorageId(),
layoutParsingRequest.pageFileStorageId()))
.build();
}
public DocumentGraph parseLayout(PDDocument originDocument, ImageServiceResponse imageServiceResponse, TableServiceResponse tableServiceResponse) {
ClassificationDocument classificationDocument = pdfParsingService.parseDocument(originDocument,
cvTableParsingAdapter.buildCvParsedTablesPerPage(tableServiceResponse),
imageServiceResponseAdapter.buildClassifiedImagesPerPage(imageServiceResponse));
classificationService.classifyDocument(classificationDocument);
sectionsBuilderService.buildSections(classificationDocument);
return documentGraphFactory.buildDocumentGraph(classificationDocument);
}
}
@@ -13,16 +13,18 @@ import org.apache.pdfbox.pdmodel.PDDocument;
import org.springframework.stereotype.Service;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.section.SectionGrid;
import com.iqser.red.storage.commons.service.StorageService;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicPositionBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.AtomicTextBlockData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.PageData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.TableOfContentsData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentStructure;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPage;
import com.knecon.fforesight.service.layoutparser.internal.api.data.taas.ResearchDocumentData;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.image.ImageServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.multitenancy.TenantContext;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingRequest;
import com.knecon.fforesight.tenantcommons.TenantContext;
import lombok.RequiredArgsConstructor;
import lombok.extern.slf4j.Slf4j;
@@ -67,34 +69,43 @@ public class LayoutParsingStorageService {
}
public void storeDocumentData(LayoutParsingRequest layoutParsingRequest, DocumentData documentData) throws IOException {
public void storeDocumentData(LayoutParsingRequest layoutParsingRequest, DocumentData documentData) {
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.structureFileStorageId(), documentData.getTableOfContents());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.textBlockFileStorageId(), documentData.getAtomicTextBlocks());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.positionBlockFileStorageId(), documentData.getAtomicPositionBlocks());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.pageFileStorageId(), documentData.getPages());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.structureFileStorageId(), documentData.getDocumentStructure());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.textBlockFileStorageId(), documentData.getDocumentTextData());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.positionBlockFileStorageId(), documentData.getDocumentPositions());
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.pageFileStorageId(), documentData.getDocumentPages());
}
public void storeSectionGrid(LayoutParsingRequest layoutParsingRequest, SectionGrid sectionGrid) {
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.sectionGridStorageId(), sectionGrid);
}
public void storeResearchDocumentData(LayoutParsingRequest layoutParsingRequest, ResearchDocumentData researchDocumentData) {
storageService.storeJSONObject(TenantContext.getTenantId(), layoutParsingRequest.researchDocumentStorageId(), researchDocumentData);
}
public DocumentData readDocumentData(LayoutParsingRequest layoutParsingRequest) throws IOException {
PageData[] pageData = storageService.readJSONObject(TenantContext.getTenantId(), layoutParsingRequest.pageFileStorageId(), PageData[].class);
AtomicTextBlockData[] atomicTextBlockData = storageService.readJSONObject(TenantContext.getTenantId(),
DocumentPage[] documentPageData = storageService.readJSONObject(TenantContext.getTenantId(), layoutParsingRequest.pageFileStorageId(), DocumentPage[].class);
DocumentTextData[] documentTextDataBlockData = storageService.readJSONObject(TenantContext.getTenantId(),
layoutParsingRequest.textBlockFileStorageId(),
AtomicTextBlockData[].class);
AtomicPositionBlockData[] atomicPositionBlockData = storageService.readJSONObject(TenantContext.getTenantId(),
DocumentTextData[].class);
DocumentPositionData[] atomicPositionBlockData = storageService.readJSONObject(TenantContext.getTenantId(),
layoutParsingRequest.positionBlockFileStorageId(),
AtomicPositionBlockData[].class);
TableOfContentsData tableOfContentsData = storageService.readJSONObject(TenantContext.getTenantId(),
layoutParsingRequest.structureFileStorageId(),
TableOfContentsData.class);
DocumentPositionData[].class);
DocumentStructure tableOfContentsData = storageService.readJSONObject(TenantContext.getTenantId(), layoutParsingRequest.structureFileStorageId(), DocumentStructure.class);
return DocumentData.builder()
.tableOfContents(tableOfContentsData)
.atomicPositionBlocks(atomicPositionBlockData)
.atomicTextBlocks(atomicTextBlockData)
.pages(pageData)
.documentStructure(tableOfContentsData)
.documentPositions(atomicPositionBlockData)
.documentTextData(documentTextDataBlockData)
.documentPages(documentPageData)
.build();
}
@@ -8,7 +8,7 @@ import java.util.Map;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.CvParsedTableCell;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableCells;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.table.TableServiceResponse;
import lombok.RequiredArgsConstructor;
@@ -19,9 +19,9 @@ import lombok.extern.slf4j.Slf4j;
@RequiredArgsConstructor
public class CvTableParsingAdapter {
public Map<Integer, List<com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.CvParsedTableCell>> buildCvParsedTablesPerPage(TableServiceResponse tableServiceResponse) {
public Map<Integer, List<TableCells>> buildCvParsedTablesPerPage(TableServiceResponse tableServiceResponse) {
Map<Integer, List<com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.CvParsedTableCell>> tableCells = new HashMap<>();
Map<Integer, List<TableCells>> tableCells = new HashMap<>();
tableServiceResponse.getData()
.forEach(tableData -> tableCells.computeIfAbsent(tableData.getPageInfo().getNumber(), tableCell -> new ArrayList<>())
.addAll(convertTableCells(tableData.getTableCells())));
@@ -30,11 +30,11 @@ public class CvTableParsingAdapter {
}
private Collection<? extends com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.CvParsedTableCell> convertTableCells(List<CvParsedTableCell> tableCells) {
private Collection<TableCells> convertTableCells(List<TableCells> tableCells) {
List<com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.CvParsedTableCell> cvParsedTableCells = new ArrayList<>();
List<TableCells> cvParsedTableCells = new ArrayList<>();
tableCells.forEach(t -> cvParsedTableCells.add(com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.CvParsedTableCell.builder()
tableCells.forEach(t -> cvParsedTableCells.add(TableCells.builder()
.y0(t.getY0())
.x1(t.getX1())
.y1(t.getY1())
@@ -9,10 +9,10 @@ import java.util.Map;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ImageType;
import com.knecon.fforesight.service.layoutparser.processor.adapter.model.image.ImageServiceResponse;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.ClassificationPage;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.image.ClassifiedImage;
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.graph.nodes.ImageType;
import lombok.RequiredArgsConstructor;
@@ -3,12 +3,9 @@ package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image
import java.util.HashMap;
import java.util.Map;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Classification {
private Map<String, Float> probabilities = new HashMap<>();
@@ -1,14 +1,11 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class FilterGeometry {
private ImageSize imageSize;
private Format imageFormat;
private ImageFormat imageFormat;
}
@@ -1,11 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Filters {
private FilterGeometry geometry;
@@ -1,11 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Geometry {
private float width;
@@ -1,12 +1,9 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Format {
public class ImageFormat {
private float quotient;
private boolean tooTall;
@@ -1,12 +1,9 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Metadata {
public class ImageMetadata {
private Classification classification;
private Position position;
@@ -3,15 +3,12 @@ package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image
import java.util.ArrayList;
import java.util.List;
import com.dslplatform.json.CompiledJson;
import com.dslplatform.json.JsonAttribute;
import com.fasterxml.jackson.annotation.JsonAlias;
import com.fasterxml.jackson.annotation.JsonProperty;
import lombok.Data;
@Data
@CompiledJson
public class ImageServiceResponse {
private String dossierId;
@@ -19,15 +16,13 @@ public class ImageServiceResponse {
@JsonProperty(value = "imageMetadata")
@JsonAlias("data")
@JsonAttribute(alternativeNames = {"imageMetadata"})
private List<Metadata> data = new ArrayList<>();
private List<ImageMetadata> data = new ArrayList<>();
private List<Metadata> dataCV = new ArrayList<>();
private List<ImageMetadata> dataCV = new ArrayList<>();
@JsonProperty(value = "imageMetadata")
@JsonAlias("data")
@JsonAttribute(alternativeNames = {"imageMetadata"})
public void setData(List<Metadata> data) {this.data = data;}
public void setData(List<ImageMetadata> data) {this.data = data;}
}
@@ -1,11 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class ImageSize {
private float quotient;
@@ -1,11 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Position {
private float x1;
@@ -1,11 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.image;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class Probability {
private boolean unconfident;
@@ -1,17 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.table;
import java.util.ArrayList;
import java.util.List;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class CvParsedTableModel {
private CvParsedPageInfo pageInfo;
private List<CvParsedTableCell> tableCells = new ArrayList<>();
}
@@ -1,12 +1,9 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.table;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class CvParsedPageInfo {
public class PageInfo {
private int number;
private int rotation;
@@ -1,4 +1,4 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto.table;
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.table;
import lombok.AllArgsConstructor;
import lombok.Builder;
@@ -9,7 +9,7 @@ import lombok.RequiredArgsConstructor;
@Builder
@AllArgsConstructor
@RequiredArgsConstructor
public class CvParsedTableCell {
public class PdfTableCell {
private float x0;
private float y0;
@@ -1,12 +1,11 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.table;
import com.dslplatform.json.CompiledJson;
import lombok.Builder;
import lombok.Data;
@Data
@CompiledJson
public class CvParsedTableCell {
@Builder
public class TableCells {
private float x0;
private float y0;
@@ -0,0 +1,14 @@
package com.knecon.fforesight.service.layoutparser.processor.adapter.model.table;
import java.util.ArrayList;
import java.util.List;
import lombok.Data;
@Data
public class TableData {
private PageInfo pageInfo;
private List<TableCells> tableCells = new ArrayList<>();
}
@@ -3,12 +3,9 @@ package com.knecon.fforesight.service.layoutparser.processor.adapter.model.table
import java.util.ArrayList;
import java.util.List;
import com.dslplatform.json.CompiledJson;
import lombok.Data;
@Data
@CompiledJson
public class TableServiceResponse {
private String dossierId;
@@ -17,6 +14,6 @@ public class TableServiceResponse {
private String targetFileExtension;
private String responseFileExtension;
private List<CvParsedTableModel> data = new ArrayList<>();
private List<TableData> data = new ArrayList<>();
}
@@ -1,71 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto;
import java.awt.geom.Rectangle2D;
import com.dslplatform.json.JsonAttribute;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.ClassificationTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextBlockOrientation;
import lombok.AllArgsConstructor;
import lombok.Data;
import lombok.NoArgsConstructor;
@Data
@AllArgsConstructor
@NoArgsConstructor
public abstract class AbstractTextContainer {
protected float minX;
protected float maxX;
protected float minY;
protected float maxY;
protected String classification;
protected int page;
private TextBlockOrientation orientation = TextBlockOrientation.NONE;
public abstract String getText();
public boolean containsBlock(ClassificationTextBlock other) {
return this.minX <= other.getPdfMinX() && this.maxX >= other.getPdfMaxX() && this.minY >= other.getPdfMinY() && this.maxY <= other.getPdfMaxY();
}
public boolean contains(AbstractTextContainer other) {
return this.minX <= other.minX && this.maxX >= other.maxX && this.minY >= other.minY && this.maxY <= other.maxY;
}
public boolean contains(Rectangle2D other) {
return other.contains(minX, minY, getWidth(), getHeight());
}
@JsonIgnore
@JsonAttribute(ignore = true)
public float getHeight() {
return maxY - minY;
}
@JsonIgnore
@JsonAttribute(ignore = true)
public float getWidth() {
return maxX - minX;
}
public boolean intersectsY(AbstractTextContainer atc) {
return this.minY <= atc.getMaxY() && this.maxY >= atc.getMinY();
}
}
@@ -1,16 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.ClassificationTextBlock;
import lombok.AllArgsConstructor;
import lombok.Data;
@Data
@AllArgsConstructor
public class ClassificationFooter {
private List<ClassificationTextBlock> textBlocks;
}
@@ -1,16 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.ClassificationTextBlock;
import lombok.AllArgsConstructor;
import lombok.Data;
@Data
@AllArgsConstructor
public class ClassificationHeader {
private List<ClassificationTextBlock> textBlocks;
}
@@ -1,38 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto;
import java.util.ArrayList;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.image.ClassifiedImage;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.Table;
import lombok.Data;
import lombok.NoArgsConstructor;
@Data
@NoArgsConstructor
public class ClassificationSection implements Comparable {
private List<AbstractTextContainer> pageBlocks = new ArrayList<>();
private List<ClassifiedImage> images = new ArrayList<>();
private String headline;
public List<Table> getTables() {
List<Table> tables = new ArrayList<>();
pageBlocks.forEach(block -> {
if (block instanceof Table) {
tables.add((Table) block);
}
});
return tables;
}
@Override
public int compareTo(Object o) {
return 0;
}
}
@@ -1,38 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto.table;
import java.awt.geom.Point2D;
import java.util.ArrayList;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.ClassificationTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.Rectangle;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
@SuppressWarnings("serial")
@Data
@EqualsAndHashCode(callSuper = true)
@NoArgsConstructor
public class TableCell extends Rectangle {
private List<ClassificationTextBlock> textBlocks = new ArrayList<>();
private List<TableCell> headerCells = new ArrayList<>();
private boolean isHeaderCell;
public TableCell(Point2D topLeft, Point2D bottomRight) {
super((float) topLeft.getY(), (float) topLeft.getX(), (float) (bottomRight.getX() - topLeft.getX()), (float) (bottomRight.getY() - topLeft.getY()));
}
public void addTextBlock(ClassificationTextBlock textBlock) {
textBlocks.add(textBlock);
}
}
@@ -1,8 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.dto.text;
public enum TextBlockOrientation {
NONE,
LEFT,
RIGHT
}
@@ -1,82 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.classification.parsing;
import java.io.IOException;
import java.util.ArrayList;
import java.util.List;
import org.apache.pdfbox.text.PDFTextStripperByArea;
import org.apache.pdfbox.text.TextPosition;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextPositionSequence;
import lombok.Getter;
import lombok.Setter;
public class PDFAreaTextStripper extends PDFTextStripperByArea {
@Getter
private List<TextPositionSequence> textPositionSequences = new ArrayList<>();
@Setter
private int pageNumber;
public PDFAreaTextStripper() throws IOException {
}
@Override
public void writeString(String text, List<TextPosition> textPositions) throws IOException {
int startIndex = 0;
for (int i = 0; i <= textPositions.size() - 1; i++) {
if (i == 0 && (textPositions.get(i).getUnicode().equals(" ") || textPositions.get(i).getUnicode().equals("\u00A0"))) {
startIndex++;
continue;
}
// Strange but sometimes this is happening, for example: Metolachlor2.pdf
if (i > 0 && textPositions.get(i).getX() < textPositions.get(i - 1).getX()) {
List<TextPosition> sublist = textPositions.subList(startIndex, i);
if (!(sublist.isEmpty() || sublist.size() == 1 && (sublist.get(0).getUnicode().equals(" ") || sublist.get(0).getUnicode().equals("\u00A0")))) {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
}
startIndex = i;
}
if (textPositions.get(i).getRotation() == 0 && i > 0 && textPositions.get(i).getX() > textPositions.get(i - 1).getEndX() + 1) {
List<TextPosition> sublist = textPositions.subList(startIndex, i);
if (!(sublist.isEmpty() || sublist.size() == 1 && (sublist.get(0).getUnicode().equals(" ") || sublist.get(0).getUnicode().equals("\u00A0")))) {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
}
startIndex = i;
}
if (i > 0 && (textPositions.get(i).getUnicode().equals(" ") || textPositions.get(i).getUnicode().equals("\u00A0")) && i <= textPositions.size() - 2) {
List<TextPosition> sublist = textPositions.subList(startIndex, i);
if (!(sublist.isEmpty() || sublist.size() == 1 && (sublist.get(0).getUnicode().equals(" ") || sublist.get(0).getUnicode().equals("\u00A0")))) {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
}
startIndex = i + 1;
}
}
List<TextPosition> sublist = textPositions.subList(startIndex, textPositions.size());
if (!sublist.isEmpty() && (sublist.get(sublist.size() - 1).getUnicode().equals(" ") || sublist.get(sublist.size() - 1).getUnicode().equals("\u00A0"))) {
sublist = sublist.subList(0, sublist.size() - 1);
}
if (!(sublist.isEmpty() || sublist.size() == 1 && (sublist.get(0).getUnicode().equals(" ") || sublist.get(0).getUnicode().equals("\u00A0")))) {
textPositionSequences.add(new TextPositionSequence(sublist, pageNumber));
}
super.writeString(text);
}
public void clearPositions() {
textPositionSequences = new ArrayList<>();
}
}
@@ -4,286 +4,125 @@ import static java.lang.String.format;
import static java.util.stream.Collectors.groupingBy;
import static java.util.stream.Collectors.toList;
import java.util.Collections;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import java.util.NoSuchElementException;
import java.util.Set;
import java.util.concurrent.atomic.AtomicInteger;
import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationDocument;
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.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Footer;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Header;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Headline;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Image;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Paragraph;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Section;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.utils.IdBuilder;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.DocumentGraph;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.TableOfContents;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.FooterNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.HeaderNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.HeadlineNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ImageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.NodeType;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.ParagraphNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SectionNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableCellNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.TableNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.AbstractTextContainer;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.ClassificationDocument;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.ClassificationFooter;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.ClassificationHeader;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.ClassificationPage;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.image.ClassifiedImage;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.Table;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.table.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.ClassificationTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextPositionSequence;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
import lombok.experimental.UtilityClass;
@Service
@UtilityClass
public class DocumentGraphFactory {
public static final double TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD = 0.05;
public Document buildDocumentGraph(ClassificationDocument document) {
Document documentGraph = new Document();
Context context = new Context(documentGraph);
public DocumentGraph buildDocumentGraph(ClassificationDocument document) {
TextBlockFactory textBlockFactory = new TextBlockFactory();
DocumentGraph documentGraph = new DocumentGraph();
Context context = new Context(new TableOfContents(documentGraph), new HashMap<>(), new LinkedList<>(), new LinkedList<>(), textBlockFactory);
document.getPages().stream().map(this::buildPage).forEach(page -> context.pages().put(page, new AtomicInteger(1)));
document.getSections().stream().flatMap(section -> section.getImages().stream()).forEach(image -> context.images().add(image));
document.getPages().forEach(context::buildAndAddPageWithCounter);
document.getSections().stream().flatMap(section -> section.getImages().stream()).forEach(image -> context.getImages().add(image));
addSections(document, context);
addHeaderAndFooterToEachPage(document, context);
documentGraph.setNumberOfPages(context.pages.size());
documentGraph.setPages(context.pages.keySet());
documentGraph.setTableOfContents(context.tableOfContents);
documentGraph.setTextBlock(documentGraph.buildTextBlock());
documentGraph.setDocumentTree(context.documentTree);
documentGraph.setTextBlock(documentGraph.getTextBlock());
return documentGraph;
}
private void addSections(ClassificationDocument document, Context context) {
document.getSections().forEach(section -> addSection(null, section.getPageBlocks(), section.getImages(), context));
document.getSections().forEach(section -> SectionNodeFactory.addSection(null, section.getPageBlocks(), section.getImages(), context));
}
private void addSection(SemanticNode parentNode, List<AbstractTextContainer> pageBlocks, List<ClassifiedImage> images, Context context) {
public void addParagraphOrHeadline(GenericSemanticNode parentNode, TextPageBlock originalTextBlock, Context context, List<TextPageBlock> textBlocksToMerge) {
Map<Integer, List<AbstractTextContainer>> blocksPerPage = pageBlocks.stream().collect(groupingBy(AbstractTextContainer::getPage));
SectionNode sectionNode = SectionNode.builder().entities(new HashSet<>()).tableOfContents(context.tableOfContents()).build();
Page page = context.getPage(originalTextBlock.getPage());
context.sections().add(sectionNode);
blocksPerPage.keySet().forEach(pageNumber -> addSectionNodeToPageNode(context, sectionNode, pageNumber));
List<Integer> tocId;
if (parentNode == null) {
tocId = context.tableOfContents.createNewMainEntryAndReturnId(NodeType.SECTION, sectionNode);
GenericSemanticNode node;
if (originalTextBlock.isHeadline()) {
node = Headline.builder().documentTree(context.getDocumentTree()).build();
} else {
tocId = context.tableOfContents.createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.SECTION, sectionNode);
}
sectionNode.setTocId(tocId);
Set<AbstractTextContainer> alreadyMerged = new HashSet<>();
for (AbstractTextContainer abstractTextContainer : pageBlocks) {
if (alreadyMerged.contains(abstractTextContainer)) {
continue;
}
if (abstractTextContainer instanceof ClassificationTextBlock) {
List<ClassificationTextBlock> textBlocks = findTextBlocksWithSameClassificationAndAlignsY(abstractTextContainer, pageBlocks);
alreadyMerged.addAll(textBlocks);
addParagraphOrHeadline(sectionNode, (ClassificationTextBlock) abstractTextContainer, context, textBlocks);
}
if (abstractTextContainer instanceof Table) {
addTable(sectionNode, (Table) abstractTextContainer, context);
}
}
for (ClassifiedImage image : images) {
addImage(sectionNode, image, context);
}
}
private static List<ClassificationTextBlock> findTextBlocksWithSameClassificationAndAlignsY(AbstractTextContainer atc, List<AbstractTextContainer> pageBlocks) {
return pageBlocks.stream()
.filter(abstractTextContainer -> !abstractTextContainer.equals(atc))
.filter(abstractTextContainer -> abstractTextContainer.getPage() == atc.getPage())
.filter(abstractTextContainer -> abstractTextContainer instanceof ClassificationTextBlock)
.filter(abstractTextContainer -> abstractTextContainer.intersectsY(atc))
.map(abstractTextContainer -> (ClassificationTextBlock) abstractTextContainer)
.toList();
}
private void addSectionNodeToPageNode(Context context, SectionNode sectionNode, Integer pageNumber) {
PageNode page = getPage(pageNumber, context);
page.getMainBody().add(sectionNode);
}
private void addTable(SemanticNode parentNode, Table table, Context context) {
PageNode page = getPage(table.getPage(), context);
TableNode tableNode = TableNode.builder().tableOfContents(context.tableOfContents()).numberOfCols(table.getColCount()).numberOfRows(table.getRowCount()).build();
if (!page.getMainBody().contains(parentNode)) {
parentNode.getPages().add(page);
}
page.getMainBody().add(tableNode);
List<Integer> tocId = context.tableOfContents().createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.TABLE, tableNode);
tableNode.setTocId(tocId);
addTableCells(table.getRows(), tableNode, context, table.getPage());
}
private void addTableCells(List<List<TableCell>> rows, SemanticNode parentNode, Context context, int pageNumber) {
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, parentNode, pageNumber, context);
}
}
}
private void addTableCell(TableCell cell, int rowIndex, int colIndex, SemanticNode parentNode, int pageNumber, Context context) {
PageNode page = getPage(pageNumber, context);
cell.getTextBlocks().stream().filter(tb -> tb.getPage() == 0).forEach(tb -> tb.setPage(pageNumber));
TableCellNode tableCellNode = TableCellNode.builder()
.tableOfContents(context.tableOfContents())
.row(rowIndex)
.col(colIndex)
.header(cell.isHeaderCell())
.bBox(cell.getBounds2D())
.build();
page.getMainBody().add(tableCellNode);
TextBlock textBlock;
List<Integer> tocId = context.tableOfContents().createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.TABLE_CELL, tableCellNode);
tableCellNode.setTocId(tocId);
if (cell.getTextBlocks().isEmpty()) {
tableCellNode.setTerminalTextBlock(context.textBlockFactory.emptyTextBlock(parentNode, context, page));
tableCellNode.setTerminal(true);
} else if (cell.getTextBlocks().size() == 1) {
textBlock = context.textBlockFactory().buildAtomicTextBlock(cell.getTextBlocks().get(0).getSequences(), tableCellNode, context, page);
tableCellNode.setTerminalTextBlock(textBlock);
tableCellNode.setTerminal(true);
} else if (firstTextBlockIsHeadline(cell)) {
addSection(tableCellNode, cell.getTextBlocks().stream().map(tb -> (AbstractTextContainer) tb).toList(), Collections.emptyList(), context);
tableCellNode.setTerminal(false);
} else if (cellAreaIsSmallerThanPageAreaTimesThreshold(cell, page)) {
List<TextPositionSequence> sequences = TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(cell.getTextBlocks());
textBlock = context.textBlockFactory().buildAtomicTextBlock(sequences, tableCellNode, context, page);
tableCellNode.setTerminalTextBlock(textBlock);
tableCellNode.setTerminal(true);
} else {
cell.getTextBlocks().forEach(tb -> addParagraphOrHeadline(tableCellNode, tb, context));
tableCellNode.setTerminal(false);
}
}
private static boolean cellAreaIsSmallerThanPageAreaTimesThreshold(TableCell cell, PageNode page) {
return cell.getArea() < TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD * page.getHeight() * page.getWidth();
}
private static boolean firstTextBlockIsHeadline(TableCell cell) {
String classification = cell.getTextBlocks().get(0).getClassification();
return classification != null && classification.startsWith("H");
}
private void addParagraphOrHeadline(SemanticNode parentNode, ClassificationTextBlock originalTextBlock, Context context) {
addParagraphOrHeadline(parentNode, originalTextBlock, context, Collections.emptyList());
}
private void addParagraphOrHeadline(SemanticNode parentNode, ClassificationTextBlock originalTextBlock, Context context, List<ClassificationTextBlock> textBlocksToMerge) {
PageNode page = getPage(originalTextBlock.getPage(), context);
SemanticNode node;
if (originalTextBlock.getClassification() != null && originalTextBlock.getClassification().startsWith("H")) {
node = HeadlineNode.builder().tableOfContents(context.tableOfContents()).build();
} else {
node = ParagraphNode.builder().tableOfContents(context.tableOfContents()).build();
node = Paragraph.builder().documentTree(context.getDocumentTree()).build();
}
page.getMainBody().add(node);
List<ClassificationTextBlock> textBlocks = new LinkedList<>(textBlocksToMerge);
List<TextPageBlock> textBlocks = new ArrayList<>();
textBlocks.add(originalTextBlock);
textBlocks.addAll(textBlocksToMerge);
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks), node, context, page);
if (node instanceof HeadlineNode headlineNode) {
List<Integer> tocId = context.tableOfContents.createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.HEADLINE, node);
headlineNode.setTerminalTextBlock(textBlock);
headlineNode.setTocId(tocId);
}
if (node instanceof ParagraphNode paragraphNode) {
List<Integer> tocId = context.tableOfContents.createNewChildEntryAndReturnId(parentNode.getTocId(), NodeType.PARAGRAPH, node);
paragraphNode.setTerminalTextBlock(textBlock);
paragraphNode.setTocId(tocId);
}
List<Integer> treeId = context.documentTree.createNewChildEntryAndReturnId(parentNode, node);
node.setLeafTextBlock(textBlock);
node.setTreeId(treeId);
}
private void addImage(SectionNode sectionNode, ClassifiedImage image, Context context) {
public void addImage(Section section, ClassifiedImage image, Context context) {
PageNode page = getPage(image.getPage(), context);
ImageNode imageNode = ImageNode.builder()
Rectangle2D position = image.getPosition();
Page page = context.getPage(image.getPage());
Image imageNode = Image.builder()
.id(IdBuilder.buildId(Set.of(page), List.of(position)))
.imageType(image.getImageType())
.position(image.getPosition())
.transparency(image.isHasTransparency())
.position(position)
.transparent(image.isHasTransparency())
.page(page)
.tableOfContents(context.tableOfContents())
.documentTree(context.getDocumentTree())
.build();
page.getMainBody().add(imageNode);
List<Integer> tocId = context.tableOfContents().createNewChildEntryAndReturnId(sectionNode.getTocId(), NodeType.IMAGE, imageNode);
imageNode.setTocId(tocId);
List<Integer> tocId = context.getDocumentTree().createNewChildEntryAndReturnId(section, imageNode);
imageNode.setTreeId(tocId);
}
private void addHeaderAndFooterToEachPage(ClassificationDocument document, Context context) {
Map<Integer, List<ClassificationTextBlock>> headers = document.getHeaders()
Map<Integer, List<TextPageBlock>> headers = document.getHeaders()
.stream()
.map(ClassificationHeader::getTextBlocks)
.flatMap(List::stream)
.collect(groupingBy(AbstractTextContainer::getPage, toList()));
.collect(groupingBy(AbstractPageBlock::getPage, toList()));
Map<Integer, List<ClassificationTextBlock>> footers = document.getFooters()
Map<Integer, List<TextPageBlock>> footers = document.getFooters()
.stream()
.map(ClassificationFooter::getTextBlocks)
.flatMap(List::stream)
.collect(groupingBy(AbstractTextContainer::getPage, toList()));
.collect(groupingBy(AbstractPageBlock::getPage, toList()));
for (int pageIndex = 1; pageIndex <= document.getPages().size(); pageIndex++) {
if (headers.containsKey(pageIndex)) {
@@ -303,85 +142,105 @@ public class DocumentGraphFactory {
}
private void addFooter(List<ClassificationTextBlock> textBlocks, Context context) {
private void addFooter(List<TextPageBlock> textBlocks, Context context) {
PageNode page = getPage(textBlocks.get(0).getPage(), context);
FooterNode footer = FooterNode.builder().tableOfContents(context.tableOfContents()).build();
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);
List<Integer> tocId = context.tableOfContents().createNewMainEntryAndReturnId(NodeType.FOOTER, footer);
footer.setTocId(tocId);
footer.setTerminalTextBlock(textBlock);
List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
footer.setTreeId(tocId);
footer.setLeafTextBlock(textBlock);
page.setFooter(footer);
}
public void addHeader(List<ClassificationTextBlock> textBlocks, Context context) {
public void addHeader(List<TextPageBlock> textBlocks, Context context) {
PageNode page = getPage(textBlocks.get(0).getPage(), context);
HeaderNode header = HeaderNode.builder().tableOfContents(context.tableOfContents()).build();
AtomicTextBlock textBlock = context.textBlockFactory.buildAtomicTextBlock(TextPositionOperations.mergeAndSortTextPositionSequenceByYThenX(textBlocks),
header,
context,
0,
page);
List<Integer> tocId = context.tableOfContents().createNewMainEntryAndReturnId(NodeType.HEADER, header);
header.setTocId(tocId);
header.setTerminalTextBlock(textBlock);
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);
List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
header.setTreeId(tocId);
header.setLeafTextBlock(textBlock);
page.setHeader(header);
}
private void addEmptyFooter(int pageIndex, Context context) {
PageNode page = getPage(pageIndex, context);
FooterNode footer = FooterNode.builder().tableOfContents(context.tableOfContents()).build();
Page page = context.getPage(pageIndex);
Footer footer = Footer.builder().documentTree(context.getDocumentTree()).build();
AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(footer, context, page);
List<Integer> tocId = context.tableOfContents().createNewMainEntryAndReturnId(NodeType.FOOTER, footer);
footer.setTocId(tocId);
footer.setTerminalTextBlock(textBlock);
List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(footer);
footer.setTreeId(tocId);
footer.setLeafTextBlock(textBlock);
page.setFooter(footer);
}
private void addEmptyHeader(int pageIndex, Context context) {
PageNode page = getPage(pageIndex, context);
HeaderNode header = HeaderNode.builder().tableOfContents(context.tableOfContents()).build();
Page page = context.getPage(pageIndex);
Header header = Header.builder().documentTree(context.getDocumentTree()).build();
AtomicTextBlock textBlock = context.textBlockFactory.emptyTextBlock(header, 0, page);
List<Integer> tocId = context.tableOfContents().createNewMainEntryAndReturnId(NodeType.HEADER, header);
header.setTocId(tocId);
header.setTerminalTextBlock(textBlock);
List<Integer> tocId = context.getDocumentTree().createNewMainEntryAndReturnId(header);
header.setTreeId(tocId);
header.setLeafTextBlock(textBlock);
page.setHeader(header);
}
private PageNode buildPage(ClassificationPage p) {
@Getter
@Builder
@AllArgsConstructor
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public final class Context {
return PageNode.builder()
.height((int) p.getPageHeight())
.width((int) p.getPageWidth())
.number(p.getPageNumber())
.rotation(p.getRotation())
.mainBody(new LinkedList<>())
.build();
}
DocumentTree documentTree;
Map<Page, Integer> pages;
List<Section> sections;
List<ClassifiedImage> images;
TextBlockFactory textBlockFactory;
private PageNode getPage(int pageIndex, Context context) {
public Context(Document document) {
return context.pages.keySet()
.stream()
.filter(page -> page.getNumber() == pageIndex)
.findFirst()
.orElseThrow(() -> new NoSuchElementException(format("ClassificationPage with number %d not found", pageIndex)));
}
documentTree = new DocumentTree(document);
pages = new HashMap<>();
sections = new LinkedList<>();
images = new LinkedList<>();
textBlockFactory = new TextBlockFactory();
}
record Context(
TableOfContents tableOfContents, Map<PageNode, AtomicInteger> pages, List<SectionNode> sections, List<ClassifiedImage> images, TextBlockFactory textBlockFactory) {
public void buildAndAddPageWithCounter(ClassificationPage classificationPage) {
Page page = Page.fromClassificationPage(classificationPage);
//this counter counts the TextBlocks per page
//initial value is set to 1, because 0 is reserved for Header
pages.put(page, 1);
}
public int getAndIncrementTextBlockNumberOnPage(Page page) {
Integer textBlockNumberOnPage = pages.get(page);
pages.merge(page, 1, Integer::sum);
return textBlockNumberOnPage;
}
public Page getPage(int pageIndex) {
return pages.keySet()
.stream()
.filter(page -> page.getNumber() == pageIndex)
.findFirst()
.orElseThrow(() -> new NoSuchElementException(format("ClassificationPage with number %d not found", pageIndex)));
}
}
@@ -1,105 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import static java.lang.String.format;
import java.awt.geom.Area;
import java.awt.geom.Rectangle2D;
import java.util.Arrays;
import java.util.List;
import java.util.Set;
import java.util.function.BiConsumer;
import java.util.function.BinaryOperator;
import java.util.function.Function;
import java.util.function.Supplier;
import java.util.stream.Collector;
import org.apache.pdfbox.pdmodel.common.PDRectangle;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.AbstractTextContainer;
public class RectangleTransformations {
public static Rectangle2D pad(Rectangle2D rectangle2D, int deltaX, int deltaY) {
return new Rectangle2D.Double(rectangle2D.getMinX() - deltaX, rectangle2D.getMinY() - deltaY, rectangle2D.getWidth() + 2 * deltaX, rectangle2D.getHeight() + 2 * deltaY);
}
public static Rectangle2D bBoxUnionAbstractTextContainer(List<AbstractTextContainer> abstractTextContainers) {
return abstractTextContainers.stream().map(RectangleTransformations::toRectangle2D).collect(new Rectangle2DUnion());
}
public static Rectangle2D rectangleUnion(List<Rectangle2D> rectangle2DList) {
return rectangle2DList.stream().collect(new Rectangle2DUnion());
}
public static Rectangle2D toRectangle2D(AbstractTextContainer abstractTextContainer) {
return new Rectangle2D.Float(abstractTextContainer.getMinX(), abstractTextContainer.getMinY(), abstractTextContainer.getWidth(), abstractTextContainer.getHeight());
}
public static Rectangle2D toRectangle2D(PDRectangle rectangle) {
return new Rectangle2D.Double(rectangle.getLowerLeftX(), rectangle.getLowerLeftY(), rectangle.getWidth(), rectangle.getHeight());
}
public static String toString(Rectangle2D rectangle2D) {
return format("%f,%f,%f,%f", rectangle2D.getX(), rectangle2D.getY(), rectangle2D.getWidth(), rectangle2D.getHeight());
}
public static Rectangle2D parseRectangle2D(String bBox) {
List<Float> floats = Arrays.stream(bBox.split(",")).map(Float::parseFloat).toList();
return new Rectangle2D.Float(floats.get(0), floats.get(1), floats.get(2), floats.get(3));
}
private static class Rectangle2DUnion implements Collector<Rectangle2D, Area, Rectangle2D> {
@Override
public Supplier<Area> supplier() {
return Area::new;
}
@Override
public BiConsumer<Area, Rectangle2D> accumulator() {
return (area, rectangle2D) -> area.add(new Area(rectangle2D));
}
@Override
public BinaryOperator<Area> combiner() {
return (area1, area2) -> {
area1.add(area2);
return area1;
};
}
@Override
public Function<Area, Rectangle2D> finisher() {
return Area::getBounds2D;
}
@Override
public Set<Characteristics> characteristics() {
return Set.of(Characteristics.CONCURRENT, Characteristics.UNORDERED);
}
}
}
@@ -0,0 +1,37 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import lombok.AccessLevel;
import lombok.Builder;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@Builder
@Getter
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class SearchTextWithTextPositionDto {
String searchText;
List<Integer> lineBreaks;
List<Integer> stringIdxToPositionIdx;
List<Boundary> boldTextBoundaries;
List<Boundary> italicTextBoundaries;
List<Rectangle2D> positions;
public static SearchTextWithTextPositionDto empty() {
return SearchTextWithTextPositionDto.builder()
.searchText("")
.lineBreaks(Collections.emptyList())
.positions(Collections.emptyList())
.stringIdxToPositionIdx(Collections.emptyList())
.build();
}
}
@@ -7,33 +7,32 @@ import java.util.LinkedList;
import java.util.List;
import java.util.Objects;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.RedTextPosition;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.RedTextPosition;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextDirection;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import lombok.experimental.UtilityClass;
@UtilityClass
public class SearchTextWithTextPositionFactory {
public static final int HEIGHT_PADDING = 2;
public final int HEIGHT_PADDING = 2;
// when checking for a hyphen linebreak, we need to check after a linebreak if the last hyphen was less than three symbols away.
// We detect a linebreak as either a "\n" character or if two adjacent symbol's position differ in y-coordinates by at least one character height.
// If there is a hyphen linebreak, the hyphen will be 1 position in front of a "\n" or 2 positions in front of the character which has a lower y-coordinate
// This is why, we need to initialize this to < -2, otherwise, if the very first symbol is a \n we would detect a hyphen linebreak that isn't there.
// Also, Integer.MIN_VALUE is a bad idea due to potential overflow during arithmetic operations. This is why the default should be -3.
public final int MAX_HYPHEN_LINEBREAK_DISTANCE = 3;
public static SearchTextWithTextPositionModel buildSearchTextToTextPositionModel(List<TextPositionSequence> sequences) {
public SearchTextWithTextPositionDto buildSearchTextToTextPositionDto(List<TextPositionSequence> sequences) {
if (sequences.isEmpty() || sequences.stream().allMatch(sequence -> sequence.getTextPositions().isEmpty())) {
return SearchTextWithTextPositionModel.builder()
.searchText("")
.lineBreaks(Collections.emptyList())
.positions(Collections.emptyList())
.stringCoordsToPositionCoords(Collections.emptyList())
.build();
return SearchTextWithTextPositionDto.empty();
}
List<Integer> stringIdxToPositionIdx = new LinkedList<>();
List<Integer> lineBreaksStringIdx = new LinkedList<>();
StringBuilder sb = new StringBuilder();
int stringIdx = 0;
int positionIdx = 0;
int lastHyphenIdx = -3;
Context context = new Context();
RedTextPosition currentTextPosition = sequences.get(0).getTextPositions().get(0);
RedTextPosition previousTextPosition = RedTextPosition.builder().unicode(" ").position(currentTextPosition.getPosition()).build();
@@ -42,60 +41,109 @@ public class SearchTextWithTextPositionFactory {
for (int i = 0; i < word.getTextPositions().size(); ++i) {
currentTextPosition = word.getTextPositions().get(i);
if (isLineBreak(currentTextPosition, previousTextPosition)) {
if (stringIdx - lastHyphenIdx < 3) {
sb.delete(lastHyphenIdx, sb.length());
stringIdxToPositionIdx = stringIdxToPositionIdx.subList(0, lastHyphenIdx);
stringIdx = lastHyphenIdx;
lastHyphenIdx = -3;
}
lineBreaksStringIdx.add(stringIdx);
removeHyphenLinebreaks(context);
context.lineBreaksStringIdx.add(context.stringIdx);
}
if (!isRepeatedWhitespace(currentTextPosition.getUnicode(), previousTextPosition.getUnicode())) {
if (isHyphen(currentTextPosition.getUnicode())) {
lastHyphenIdx = stringIdx;
context.lastHyphenIdx = context.stringIdx;
}
sb.append(currentTextPosition.getUnicode());
stringIdxToPositionIdx.add(positionIdx);
++stringIdx;
appendCurrentTextPosition(context, currentTextPosition);
}
previousTextPosition = currentTextPosition;
++positionIdx;
++context.positionIdx;
}
previousTextPosition = RedTextPosition.builder().unicode(" ").position(previousTextPosition.getPosition()).build();
sb.append(previousTextPosition.getUnicode());
stringIdxToPositionIdx.add(positionIdx);
++stringIdx;
context.stringBuilder.append(" ");
context.stringIdxToPositionIdx.add(context.positionIdx);
++context.stringIdx;
}
assert sb.length() == stringIdxToPositionIdx.size();
assert context.stringBuilder.length() == context.stringIdxToPositionIdx.size();
List<Rectangle2D> positions = sequences.stream()
.flatMap(sequence -> sequence.getTextPositions().stream().map(textPosition -> mapRedTextPositionToInitialUserSpace(textPosition, sequence)))
.toList();
return SearchTextWithTextPositionModel.builder()
.searchText(sb.toString())
.lineBreaks(lineBreaksStringIdx)
.stringCoordsToPositionCoords(stringIdxToPositionIdx)
return SearchTextWithTextPositionDto.builder()
.searchText(context.stringBuilder.toString())
.lineBreaks(context.lineBreaksStringIdx)
.stringIdxToPositionIdx(context.stringIdxToPositionIdx)
.positions(positions)
.boldTextBoundaries(mergeToBoundaries(context.boldTextsStringIdx))
.italicTextBoundaries(mergeToBoundaries(context.italicTextStringIdx))
.build();
}
private static boolean isLineBreak(RedTextPosition currentTextPosition, RedTextPosition previousTextPosition) {
private void appendCurrentTextPosition(Context context, RedTextPosition currentTextPosition) {
context.stringBuilder.append(currentTextPosition.getUnicode());
// unicode characters with more than 16-bit encoding have a length > 1 in java strings
for (int j = 0; j < currentTextPosition.getUnicode().length(); j++) {
context.stringIdxToPositionIdx.add(context.positionIdx);
addTextPositionWithFontType(currentTextPosition, "bold", context.boldTextsStringIdx, context.stringIdx);
addTextPositionWithFontType(currentTextPosition, "italic", context.italicTextStringIdx, context.stringIdx);
}
context.stringIdx += currentTextPosition.getUnicode().length();
}
private void removeHyphenLinebreaks(Context context) {
if (lastHyphenDirectlyBeforeLineBreak(context)) {
context.stringBuilder.delete(context.lastHyphenIdx, context.stringBuilder.length());
context.stringIdxToPositionIdx = context.stringIdxToPositionIdx.subList(0, context.lastHyphenIdx);
context.stringIdx = context.lastHyphenIdx;
context.lastHyphenIdx = -MAX_HYPHEN_LINEBREAK_DISTANCE;
}
}
private boolean lastHyphenDirectlyBeforeLineBreak(Context context) {
return context.stringIdx - context.lastHyphenIdx < MAX_HYPHEN_LINEBREAK_DISTANCE;
}
private static List<Boundary> mergeToBoundaries(List<Integer> integers) {
if (integers.isEmpty()) {
return Collections.emptyList();
}
List<Boundary> boundaries = new LinkedList<>();
int start = integers.get(0);
int end = integers.get(0) + 1;
for (int current : integers) {
if (current > end + 1) {
boundaries.add(new Boundary(start, end));
start = current;
}
end = current + 1;
}
if (boundaries.isEmpty())
boundaries.add(new Boundary(start, end));
return boundaries;
}
private static void addTextPositionWithFontType(RedTextPosition currentTextPosition, String fontType, List<Integer> fontTypePositions, int stringIdx) {
if (currentTextPosition.getFontName().toLowerCase().contains(fontType)) {
fontTypePositions.add(stringIdx);
}
}
private boolean isLineBreak(RedTextPosition currentTextPosition, RedTextPosition previousTextPosition) {
return Objects.equals(currentTextPosition.getUnicode(), "\n") || isDeltaYLargerThanTextHeight(currentTextPosition, previousTextPosition);
}
private static boolean isDeltaYLargerThanTextHeight(RedTextPosition currentPosition, RedTextPosition previousPosition) {
private boolean isDeltaYLargerThanTextHeight(RedTextPosition currentPosition, RedTextPosition previousPosition) {
if (previousPosition == null) {
return false;
@@ -106,13 +154,13 @@ public class SearchTextWithTextPositionFactory {
}
private static boolean isRepeatedWhitespace(String currentUnicode, String previousUnicode) {
private boolean isRepeatedWhitespace(String currentUnicode, String previousUnicode) {
return Objects.equals(previousUnicode, " ") && Objects.equals(currentUnicode, " ");
}
private static boolean isHyphen(String unicodeCharacter) {
private boolean isHyphen(String unicodeCharacter) {
return Objects.equals(unicodeCharacter, "-") || //
Objects.equals(unicodeCharacter, "~") || //
@@ -128,7 +176,7 @@ public class SearchTextWithTextPositionFactory {
}
private static Rectangle2D mapRedTextPositionToInitialUserSpace(RedTextPosition textPosition, TextPositionSequence sequence) {
private Rectangle2D mapRedTextPositionToInitialUserSpace(RedTextPosition textPosition, TextPositionSequence sequence) {
float textHeight = sequence.getTextHeight() + HEIGHT_PADDING;
Rectangle2D rectangle2D = new Rectangle2D.Double(textPosition.getXDirAdj(),
@@ -153,4 +201,20 @@ public class SearchTextWithTextPositionFactory {
return transform.createTransformedShape(rectangle2D).getBounds2D();
}
private class Context {
List<Integer> stringIdxToPositionIdx = new LinkedList<>();
List<Integer> lineBreaksStringIdx = new LinkedList<>();
List<Integer> boldTextsStringIdx = new LinkedList<>();
List<Integer> italicTextStringIdx = new LinkedList<>();
StringBuilder stringBuilder = new StringBuilder();
int stringIdx;
int positionIdx;
int lastHyphenIdx = -MAX_HYPHEN_LINEBREAK_DISTANCE;
}
}
@@ -1,20 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import java.awt.geom.Rectangle2D;
import java.util.List;
import lombok.AccessLevel;
import lombok.Builder;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@Builder
@Getter
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
public class SearchTextWithTextPositionModel {
String searchText;
List<Integer> lineBreaks;
List<Integer> stringCoordsToPositionCoords;
List<Rectangle2D> positions;
}
@@ -0,0 +1,184 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import static java.lang.String.format;
import static java.util.Collections.emptyList;
import static java.util.stream.Collectors.groupingBy;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
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.image.ClassifiedImage;
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.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Section;
import com.knecon.fforesight.service.layoutparser.processor.utils.TableMergingUtility;
import lombok.experimental.UtilityClass;
@UtilityClass
public class SectionNodeFactory {
public void addSection(GenericSemanticNode parentNode, List<AbstractPageBlock> pageBlocks, List<ClassifiedImage> images, DocumentGraphFactory.Context context) {
if (pageBlocks.isEmpty()) {
return;
}
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));
section.setTreeId(getTreeId(parentNode, context, section));
addFirstHeadlineDirectlyToSection(pageBlocks, context, section);
if (containsTablesAndTextBlocks(pageBlocks)) {
splitPageBlocksIntoSubSections(pageBlocks).forEach(subSectionPageBlocks -> addSection(section, subSectionPageBlocks, emptyList(), context));
} else {
addTablesAndParagraphsAndHeadlinesToSection(pageBlocks, context, section);
}
images.stream().distinct().forEach(image -> DocumentGraphFactory.addImage(section, image, context));
}
private List<Integer> getTreeId(GenericSemanticNode parentNode, DocumentGraphFactory.Context context, Section section) {
if (parentNode == null) {
return context.getDocumentTree().createNewMainEntryAndReturnId(section);
} else {
return context.getDocumentTree().createNewChildEntryAndReturnId(parentNode, section);
}
}
private void addFirstHeadlineDirectlyToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
if (pageBlocks.get(0).isHeadline()) {
addTablesAndParagraphsAndHeadlinesToSection(List.of(pageBlocks.get(0)), context, section);
pageBlocks.remove(0);
}
}
private void addTablesAndParagraphsAndHeadlinesToSection(List<AbstractPageBlock> pageBlocks, DocumentGraphFactory.Context context, Section section) {
Set<AbstractPageBlock> alreadyMerged = new HashSet<>();
List<AbstractPageBlock> remainingBlocks = new LinkedList<>(pageBlocks);
for (AbstractPageBlock abstractPageBlock : pageBlocks) {
if (alreadyMerged.contains(abstractPageBlock)) {
continue;
}
remainingBlocks.removeAll(alreadyMerged);
if (abstractPageBlock instanceof TextPageBlock) {
List<TextPageBlock> textBlocks = findTextBlocksWithSameClassificationAndAlignsY((TextPageBlock) abstractPageBlock, remainingBlocks);
alreadyMerged.addAll(textBlocks);
DocumentGraphFactory.addParagraphOrHeadline(section, (TextPageBlock) abstractPageBlock, context, textBlocks);
} else if (abstractPageBlock instanceof TablePageBlock tablePageBlock) {
List<TablePageBlock> tablesToMerge = TableMergingUtility.findConsecutiveTablesWithSameColCountAndSameHeaders(tablePageBlock, remainingBlocks);
alreadyMerged.addAll(tablesToMerge);
TableNodeFactory.addTable(section, tablesToMerge, context);
} else {
throw new RuntimeException(format("Unhandled AbstractPageBlockType %s!", abstractPageBlock.getClass()));
}
}
}
private boolean containsTablesAndTextBlocks(List<AbstractPageBlock> pageBlocks) {
return pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TablePageBlock) && pageBlocks.stream().anyMatch(pageBlock -> pageBlock instanceof TextPageBlock);
}
/**
* This function splits the list of PageBlocks around TablePageBlocks, such that SubSections can be created, that don't include tables.
* This is needed so we can execute rules on sections, that do not contain tables.
* See: <a href="https://knecon.atlassian.net/wiki/spaces/RED/pages/14765218/Document+Structure">document structure wiki</a>
*
* @param pageBlocks a List of AbstractPageBlocks, which have at least one TablePageBlock and one ClassificationTextBlock
* @return List of Lists of AbstractPageBlocks, which include either a single Headline ClassificationTextBlock and a TablePageBlock or only ClassificationTextBlocks.
*/
private List<List<AbstractPageBlock>> splitPageBlocksIntoSubSections(List<AbstractPageBlock> pageBlocks) {
List<List<AbstractPageBlock>> splitList = splitIntoCoherentList(pageBlocks);
movePrecedingHeadlineToTableList(splitList);
return splitList.stream().filter(list -> !list.isEmpty()).toList();
}
private void movePrecedingHeadlineToTableList(List<List<AbstractPageBlock>> splitList) {
for (int i = 0; i < splitList.size(); i++) {
if (listIsTablesOnly(splitList.get(i)) && i > 0) {
List<AbstractPageBlock> previousList = splitList.get(i - 1);
AbstractPageBlock lastPageBlockInPreviousList = previousList.get(previousList.size() - 1);
if (lastPageBlockInPreviousList.isHeadline()) {
previousList.remove(previousList.size() - 1);
splitList.get(i).add(0, lastPageBlockInPreviousList);
}
}
}
}
private boolean listIsTablesOnly(List<AbstractPageBlock> abstractPageBlocks) {
return abstractPageBlocks.stream().allMatch(abstractPageBlock -> abstractPageBlock instanceof TablePageBlock);
}
/**
* @param pageBlocks a List of AbstractPageBlocks, which have at least one TablePageBlock and one ClassificationTextBlock
* @return List of Lists of AbstractPageBlocks, which are exclusively of type ClassificationTextBlock or TablePageBlock
*/
private List<List<AbstractPageBlock>> splitIntoCoherentList(List<AbstractPageBlock> pageBlocks) {
List<List<AbstractPageBlock>> splitList = new LinkedList<>();
List<AbstractPageBlock> currentList = new LinkedList<>();
splitList.add(currentList);
Class<? extends AbstractPageBlock> lastPageBlockClass = pageBlocks.get(0).getClass();
for (AbstractPageBlock pageBlock : pageBlocks) {
if (lastPageBlockClass.isInstance(pageBlock)) {
currentList.add(pageBlock);
} else {
currentList = new LinkedList<>();
currentList.add(pageBlock);
splitList.add(currentList);
lastPageBlockClass = pageBlock.getClass();
}
}
return splitList;
}
private List<TextPageBlock> findTextBlocksWithSameClassificationAndAlignsY(TextPageBlock atc, List<AbstractPageBlock> pageBlocks) {
return pageBlocks.stream()
.filter(abstractTextContainer -> !abstractTextContainer.equals(atc))
.filter(abstractTextContainer -> abstractTextContainer.getPage() == atc.getPage())
.filter(abstractTextContainer -> abstractTextContainer instanceof TextPageBlock)
.filter(abstractTextContainer -> abstractTextContainer.intersectsY(atc))
.map(abstractTextContainer -> (TextPageBlock) abstractTextContainer)
.filter(abstractTextContainer -> abstractTextContainer.getDir() == atc.getDir())
.toList();
}
private void addSectionNodeToPageNode(DocumentGraphFactory.Context context, Section section, Integer pageNumber) {
Page page = context.getPage(pageNumber);
page.getMainBody().add(section);
}
}
@@ -0,0 +1,136 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import static java.util.Collections.emptyList;
import java.util.Collection;
import java.util.List;
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.table.Cell;
import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionOperations;
import lombok.experimental.UtilityClass;
@UtilityClass
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) {
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();
Table table = Table.builder().documentTree(context.getDocumentTree()).numberOfCols(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);
ifTableHasNoHeadersSetFirstRowAsHeaders(table);
}
private void setPageNumberInCells(List<TablePageBlock> tablesToMerge) {
// For some reason I can't figure out, in some table cells, the ClassificationTextBlocks have 0 as page number
// So I am fixing this here, but this should actually be fixed upstream.
tablesToMerge.forEach(table -> table.getRows()
.stream()
.flatMap(Collection::stream)
.peek(cell -> cell.setPageNumber(table.getPage()))
.forEach(cell -> setPageNumberInTextBlocksWithPageNumberSetTo0(table, cell)));
}
private void setPageNumberInTextBlocksWithPageNumberSetTo0(TablePageBlock table, Cell cell) {
cell.getTextBlocks().stream()//
.filter(tb -> tb.getPage() == 0)//
.forEach(tb -> tb.setPage(table.getPage()));
}
@SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
private void addTableToPage(Page page, SemanticNode parentNode, Table table) {
if (!page.getMainBody().contains(parentNode)) {
parentNode.getPages().add(page);
}
page.getMainBody().add(table);
}
private void ifTableHasNoHeadersSetFirstRowAsHeaders(Table table) {
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) {
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);
}
}
}
@SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
private void addTableCell(Cell cell, int rowIndex, int colIndex, Table tableNode, DocumentGraphFactory.Context context) {
Page page = context.getPage(cell.getPageNumber());
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);
tableCell.setTreeId(treeId);
TextBlock textBlock;
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);
tableCell.setLeafTextBlock(textBlock);
} else if (firstTextBlockIsHeadline(cell)) {
SectionNodeFactory.addSection(tableCell, cell.getTextBlocks().stream().map(tb -> (AbstractPageBlock) tb).toList(), emptyList(), context);
} 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()));
}
}
private boolean cellAreaIsSmallerThanPageAreaTimesThreshold(Cell cell, Page page) {
return cell.getArea() < TABLE_CELL_MERGE_CONTENTS_SIZE_THRESHOLD * page.getHeight() * page.getWidth();
}
private boolean firstTextBlockIsHeadline(Cell cell) {
return cell.getTextBlocks().get(0).isHeadline();
}
}
@@ -1,79 +1,74 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import java.util.Collections;
import java.util.List;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.AtomicTextBlock;
import lombok.AccessLevel;
import lombok.experimental.FieldDefaults;
@FieldDefaults(level = AccessLevel.PRIVATE)
public class TextBlockFactory {
AtomicInteger stringOffset;
AtomicLong textBlockIdx;
int stringOffset;
long textBlockIdx;
public TextBlockFactory() {
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
stringOffset = new AtomicInteger();
textBlockIdx = new AtomicLong();
Integer numberOnPage = context.getAndIncrementTextBlockNumberOnPage(page);
return buildAtomicTextBlock(sequences, parent, numberOnPage, page);
}
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, DocumentGraphFactory.Context context, PageNode page) {
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences, SemanticNode parent, Integer numberOnPage, Page page) {
Integer numberOnPage = context.pages().get(page).getAndIncrement();
return buildAtomicTextBlock(sequences, parent, context, numberOnPage, page);
SearchTextWithTextPositionDto searchTextWithTextPositionDto = SearchTextWithTextPositionFactory.buildSearchTextToTextPositionDto(sequences);
int offset = stringOffset;
stringOffset += searchTextWithTextPositionDto.getSearchText().length();
long idx = textBlockIdx;
textBlockIdx++;
String orientation;
int textDirection;
if (sequences.isEmpty()) {
orientation = null;
textDirection = 0;
} else {
orientation = sequences.get(0).getDir().toString();
textDirection = sequences.get(0).getRotation();
}
return AtomicTextBlock.fromSearchTextWithTextPosition(searchTextWithTextPositionDto.getSearchText(),
searchTextWithTextPositionDto.getLineBreaks(),
searchTextWithTextPositionDto.getBoldTextBoundaries(),
searchTextWithTextPositionDto.getItalicTextBoundaries(),
searchTextWithTextPositionDto.getPositions(),
searchTextWithTextPositionDto.getStringIdxToPositionIdx(),
idx,
parent,
numberOnPage,
page,
offset,
orientation,
textDirection);
}
public AtomicTextBlock buildAtomicTextBlock(List<TextPositionSequence> sequences,
SemanticNode parent,
DocumentGraphFactory.Context context,
Integer numberOnPage,
PageNode page) {
public AtomicTextBlock emptyTextBlock(SemanticNode parent, DocumentGraphFactory.Context context, Page page) {
SearchTextWithTextPositionModel searchTextWithTextPositionModel = SearchTextWithTextPositionFactory.buildSearchTextToTextPositionModel(sequences);
int offset = stringOffset.getAndAdd(searchTextWithTextPositionModel.getSearchText().length());
return AtomicTextBlock.builder()
.id(textBlockIdx.getAndIncrement())
.parent(parent)
.searchText(searchTextWithTextPositionModel.getSearchText())
.numberOnPage(numberOnPage)
.page(page)
.lineBreaks(searchTextWithTextPositionModel.getLineBreaks())
.positions(searchTextWithTextPositionModel.getPositions())
.stringIdxToPositionIdx(searchTextWithTextPositionModel.getStringCoordsToPositionCoords())
.boundary(new Boundary(offset, offset + searchTextWithTextPositionModel.getSearchText().length()))
.build();
long idx = textBlockIdx;
textBlockIdx++;
return AtomicTextBlock.empty(idx, stringOffset, page, context.getAndIncrementTextBlockNumberOnPage(page), parent);
}
public AtomicTextBlock emptyTextBlock(SemanticNode parent, DocumentGraphFactory.Context context, PageNode page) {
public AtomicTextBlock emptyTextBlock(SemanticNode parent, Integer numberOnPage, Page page) {
return emptyTextBlock(parent, context.pages().get(page).getAndIncrement(), page);
}
public AtomicTextBlock emptyTextBlock(SemanticNode parent, Integer numberOnPage, PageNode page) {
return AtomicTextBlock.builder()
.id(textBlockIdx.getAndIncrement())
.boundary(new Boundary(stringOffset.get(), stringOffset.get()))
.searchText("")
.lineBreaks(Collections.emptyList())
.page(page)
.numberOnPage(numberOnPage)
.stringIdxToPositionIdx(Collections.emptyList())
.positions(Collections.emptyList())
.parent(parent)
.build();
long idx = textBlockIdx;
textBlockIdx++;
return AtomicTextBlock.empty(idx, stringOffset, page, numberOnPage, parent);
}
}
@@ -1,20 +0,0 @@
package com.knecon.fforesight.service.layoutparser.processor.factory;
import java.util.Comparator;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.ClassificationTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.classification.dto.text.TextPositionSequence;
public class TextPositionOperations {
public static List<TextPositionSequence> mergeAndSortTextPositionSequenceByYThenX(List<ClassificationTextBlock> textBlocks) {
return textBlocks.stream()//
.flatMap(tb -> tb.getSequences().stream())//
.sorted(Comparator.comparingDouble(TextPositionSequence::getMaxYDirAdj)//
.thenComparing(TextPositionSequence::getMaxXDirAdj))//
.toList();
}
}
@@ -1,11 +1,17 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph;
package com.knecon.fforesight.service.layoutparser.processor.graph;
import static java.lang.String.format;
import java.util.Collection;
import java.util.LinkedList;
import java.util.List;
import java.util.stream.IntStream;
import lombok.EqualsAndHashCode;
import lombok.Setter;
@Setter
@EqualsAndHashCode
public class Boundary implements Comparable<Boundary> {
private int start;
@@ -15,7 +21,7 @@ public class Boundary implements Comparable<Boundary> {
public Boundary(int start, int end) {
if (start > end) {
throw new IllegalArgumentException(String.format("start: %d > end: %d", start, end));
throw new IllegalArgumentException(format("start: %d > end: %d", start, end));
}
this.start = start;
this.end = end;
@@ -55,7 +61,7 @@ public class Boundary implements Comparable<Boundary> {
public boolean contains(int start, int end) {
if (start > end) {
throw new IllegalArgumentException(String.format("start: %d > end: %d", start, end));
throw new IllegalArgumentException(format("start: %d > end: %d", start, end));
}
return this.start <= start && end <= this.end;
}
@@ -64,7 +70,7 @@ public class Boundary implements Comparable<Boundary> {
public boolean containedBy(int start, int end) {
if (start > end) {
throw new IllegalArgumentException(String.format("start: %d > end: %d", start, end));
throw new IllegalArgumentException(format("start: %d > end: %d", start, end));
}
return start <= this.start && this.end <= end;
}
@@ -78,14 +84,14 @@ public class Boundary implements Comparable<Boundary> {
public boolean intersects(Boundary boundary) {
return contains(boundary.start()) || contains(boundary.end() - 1);
return boundary.start() < this.end && this.start < boundary.end();
}
public List<Boundary> split(List<Integer> splitIndices) {
if (splitIndices.stream().anyMatch(idx -> !this.contains(idx))) {
throw new IndexOutOfBoundsException(String.format("%s splitting indices are out of range for %s", splitIndices.stream().filter(idx -> !this.contains(idx)).toList(), this));
throw new IndexOutOfBoundsException(format("%s splitting indices are out of range for %s", splitIndices.stream().filter(idx -> !this.contains(idx)).toList(), this));
}
List<Boundary> splitBoundaries = new LinkedList<>();
int previousIndex = start;
@@ -102,8 +108,12 @@ public class Boundary implements Comparable<Boundary> {
return splitBoundaries;
}
public IntStream intStream() {
public static Boundary merge(List<Boundary> boundaries) {
return IntStream.range(start, end);
}
public static Boundary merge(Collection<Boundary> boundaries) {
int minStart = boundaries.stream().mapToInt(Boundary::start).min().orElseThrow(IllegalArgumentException::new);
int maxEnd = boundaries.stream().mapToInt(Boundary::end).max().orElseThrow(IllegalArgumentException::new);
@@ -114,7 +124,7 @@ public class Boundary implements Comparable<Boundary> {
@Override
public String toString() {
return String.format("Boundary [%d|%d)", start, end);
return format("Boundary [%d|%d)", start, end);
}
@@ -131,18 +141,4 @@ public class Boundary implements Comparable<Boundary> {
return 0;
}
@Override
public int hashCode() {
return toString().hashCode();
}
@Override
public boolean equals(Object object) {
return hashCode() == object.hashCode();
}
}
@@ -0,0 +1,217 @@
package com.knecon.fforesight.service.layoutparser.processor.graph;
import static java.lang.String.format;
import java.util.Collections;
import java.util.LinkedList;
import java.util.List;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Document;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.GenericSemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Table;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.TableCell;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.Getter;
import lombok.experimental.FieldDefaults;
@Data
@EqualsAndHashCode
public class DocumentTree {
private final Entry root;
public DocumentTree(Document document) {
root = Entry.builder().treeId(Collections.emptyList()).children(new LinkedList<>()).node(document).build();
}
public TextBlock buildTextBlock() {
return allEntriesInOrder().map(Entry::getNode).filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
}
public List<Integer> createNewMainEntryAndReturnId(GenericSemanticNode node) {
return createNewChildEntryAndReturnIdImpl(Collections.emptyList(), node);
}
public List<Integer> createNewChildEntryAndReturnId(GenericSemanticNode parentNode, GenericSemanticNode node) {
return createNewChildEntryAndReturnIdImpl(parentNode.getTreeId(), node);
}
public List<Integer> createNewChildEntryAndReturnId(GenericSemanticNode parentNode, Table node) {
return createNewChildEntryAndReturnIdImpl(parentNode.getTreeId(), node);
}
public List<Integer> createNewTableChildEntryAndReturnId(Table parentTable, TableCell tableCell) {
return createNewChildEntryAndReturnIdImpl(parentTable.getTreeId(), tableCell);
}
@SuppressWarnings("PMD.UnusedPrivateMethod") // PMD actually flags this wrong
private List<Integer> createNewChildEntryAndReturnIdImpl(List<Integer> parentId, SemanticNode node) {
if (!entryExists(parentId)) {
throw new IllegalArgumentException(format("parentId %s does not exist!", parentId));
}
Entry parent = getEntryById(parentId);
List<Integer> newId = new LinkedList<>(parentId);
newId.add(parent.children.size());
parent.children.add(Entry.builder().treeId(newId).node(node).build());
return newId;
}
private boolean entryExists(List<Integer> treeId) {
if (treeId.isEmpty()) {
return root != null;
}
Entry entry = root.children.get(treeId.get(0));
for (int id : treeId.subList(1, treeId.size())) {
if (id >= entry.children.size() || 0 > id) {
return false;
}
entry = entry.children.get(id);
}
return true;
}
public Entry getParentEntryById(List<Integer> treeId) {
return getEntryById(getParentId(treeId));
}
public boolean hasParentById(List<Integer> treeId) {
return !treeId.isEmpty();
}
public Stream<SemanticNode> childNodes(List<Integer> treeId) {
return getEntryById(treeId).children.stream().map(Entry::getNode);
}
public Stream<SemanticNode> childNodesOfType(List<Integer> treeId, NodeType nodeType) {
return getEntryById(treeId).children.stream().filter(entry -> entry.node.getType().equals(nodeType)).map(Entry::getNode);
}
private static List<Integer> getParentId(List<Integer> treeId) {
if (treeId.isEmpty()) {
throw new UnsupportedOperationException("Root has no parent!");
}
if (treeId.size() < 2) {
return Collections.emptyList();
}
return treeId.subList(0, treeId.size() - 1);
}
public Entry getEntryById(List<Integer> treeId) {
if (treeId.isEmpty()) {
return root;
}
Entry entry = root.children.get(treeId.get(0));
for (int id : treeId.subList(1, treeId.size())) {
entry = entry.children.get(id);
}
return entry;
}
public Stream<Entry> mainEntries() {
return root.children.stream();
}
public Stream<Entry> allEntriesInOrder() {
return Stream.of(root).flatMap(DocumentTree::flatten);
}
public Stream<Entry> allSubEntriesInOrder(List<Integer> parentId) {
return getEntryById(parentId).children.stream().flatMap(DocumentTree::flatten);
}
@Override
public String toString() {
return String.join("\n", allEntriesInOrder().map(Entry::toString).toList());
}
private static Stream<Entry> flatten(Entry entry) {
return Stream.concat(Stream.of(entry), entry.children.stream().flatMap(DocumentTree::flatten));
}
public SemanticNode getHighestParentById(List<Integer> treeId) {
if (treeId.isEmpty()) {
return root.node;
}
return root.children.get(treeId.get(0)).node;
}
@Builder
@Getter
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE, makeFinal = true)
public static class Entry {
List<Integer> treeId;
SemanticNode node;
@Builder.Default
List<Entry> children = new LinkedList<>();
@Override
public String toString() {
return node.toString();
}
public NodeType getType() {
return node.getType();
}
}
}
@@ -0,0 +1,8 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.entity;
public enum EntityType {
ENTITY,
RECOMMENDATION,
FALSE_POSITIVE,
FALSE_RECOMMENDATION
}
@@ -0,0 +1,228 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.entity;
import java.awt.geom.Rectangle2D;
import java.util.Collection;
import java.util.Comparator;
import java.util.Deque;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import java.util.Set;
import com.iqser.red.service.persistence.service.v1.api.shared.model.redactionlog.Engine;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.utils.IdBuilder;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
@EqualsAndHashCode(onlyExplicitlyIncluded = true)
public class RedactionEntity {
// initial values
@EqualsAndHashCode.Include
final Boundary boundary;
@EqualsAndHashCode.Include
final String type;
@EqualsAndHashCode.Include
final EntityType entityType;
// empty defaults
boolean redaction;
boolean removed;
boolean ignored;
boolean resized;
boolean skipRemoveEntitiesContainedInLarger;
boolean dictionaryEntry;
boolean dossierDictionaryEntry;
Set<Engine> engines;
Set<RedactionEntity> references;
@Builder.Default
Deque<Integer> matchedRules = new LinkedList<>();
String redactionReason;
String legalBasis;
// inferred on graph insertion
@EqualsAndHashCode.Include
String value;
String textBefore;
String textAfter;
@Builder.Default
Set<Page> pages = new HashSet<>();
List<RedactionPosition> redactionPositionsPerPage;
@Builder.Default
List<SemanticNode> intersectingNodes = new LinkedList<>();
SemanticNode deepestFullyContainingNode;
public static RedactionEntity initialEntityNode(Boundary boundary, String type, EntityType entityType) {
return RedactionEntity.builder().type(type).entityType(entityType).boundary(boundary).engines(new HashSet<>()).references(new HashSet<>()).build();
}
public boolean occursInNodeOfType(Class<? extends SemanticNode> clazz) {
return intersectingNodes.stream().anyMatch(clazz::isInstance);
}
public boolean occursInNode(SemanticNode semanticNode) {
return intersectingNodes.stream().anyMatch(node -> node.equals(semanticNode));
}
public boolean isType(String type) {
return this.type.equals(type);
}
public boolean isAnyType(List<String> types) {
return types.contains(type);
}
public void addIntersectingNode(SemanticNode containingNode) {
intersectingNodes.add(containingNode);
}
public void removeFromGraph() {
intersectingNodes.forEach(node -> node.getEntities().remove(this));
pages.forEach(page -> page.getEntities().remove(this));
intersectingNodes = new LinkedList<>();
deepestFullyContainingNode = null;
pages = new HashSet<>();
removed = true;
ignored = true;
}
public void addMatchedRule(int ruleNumber) {
matchedRules.add(ruleNumber);
}
public int getMatchedRule() {
if (matchedRules.isEmpty()) {
return 0;
}
return matchedRules.getLast();
}
public List<RedactionPosition> getRedactionPositionsPerPage() {
if (redactionPositionsPerPage == null || redactionPositionsPerPage.isEmpty()) {
Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = deepestFullyContainingNode.getTextBlock().getPositionsPerPage(boundary);
Page firstPage = rectanglesPerLinePerPage.keySet()
.stream()
.min(Comparator.comparingInt(Page::getNumber))
.orElseThrow(() -> new RuntimeException("No Positions found on any page!"));
String id = IdBuilder.buildId(pages, rectanglesPerLinePerPage.values().stream().flatMap(Collection::stream).toList());
redactionPositionsPerPage = rectanglesPerLinePerPage.entrySet().stream().map(entry -> buildRedactionPosition(firstPage, id, entry)).toList();
}
return redactionPositionsPerPage;
}
private static RedactionPosition buildRedactionPosition(Page firstPage, String id, Map.Entry<Page, List<Rectangle2D>> entry) {
if (entry.getKey().equals(firstPage)) {
return new RedactionPosition(id, entry.getKey(), entry.getValue());
} else {
return new RedactionPosition(id + "-" + entry.getKey().getNumber(), entry.getKey(), entry.getValue());
}
}
public boolean containedBy(RedactionEntity redactionEntity) {
return this.boundary.containedBy(redactionEntity.getBoundary());
}
public boolean contains(RedactionEntity redactionEntity) {
return this.boundary.contains(redactionEntity.getBoundary());
}
public boolean intersects(RedactionEntity redactionEntity) {
return this.boundary.intersects(redactionEntity.getBoundary());
}
public void addEngine(Engine engine) {
engines.add(engine);
}
public void addEngines(Set<Engine> engines) {
this.engines.addAll(engines);
}
public void addReference(RedactionEntity reference) {
references.add(reference);
}
public void addReferences(List<RedactionEntity> references) {
this.references.addAll(references);
}
public boolean matchesAnnotationId(String manualRedactionId) {
return getRedactionPositionsPerPage().stream().anyMatch(entityPosition -> entityPosition.getId().equals(manualRedactionId));
}
@Override
public String toString() {
StringBuilder sb = new StringBuilder();
sb.append("Entity[\"");
sb.append(value);
sb.append("\", ");
sb.append(boundary);
sb.append(", pages[");
pages.forEach(page -> {
sb.append(page.getNumber());
sb.append(", ");
});
sb.delete(sb.length() - 2, sb.length());
sb.append("], type = \"");
sb.append(type);
sb.append("\", EntityType.");
sb.append(entityType);
sb.append("]");
return sb.toString();
}
}
@@ -0,0 +1,24 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.entity;
import java.awt.geom.Rectangle2D;
import java.util.List;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Data;
import lombok.experimental.FieldDefaults;
@Data
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class RedactionPosition {
final String id;
Page page;
// Each entry in this list corresponds to an entry in the redaction log, this means:
// An entity might be represented by multiple redaction log entries
List<Rectangle2D> rectanglePerLine;
}
@@ -0,0 +1,120 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.amazonaws.services.kms.model.NotFoundException;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Document implements GenericSemanticNode {
Set<Page> pages;
DocumentTree documentTree;
Integer numberOfPages;
TextBlock textBlock;
@Builder.Default
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.DOCUMENT;
}
public TextBlock getTextBlock() {
if (textBlock == null) {
textBlock = streamTerminalTextBlocksInOrder().collect(new TextBlockCollector());
}
return textBlock;
}
public List<Section> getMainSections() {
return streamChildrenOfType(NodeType.SECTION).map(node -> (Section) node).collect(Collectors.toList());
}
public Stream<TextBlock> streamTerminalTextBlocksInOrder() {
return streamAllNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock);
}
@Override
public List<Integer> getTreeId() {
return Collections.emptyList();
}
@Override
public void setTreeId(List<Integer> tocId) {
throw new UnsupportedOperationException("Document is always the root of the TablePageBlock of Contents");
}
@Override
public Headline getHeadline() {
return streamAllSubNodesOfType(NodeType.HEADLINE).map(node -> (Headline) node).findFirst().orElseThrow(() -> new NotFoundException("No Headlines found in this document!"));
}
private Stream<SemanticNode> streamAllNodes() {
return documentTree.allEntriesInOrder().map(DocumentTree.Entry::getNode);
}
public Stream<Image> streamAllImages() {
return streamAllSubNodesOfType(NodeType.IMAGE).map(node -> (Image) node);
}
@Override
public String toString() {
return NodeType.DOCUMENT + ": " + this.getTextBlock().buildSummary();
}
@Override
public Map<Page, Rectangle2D> getBBox() {
Map<Page, Rectangle2D> bBox = new HashMap<>();
for (Page page : pages) {
bBox.put(page, new Rectangle2D.Double(0, 0, page.getWidth(), page.getHeight()));
}
return bBox;
}
}
@@ -0,0 +1,65 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Footer implements GenericSemanticNode {
List<Integer> treeId;
TextBlock leafTextBlock;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.FOOTER;
}
@Override
public boolean isLeaf() {
return true;
}
@Override
public TextBlock getTextBlock() {
return leafTextBlock;
}
@Override
public String toString() {
return treeId + ": " + NodeType.FOOTER + ": " + leafTextBlock.buildSummary();
}
}
@@ -0,0 +1,5 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
public interface GenericSemanticNode extends SemanticNode {
}
@@ -0,0 +1,65 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Header implements GenericSemanticNode {
List<Integer> treeId;
TextBlock leafTextBlock;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public boolean isLeaf() {
return true;
}
@Override
public NodeType getType() {
return NodeType.HEADER;
}
@Override
public TextBlock getTextBlock() {
return leafTextBlock;
}
@Override
public String toString() {
return treeId + ": " + NodeType.HEADER + ": " + leafTextBlock.buildSummary();
}
}
@@ -0,0 +1,72 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Headline implements GenericSemanticNode {
List<Integer> treeId;
TextBlock leafTextBlock;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.HEADLINE;
}
@Override
public boolean isLeaf() {
return true;
}
@Override
public TextBlock getTextBlock() {
return leafTextBlock;
}
@Override
public String toString() {
return treeId + ": " + NodeType.HEADLINE + ": " + leafTextBlock.buildSummary();
}
@Override
public Headline getHeadline() {
return this;
}
}
@@ -0,0 +1,95 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.awt.geom.Rectangle2D;
import java.util.Collections;
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.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.NoArgsConstructor;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@NoArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Image implements GenericSemanticNode {
List<Integer> treeId;
String id;
ImageType imageType;
boolean transparent;
Rectangle2D position;
boolean redaction;
boolean ignored;
@Builder.Default
String redactionReason = "";
@Builder.Default
String legalBasis = "";
@Builder.Default
int matchedRule = -1;
@EqualsAndHashCode.Exclude
Page page;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.IMAGE;
}
@Override
public TextBlock getTextBlock() {
return streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
}
@Override
public Set<Page> getPages() {
return Collections.singleton(page);
}
@Override
public String toString() {
return treeId + ": " + NodeType.IMAGE + ": " + imageType.toString() + " " + position;
}
@Override
public Map<Page, Rectangle2D> getBBox() {
Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
bBoxPerPage.put(page, position);
return bBoxPerPage;
}
}
@@ -0,0 +1,21 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
public enum ImageType {
LOGO,
FORMULA,
SIGNATURE,
OTHER,
OCR;
public static ImageType fromString(String imageType) {
return switch (imageType.toLowerCase()) {
case "logo" -> ImageType.LOGO;
case "formula" -> ImageType.FORMULA;
case "signature" -> ImageType.SIGNATURE;
case "ocr" -> ImageType.OCR;
default -> ImageType.OTHER;
};
}
}
@@ -0,0 +1,87 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.util.HashSet;
import java.util.LinkedList;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.EqualsAndHashCode;
import lombok.Getter;
import lombok.Setter;
import lombok.experimental.FieldDefaults;
@Getter
@Setter
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Page {
Integer number;
Integer height;
Integer width;
Integer rotation;
@EqualsAndHashCode.Exclude
List<SemanticNode> mainBody;
@EqualsAndHashCode.Exclude
Header header;
@EqualsAndHashCode.Exclude
Footer footer;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Builder.Default
@EqualsAndHashCode.Exclude
Set<Image> images = new HashSet<>();
public static Page fromClassificationPage(ClassificationPage classificationPage) {
return Page.builder()
.height((int) classificationPage.getPageHeight())
.width((int) classificationPage.getPageWidth())
.number(classificationPage.getPageNumber())
.rotation(classificationPage.getRotation())
.mainBody(new LinkedList<>())
.build();
}
public TextBlock getMainBodyTextBlock() {
return mainBody.stream().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
}
@Override
public String toString() {
return String.valueOf(number);
}
@Override
public int hashCode() {
return number;
}
@Override
public boolean equals(Object o) {
return o instanceof Page && o.hashCode() == this.hashCode();
}
}
@@ -0,0 +1,63 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Paragraph implements GenericSemanticNode {
List<Integer> treeId;
TextBlock leafTextBlock;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.PARAGRAPH;
}
@Override
public boolean isLeaf() {
return true;
}
@Override
public TextBlock getTextBlock() {
return leafTextBlock;
}
@Override
public String toString() {
return treeId + ": " + NodeType.PARAGRAPH + ": " + leafTextBlock.buildSummary();
}
}
@@ -0,0 +1,77 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
import lombok.extern.slf4j.Slf4j;
@Slf4j
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Section implements GenericSemanticNode {
List<Integer> treeId;
TextBlock textBlock;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public NodeType getType() {
return NodeType.SECTION;
}
public boolean hasTables() {
return streamAllSubNodesOfType(NodeType.TABLE).findAny().isPresent();
}
@Override
public TextBlock getTextBlock() {
if (textBlock == null) {
textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
}
return textBlock;
}
@Override
public String toString() {
return treeId.toString() + ": " + NodeType.SECTION + ": " + this.getTextBlock().buildSummary();
}
public Headline getHeadline() {
return streamChildrenOfType(NodeType.HEADLINE)//
.map(node -> (Headline) node)//
.findFirst()//
.orElseGet(() -> getParent().getHeadline());
}
}
@@ -0,0 +1,452 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.Comparator;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.EntityType;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.AtomicTextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
public interface SemanticNode {
/**
* Returns the type of this node, such as Section, Paragraph, etc.
*
* @return NodeType of this node
*/
NodeType getType();
/**
* Searches all Nodes located underneath this Node in the DocumentTree and concatenates their AtomicTextBlocks into a single TextBlock.
* So, for a Section all TextBlocks of Subsections, Paragraphs, and Tables are concatenated into a single TextBlock
* If the Node is a Leaf, the LeafTextBlock will be returned instead.
*
* @return TextBlock containing all AtomicTextBlocks that are located under this Node.
*/
TextBlock getTextBlock();
/**
* Any Node maintains its own Set of Entities.
* This Set contains all Entities whose boundary intersects the boundary of this node.
*
* @return Set of all Entities associated with this Node
*/
Set<RedactionEntity> getEntities();
/**
* Each AtomicTextBlock is assigned a page, so to get the pages this node appears on, it collects the PageNodes from each AtomicTextBlock belonging to this node's TextBlock.
*
* @return Set of PageNodes this node appears on.
*/
default Set<Page> getPages() {
return getTextBlock().getPages();
}
default Page getFirstPage() {
return getTextBlock().getPages().stream().min(Comparator.comparingInt(Page::getNumber)).orElseThrow(() -> new IllegalStateException("SemanticNode has no Page!"));
}
/**
* Each AtomicTextBlock is assigned a page, so to get the pages for this boundary, it collects the PageNodes from each AtomicTextBlock belonging to this node's TextBlock.
*
* @return Set of PageNodes this node appears on.
*/
default Set<Page> getPages(Boundary boundary) {
if (!getBoundary().contains(boundary)) {
throw new IllegalArgumentException(format("%s which was used to query for pages is not contained in the %s of this node!", boundary, getBoundary()));
}
return getTextBlock().getPages(boundary);
}
default boolean isOnPage(int pageNumber) {
return getPages().stream().anyMatch(page -> page.getNumber() == pageNumber);
}
/**
* Returns the DocumentTree Object.
*
* @return the DocumentTree of the Document this node belongs to
*/
DocumentTree getDocumentTree();
/**
* The id is a List of Integers uniquely identifying this node in the DocumentTree.
*
* @return the DocumentTree ID
*/
List<Integer> getTreeId();
/**
* This should only be used during graph construction.
*
* @param tocId List of Integers
*/
void setTreeId(List<Integer> tocId);
/**
* Traverses the Tree up, until it hits a Headline or hits a Section which will then return the first Headline from its children.
* Throws NotFoundException if no Headline is found this way
*
* @return First Headline found
*/
default Headline getHeadline() {
return getParent().getHeadline();
}
/**
* Checks if its TocId has a length greater than zero.
*
* @return boolean indicating whether this Node has a Parent in the DocumentTree
*/
default boolean hasParent() {
return getDocumentTree().hasParentById(getTreeId());
}
/**
* @return The SemanticNode representing the Parent in the DocumentTree
* throws NotFoundException, when no parent is present
*/
default SemanticNode getParent() {
return getDocumentTree().getParentEntryById(getTreeId()).getNode();
}
/**
* @return The SemanticNode which is directly underneath the document and also under which this node is.
* if this is the highest child node or the document itself, it returns itself.
*/
default SemanticNode getHighestParent() {
return getDocumentTree().getHighestParentById(getTreeId());
}
/**
* Leaf means a SemanticNode has direct access to a TextBlock, by default this is false and must be overridden.
* Currently only Sections, Images, and Tables are not leaves.
* A TableCell might be a leaf depending on its area compared to the page.
*
* @return boolean, indicating if a Node has direct access to a TextBlock
*/
default boolean isLeaf() {
return false;
}
/**
* Leaf means a SemanticNode has direct access to a TextBlock, by default this is false and must be overridden.
* Currently only Sections and Tables are no leaves.
*
* @return AtomicTextBlock
*/
default TextBlock getLeafTextBlock() {
throw new UnsupportedOperationException("Only leaf Nodes have access to LeafTextBlocks!");
}
/**
* Should only be used during construction of the Graph. Sets the LeafTextBlock of this SemanticNode.
*
* @param textBlock the TextBlock to set as the LeafTextBlock of this SemanticNode
*/
default void setLeafTextBlock(TextBlock textBlock) {
throw new UnsupportedOperationException();
}
/**
* Checks whether this SemanticNode has any Entity with EntityType.ENTITY of the provided type.
*
* @param type string representing the type of entity to check for
* @return true, if this SemanticNode has at least one Entity of the provided type
*/
default boolean hasEntitiesOfType(String type) {
return getEntities().stream().filter(entity -> entity.getEntityType().equals(EntityType.ENTITY)).anyMatch(redactionEntity -> redactionEntity.getType().equals(type));
}
/**
* Returns a List of Entities in this SemanticNode which are of the provided type such as "CBI_author".
*
* @param type string representing the type of entities to return
* @return List of RedactionEntities of any the type
*/
default List<RedactionEntity> getEntitiesOfType(String type) {
return getEntities().stream().filter(redactionEntity -> redactionEntity.getType().equals(type)).toList();
}
/**
* Returns a List of Entities in this SemanticNode which have any of the provided types such as "CBI_author".
*
* @param types A list of strings representing the types of entities to return
* @return List of RedactionEntities of any provided type
*/
default List<RedactionEntity> getEntitiesOfType(List<String> types) {
return getEntities().stream().filter(redactionEntity -> redactionEntity.isAnyType(types)).toList();
}
/**
* Each AtomicTextBlock has an index on its page, this returns the number of the first AtomicTextBlock underneath this node.
* If this node does not have any AtomicTexBlocks underneath it, e.g. an empty TableCell. It returns -1.
*
* @return Integer representing the number on the page
*/
default Integer getNumberOnPage() {
TextBlock textBlock = getTextBlock();
if (!textBlock.getAtomicTextBlocks().isEmpty()) {
return getTextBlock().getAtomicTextBlocks().get(0).getNumberOnPage();
} else {
return -1;
}
}
/**
* Checks if the SemanticNode contains any text.
*
* @return true, if this node's TextBlock is not empty
*/
default boolean hasText() {
return !getTextBlock().isEmpty();
}
/**
* Checks whether this SemanticNode contains the provided String.
*
* @param string A String which the TextBlock might contain
* @return true, if this node's TextBlock contains the string
*/
default boolean containsString(String string) {
return getTextBlock().getSearchText().contains(string);
}
/**
* Checks whether this SemanticNode contains all the provided Strings.
*
* @param strings A List of Strings which the TextBlock might contain
* @return true, if this node's TextBlock contains all strings
*/
default boolean containsStrings(List<String> strings) {
return strings.stream().allMatch(this::containsString);
}
/**
* Checks whether this SemanticNode contains all the provided Strings ignoring case.
*
* @param string A String which the TextBlock might contain
* @return true, if this node's TextBlock contains the string ignoring case
*/
default boolean containsStringIgnoreCase(String string) {
return getTextBlock().getSearchText().toLowerCase().contains(string.toLowerCase());
}
/**
* Checks whether this SemanticNode contains any of the provided Strings.
*
* @param strings A List of Strings which the TextBlock might contain
* @return true, if this node's TextBlock contains any of the strings
*/
default boolean containsAnyString(List<String> strings) {
return strings.stream().anyMatch(this::containsString);
}
/**
* Checks whether this SemanticNode contains any of the provided Strings ignoring case.
*
* @param strings A List of Strings which the TextBlock might contain
* @return true, if this node's TextBlock contains any of the strings
*/
default boolean containsAnyStringIgnoreCase(List<String> strings) {
return strings.stream().anyMatch(this::containsStringIgnoreCase);
}
/**
* This function is used during insertion of EntityNodes into the graph, it checks if the boundary of the RedactionEntity intersects or even contains the RedactionEntity.
* It sets the fields accordingly and recursively calls this function on all its children.
*
* @param redactionEntity RedactionEntity, which is being inserted into the graph
*/
default void addThisToEntityIfIntersects(RedactionEntity redactionEntity) {
TextBlock textBlock = getTextBlock();
if (textBlock.getBoundary().intersects(redactionEntity.getBoundary())) {
if (textBlock.containsBoundary(redactionEntity.getBoundary())) {
redactionEntity.setDeepestFullyContainingNode(this);
}
redactionEntity.addIntersectingNode(this);
streamChildren().filter(semanticNode -> semanticNode.getBoundary().intersects(redactionEntity.getBoundary()))
.forEach(node -> node.addThisToEntityIfIntersects(redactionEntity));
}
}
/**
* Streams all children located directly underneath this node in the DocumentTree.
*
* @return Stream of all children
*/
default Stream<SemanticNode> streamChildren() {
return getDocumentTree().childNodes(getTreeId());
}
/**
* Streams all children located directly underneath this node in the DocumentTree of the provided type.
*
* @return Stream of all children
*/
default Stream<SemanticNode> streamChildrenOfType(NodeType nodeType) {
return getDocumentTree().childNodesOfType(getTreeId(), nodeType);
}
/**
* Recursively streams all SemanticNodes located underneath this node in the DocumentTree in order.
*
* @return Stream of all SubNodes
*/
default Stream<SemanticNode> streamAllSubNodes() {
return getDocumentTree().allSubEntriesInOrder(getTreeId()).map(DocumentTree.Entry::getNode);
}
/**
* Recursively streams all SemanticNodes of the provided type located underneath this node in the DocumentTree in order.
*
* @return Stream of all SubNodes
*/
default Stream<SemanticNode> streamAllSubNodesOfType(NodeType nodeType) {
return getDocumentTree().allSubEntriesInOrder(getTreeId()).filter(entry -> entry.getType().equals(nodeType)).map(DocumentTree.Entry::getNode);
}
/**
* The Boundary is the start and end string offsets in the reading order of the document.
*
* @return Boundary of this Node's TextBlock
*/
default Boundary getBoundary() {
return getTextBlock().getBoundary();
}
/**
* If this Node is a Leaf it will calculate the boundingBox of its LeafTextBlock, otherwise it will calculate the Union of the BoundingBoxes of all its Children.
* If called on the Document, it will return the cropbox of each page
*
* @return Rectangle2D fully encapsulating this Node for each page.
*/
default Map<Page, Rectangle2D> getBBox() {
Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
if (isLeaf()) {
return getBBoxFromLeafTextBlock(bBoxPerPage);
}
return getBBoxFromChildren(bBoxPerPage);
}
/**
* Checks whether the Bounding Box of this SemanticNode contains the provided rectangle on the provided page.
*
* @param rectangle2D The rectangle to check if it is contained
* @param pageNumber The Page number on which the rectangle should be checked
* @return boolean
*/
default boolean containsRectangle(Rectangle2D rectangle2D, Integer pageNumber) {
Page helperPage = Page.builder().number(pageNumber).build();
if (!getPages().contains(helperPage)) {
return false;
}
return getBBox().get(helperPage).contains(rectangle2D);
}
/**
* TODO: this produces unwanted results for sections spanning multiple columns.
*
* @param bBoxPerPage initial empty BoundingBox
* @return The union of the BoundingBoxes of all children
*/
private Map<Page, Rectangle2D> getBBoxFromChildren(Map<Page, Rectangle2D> bBoxPerPage) {
return streamChildren().map(SemanticNode::getBBox).reduce((map1, map2) -> {
map1.forEach((page, rectangle) -> map2.merge(page, rectangle, (rect1, rect2) -> rect1.createUnion(rect2).getBounds2D()));
return map2;
}).orElse(bBoxPerPage);
}
/**
* @param bBoxPerPage initial empty BoundingBox
* @return The union of all BoundingBoxes of the TextBlock of this node
*/
private Map<Page, Rectangle2D> getBBoxFromLeafTextBlock(Map<Page, Rectangle2D> bBoxPerPage) {
Map<Page, List<AtomicTextBlock>> atomicTextBlockPerPage = getTextBlock().getAtomicTextBlocks().stream().collect(Collectors.groupingBy(AtomicTextBlock::getPage));
atomicTextBlockPerPage.forEach((page, atbs) -> bBoxPerPage.put(page, RectangleTransformations.bBoxUnionAtomicTextBlock(atbs)));
return bBoxPerPage;
}
}
@@ -0,0 +1,316 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import static java.lang.String.format;
import java.util.Collection;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import java.util.stream.IntStream;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class Table implements SemanticNode {
List<Integer> treeId;
DocumentTree documentTree;
int numberOfRows;
int numberOfCols;
TextBlock textBlock;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
/**
* Streams all entities in this table, that appear in a row, which contains any of the provided strings.
*
* @param strings Strings to check whether a row contains them
* @return Stream of all entities in this table, that appear in a row, which contains any of the provided strings
*/
public Stream<RedactionEntity> streamEntitiesWhereRowContainsStringsIgnoreCase(List<String> strings) {
return IntStream.range(0, numberOfRows)
.boxed()
.filter(row -> rowContainsStringsIgnoreCase(row, strings))
.flatMap(this::streamRow)
.map(TableCell::getEntities)
.flatMap(Collection::stream);
}
/**
* Checks whether the specified row contains all the provided strings.
*
* @param row the row to check as an Integer, must be smaller than numberOfRows
* @param strings a list of strings to check for
* @return true, if all strings appear in the provided row
*/
public boolean rowContainsStringsIgnoreCase(Integer row, List<String> strings) {
String rowText = streamRow(row).map(TableCell::getTextBlock).collect(new TextBlockCollector()).getSearchText().toLowerCase();
return strings.stream().map(String::toLowerCase).allMatch(rowText::contains);
}
/**
* Streams all entities which appear in a row where at least one cell has the provided header and the provided value.
*
* @param header the header value to search for
* @param value the string which the table cell should contain
* @return a stream of all entities, which appear in a row where at least one cell has the provided header and the provided value.
*/
public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndValue(String header, String value) {
List<Integer> vertebrateStudyCols = streamHeaders().filter(headerNode -> headerNode.containsString(header)).map(TableCell::getCol).toList();
return streamTableCells().filter(tableCellNode -> vertebrateStudyCols.stream()
.anyMatch(vertebrateStudyCol -> getCell(tableCellNode.getRow(), vertebrateStudyCol).containsString(value))).map(TableCell::getEntities).flatMap(Collection::stream);
}
/**
* Streams all entities which appear in a row where at least one cell has the provided header and any provided value.
*
* @param header the header value to search for
* @param values the strings which the table cell should contain
* @return a stream of all entities, which appear in a row where at least one cell has the provided header and any provided value.
*/
public Stream<RedactionEntity> streamEntitiesWhereRowHasHeaderAndAnyValue(String header, List<String> values) {
List<Integer> colsWithHeader = streamHeaders().filter(headerNode -> headerNode.containsString(header)).map(TableCell::getCol).toList();
return streamTableCells().filter(tableCellNode -> colsWithHeader.stream()
.anyMatch(colWithHeader -> getCell(tableCellNode.getRow(), colWithHeader).containsAnyString(values))).map(TableCell::getEntities).flatMap(Collection::stream);
}
/**
* Streams all entities in this table, that appear in a row, which contains at least one entity with any of the provided types.
*
* @param types type strings to check whether a row contains an entity like them
* @return Stream of all entities in this table, that appear in a row, which contains at least one entity with any of the provided types.
*/
public Stream<RedactionEntity> streamEntitiesWhereRowContainsEntitiesOfType(List<String> types) {
List<Integer> rowsWithEntityOfType = IntStream.range(0, numberOfRows)
.boxed()
.filter(rowNumber -> streamEntityTypesInRow(rowNumber).anyMatch(existingType -> types.stream().anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
.toList();
return rowsWithEntityOfType.stream().flatMap(this::streamRow).map(TableCell::getEntities).flatMap(Collection::stream);
}
/**
* Streams all entities in this table, that appear in a row, which does not contain any entity with any of the provided types.
*
* @param types type strings to check whether a row doesn't contain an entity like it
* @return Stream of all entities in this table, that appear in a row, which does not contain any entity with any of the provided types.
*/
public Stream<RedactionEntity> streamEntitiesWhereRowContainsNoEntitiesOfType(List<String> types) {
List<Integer> rowsWithNoEntityOfType = IntStream.range(0, numberOfRows)
.boxed()
.filter(rowNumber -> streamEntityTypesInRow(rowNumber).noneMatch(existingType -> types.stream().anyMatch(typeToCheck -> typeToCheck.equals(existingType))))
.toList();
return rowsWithNoEntityOfType.stream().flatMap(this::streamRow).map(TableCell::getEntities).flatMap(Collection::stream);
}
private Stream<String> streamEntityTypesInRow(Integer rowNumber) {
return streamRow(rowNumber).map(TableCell::getEntities).flatMap(Collection::stream).map(RedactionEntity::getType).distinct();
}
/**
* Returns a TableCell at the provided row and column location.
*
* @param row int representing the row, must be smaller than numberOfRows
* @param col int representing the col, must be smaller than numberOfCols
* @return TableCell at the provided location in the table
*/
public TableCell getCell(int row, int col) {
if (numberOfRows - row < 0 || numberOfCols - col < 0) {
throw new IllegalArgumentException(format("row %d, col %d is out of bounds for number of rows of %d and number of cols %d", row, col, numberOfRows, numberOfCols));
}
int idx = row * numberOfCols + col;
return (TableCell) documentTree.getEntryById(treeId).getChildren().get(idx).getNode();
}
/**
* Streams all TableCells in this Table row-wise.
*
* @return Stream of all TableCells
*/
public Stream<TableCell> streamTableCells() {
return streamChildrenOfType(NodeType.TABLE_CELL).map(node -> (TableCell) node);
}
/**
* Streams all TableCells in this Table which have the provided header row-wise.
*
* @return Stream of all TableCells which have the provided header
*/
public Stream<TableCell> streamTableCellsWithHeader(String header) {
return streamHeaders().filter(tableCellNode -> tableCellNode.getTextBlock().getSearchText().contains(header))
.map(TableCell::getCol)
.flatMap(this::streamCol)
.filter(tableCellNode -> !tableCellNode.isHeader());
}
/**
* Streams all TableCells belonging to the provided column from top down.
*
* @param col int representing the column
* @return Stream of all TableCell in the provided column
*/
public Stream<TableCell> streamCol(int col) {
return IntStream.range(0, numberOfRows).boxed().map(row -> getCell(row, col));
}
/**
* Streams all TableCells belonging to the provided row from left to right.
*
* @param row int representing the row
* @return Stream of all TableCell in the provided row
*/
public Stream<TableCell> streamRow(int row) {
return IntStream.range(0, numberOfCols).boxed().map(col -> getCell(row, col));
}
/**
* Streams all TableCells row-wise and filters them with header == true.
*
* @return Stream of all TableCells with header == true
*/
public Stream<TableCell> streamHeaders() {
return streamTableCells().filter(TableCell::isHeader);
}
/**
* Streams all TableCells of the provided row and column and filters them with header == true.
*
* @param row int representing the row
* @param col int representing the column
* @return Stream of all TableCells with header == true in the provided row or col
*/
public Stream<TableCell> streamHeadersForCell(int row, int col) {
return Stream.concat(streamRow(row), streamCol(col)).filter(TableCell::isHeader);
}
/**
* Streams all Headers and checks if any equal the provided string.
*
* @param header string to check the headers for
* @return true, if at least one header equals the provided string
*/
public boolean hasHeader(String header) {
return streamHeaders().anyMatch(tableCellNode -> tableCellNode.getTextBlock().getSearchText().strip().equals(header));
}
/**
* Checks if this table has a column with the provided header and any of the table cells in that column contain the provided value.
*
* @param header string to find header cells
* @param value string to check cells with provided header
* @return true, if this table has a column with the provided header and any of the table cells in that column contain the provided value
*/
public boolean hasRowWithHeaderAndValue(String header, String value) {
return streamTableCellsWithHeader(header).anyMatch(tableCellNode -> tableCellNode.containsString(value));
}
/**
* Checks if this table has a column with the provided header and any of the table cells in that column contains any of the provided values.
*
* @param header string to find header cells
* @param values List of strings to check cells with provided header
* @return true, if this table has a column with the provided header and any of the table cells in that column contains any of the provided values.
*/
public boolean hasRowWithHeaderAndAnyValue(String header, List<String> values) {
return streamTableCellsWithHeader(header).anyMatch(tableCellNode -> tableCellNode.containsAnyString(values));
}
/**
* Finds all entities of the provided type, which appear in the same row that the provided entity appears in.
*
* @param type the type of entities to search for
* @param redactionEntity the entity, which appears in the row to search
* @return List of all entities of the provided type, which appear in the same row that the provided entity appears in.
*/
public List<RedactionEntity> getEntitiesOfTypeInSameRow(String type, RedactionEntity redactionEntity) {
return redactionEntity.getIntersectingNodes()
.stream()
.filter(node -> node instanceof TableCell)
.map(node -> (TableCell) node)
.flatMap(tableCellNode -> streamRow(tableCellNode.getRow()))
.map(cell -> cell.getEntitiesOfType(type))
.flatMap(Collection::stream)
.toList();
}
@Override
public NodeType getType() {
return NodeType.TABLE;
}
@Override
public TextBlock getTextBlock() {
if (textBlock == null) {
textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
}
return textBlock;
}
@Override
public String toString() {
return treeId.toString() + ": " + NodeType.TABLE + ": #cols: " + numberOfCols + ", #rows: " + numberOfRows + ", " + this.getTextBlock().buildSummary();
}
}
@@ -0,0 +1,91 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.nodes;
import java.awt.geom.Rectangle2D;
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.internal.api.data.redaction.NodeType;
import com.knecon.fforesight.service.layoutparser.processor.graph.DocumentTree;
import com.knecon.fforesight.service.layoutparser.processor.graph.entity.RedactionEntity;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlock;
import com.knecon.fforesight.service.layoutparser.processor.graph.textblock.TextBlockCollector;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class TableCell implements GenericSemanticNode {
List<Integer> treeId;
int row;
int col;
boolean header;
Rectangle2D bBox;
TextBlock leafTextBlock;
TextBlock textBlock;
@EqualsAndHashCode.Exclude
DocumentTree documentTree;
@Builder.Default
@EqualsAndHashCode.Exclude
Set<RedactionEntity> entities = new HashSet<>();
@Override
public Map<Page, Rectangle2D> getBBox() {
Map<Page, Rectangle2D> bBoxPerPage = new HashMap<>();
getPages().forEach(page -> bBoxPerPage.put(page, bBox));
return bBoxPerPage;
}
@Override
public NodeType getType() {
return NodeType.TABLE_CELL;
}
@Override
public boolean isLeaf() {
return getDocumentTree().getEntryById(getTreeId()).getChildren().isEmpty();
}
@Override
public TextBlock getTextBlock() {
if (isLeaf()) {
return leafTextBlock;
}
if (textBlock == null) {
textBlock = streamAllSubNodes().filter(SemanticNode::isLeaf).map(SemanticNode::getLeafTextBlock).collect(new TextBlockCollector());
}
return textBlock;
}
@Override
public String toString() {
return treeId + ": " + NodeType.TABLE_CELL + ": " + this.getTextBlock().buildSummary();
}
}
@@ -0,0 +1,232 @@
package com.knecon.fforesight.service.layoutparser.processor.graph.textblock;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentPositionData;
import com.knecon.fforesight.service.layoutparser.internal.api.data.redaction.DocumentTextData;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.SemanticNode;
import com.knecon.fforesight.service.layoutparser.processor.utils.RectangleTransformations;
import lombok.AccessLevel;
import lombok.AllArgsConstructor;
import lombok.Builder;
import lombok.Data;
import lombok.EqualsAndHashCode;
import lombok.experimental.FieldDefaults;
@Data
@Builder
@AllArgsConstructor
@FieldDefaults(level = AccessLevel.PRIVATE)
public class AtomicTextBlock implements TextBlock {
Long id;
Integer numberOnPage;
Page page;
//string coordinates
Boundary boundary;
String searchText;
@Builder.Default
List<Integer> lineBreaks = new ArrayList<>();
@Builder.Default
List<Boundary> boldTextBoundaries = new ArrayList<>();
@Builder.Default
List<Boundary> italicTextBoundaries = new ArrayList<>();
String orientation;
int textDirection;
//position coordinates
@Builder.Default
List<Integer> stringIdxToPositionIdx = new ArrayList<>();
@Builder.Default
List<Rectangle2D> positions = new ArrayList<>();
@EqualsAndHashCode.Exclude
SemanticNode parent;
@Override
public int numberOfLines() {
return lineBreaks.size() + 1;
}
public static AtomicTextBlock fromSearchTextWithTextPosition(String searchText,
List<Integer> lineBreaks,
List<Boundary> boldTextBoundaries,
List<Boundary> italicTextBoundaries,
List<Rectangle2D> positions,
List<Integer> stringIdxToPositionIdx,
long idx,
SemanticNode parent,
int numberOnPage,
Page page,
int offset,
String orientation,
int textDirection) {
return AtomicTextBlock.builder()
.id(idx)
.parent(parent)
.searchText(searchText)
.numberOnPage(numberOnPage)
.page(page)
.lineBreaks(lineBreaks)
.boldTextBoundaries(boldTextBoundaries)
.italicTextBoundaries(italicTextBoundaries)
.positions(positions)
.stringIdxToPositionIdx(stringIdxToPositionIdx)
.boundary(new Boundary(offset, offset + searchText.length()))
.textDirection(textDirection)
.orientation(orientation)
.build();
}
public static AtomicTextBlock empty(Long textBlockIdx, int stringOffset, Page page, int numberOnPage, SemanticNode parent) {
return AtomicTextBlock.builder()
.id(textBlockIdx)
.boundary(new Boundary(stringOffset, stringOffset))
.searchText("")
.page(page)
.numberOnPage(numberOnPage)
.parent(parent)
.build();
}
public static AtomicTextBlock fromAtomicTextBlockData(DocumentTextData documentTextData,
DocumentPositionData documentPositionData,
SemanticNode parent,
Page page) {
return AtomicTextBlock.builder()
.id(documentTextData.getId())
.numberOnPage(documentTextData.getNumberOnPage())
.page(page)
.boundary(new Boundary(documentTextData.getStart(), documentTextData.getEnd()))
.searchText(documentTextData.getSearchText())
.lineBreaks(Arrays.stream(documentTextData.getLineBreaks()).boxed().toList())
.stringIdxToPositionIdx(Arrays.stream(documentPositionData.getStringIdxToPositionIdx()).boxed().toList())
.positions(toRectangle2DList(documentPositionData.getPositions()))
.parent(parent)
.build();
}
private static List<Rectangle2D> toRectangle2DList(float[][] positions) {
return Arrays.stream(positions).map(floatArr -> (Rectangle2D) new Rectangle2D.Float(floatArr[0], floatArr[1], floatArr[2], floatArr[3])).toList();
}
public CharSequence getLine(int lineNumber) {
if (lineNumber >= numberOfLines() || lineNumber < 0) {
throw new IndexOutOfBoundsException(format("line %d out of range for AtomicTextBlock with %d lines", lineNumber, numberOfLines()));
}
if (lineNumber == 0) {
return subSequence(boundary.start(), lineBreaks.get(0) + boundary.start());
} else if (lineNumber == numberOfLines() - 1) {
return subSequence(lineBreaks.get(lineBreaks.size() - 1) + boundary.start(), boundary.end());
}
return subSequence(lineBreaks.get(lineNumber - 1) + boundary.start(), lineBreaks.get(lineNumber) + boundary.start());
}
@Override
public List<AtomicTextBlock> getAtomicTextBlocks() {
return List.of(this);
}
@Override
public int getNextLinebreak(int fromIndex) {
return lineBreaks.stream()//
.filter(linebreak -> linebreak > fromIndex - boundary.start()) //
.findFirst() //
.orElse(searchText.length()) + boundary.start();
}
@Override
public int getPreviousLinebreak(int fromIndex) {
return lineBreaks.stream()//
.filter(linebreak -> linebreak <= fromIndex - boundary.start())//
.reduce((a, b) -> b)//
.orElse(0) + boundary.start();
}
@Override
public Rectangle2D getPosition(int stringIdx) {
return positions.get(stringIdxToPositionIdx.get(stringIdx - boundary.start()));
}
@Override
public List<Rectangle2D> getPositions(Boundary stringBoundary) {
if (!containsBoundary(stringBoundary)) {
throw new IndexOutOfBoundsException(format("%s is out of bounds for %s", stringBoundary, this.boundary));
}
if (stringBoundary.length() == 0) {
return Collections.emptyList();
}
int startPositionIdx = stringIdxToPositionIdx.get(stringBoundary.start() - this.boundary.start());
if (stringBoundary.end() == this.boundary.end()) {
return positions.subList(startPositionIdx, positions.size());
}
return positions.subList(startPositionIdx, stringIdxToPositionIdx.get(stringBoundary.end() - this.boundary.start()));
}
public Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary) {
List<Rectangle2D> rectanglesPerLine = stringBoundary.split(getAllLineBreaksInBoundary(stringBoundary))
.stream()
.map(this::getPositions)
.map(RectangleTransformations::rectangleBBoxWithGaps)
.flatMap(Collection::stream)
.toList();
Map<Page, List<Rectangle2D>> rectanglePerLinePerPage = new HashMap<>();
rectanglePerLinePerPage.put(page, rectanglesPerLine);
return rectanglePerLinePerPage;
}
private List<Integer> getAllLineBreaksInBoundary(Boundary boundary) {
return getLineBreaks().stream().map(linebreak -> linebreak + this.boundary.start()).filter(boundary::contains).toList();
}
@Override
public String toString() {
return searchText;
}
}
@@ -1,14 +1,18 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock;
package com.knecon.fforesight.service.layoutparser.processor.graph.textblock;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityPosition;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
import lombok.AccessLevel;
import lombok.Data;
@@ -23,6 +27,12 @@ public class ConcatenatedTextBlock implements TextBlock {
Boundary boundary;
public static ConcatenatedTextBlock empty() {
return new ConcatenatedTextBlock(Collections.emptyList());
}
public ConcatenatedTextBlock(List<AtomicTextBlock> atomicTextBlocks) {
this.atomicTextBlocks = new LinkedList<>();
@@ -44,10 +54,11 @@ public class ConcatenatedTextBlock implements TextBlock {
boundary.setStart(textBlock.getBoundary().start());
boundary.setEnd(textBlock.getBoundary().end());
} else if (boundary.end() != textBlock.getBoundary().start()) {
throw new UnsupportedOperationException(String.format("Can only concat consecutive TextBlocks, trying to concat %s and %s", boundary, textBlock.getBoundary()));
throw new UnsupportedOperationException(format("Can only concat consecutive TextBlocks, trying to concat %s and %s", boundary, textBlock.getBoundary()));
}
this.atomicTextBlocks.addAll(textBlock.getAtomicTextBlocks());
boundary.setEnd(textBlock.getBoundary().end());
this.searchText = null;
return this;
}
@@ -135,38 +146,34 @@ public class ConcatenatedTextBlock implements TextBlock {
@Override
public List<EntityPosition> getEntityPositionsPerPage(Boundary stringBoundary) {
public Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary) {
List<AtomicTextBlock> textBlocks = getAllAtomicTextBlocksPartiallyInStringBoundary(stringBoundary);
if (textBlocks.size() == 1) {
return textBlocks.get(0).getEntityPositionsPerPage(stringBoundary);
return textBlocks.get(0).getPositionsPerPage(stringBoundary);
}
AtomicTextBlock firstTextBlock = textBlocks.get(0);
List<EntityPosition> positions = new LinkedList<>(firstTextBlock.getEntityPositionsPerPage(new Boundary(stringBoundary.start(), firstTextBlock.getBoundary().end())));
Map<Page, List<Rectangle2D>> rectanglesPerLinePerPage = firstTextBlock.getPositionsPerPage(new Boundary(stringBoundary.start(), firstTextBlock.getBoundary().end()));
for (AtomicTextBlock textBlock : textBlocks.subList(1, textBlocks.size() - 1)) {
positions.addAll(textBlock.getEntityPositionsPerPage(textBlock.getBoundary()));
rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage, textBlock.getPositionsPerPage(textBlock.getBoundary()));
}
AtomicTextBlock lastTextBlock = textBlocks.get(textBlocks.size() - 1);
positions.addAll(lastTextBlock.getEntityPositionsPerPage(new Boundary(lastTextBlock.getBoundary().start(), stringBoundary.end())));
rectanglesPerLinePerPage = mergeEntityPositionsWithSamePageNode(rectanglesPerLinePerPage,
lastTextBlock.getPositionsPerPage(new Boundary(lastTextBlock.getBoundary().start(), stringBoundary.end())));
return mergeEntityPositionsWithSamePageNode(positions);
return rectanglesPerLinePerPage;
}
private List<EntityPosition> mergeEntityPositionsWithSamePageNode(List<EntityPosition> positions) {
Map<PageNode, List<Rectangle2D>> entityPositionsPerPage = positions.stream().collect(//
Collectors.groupingBy(EntityPosition::getPageNode, //
Collectors.flatMapping(entityPosition -> entityPosition.getRectanglePerLine().stream(), Collectors.toList())));
return entityPositionsPerPage.entrySet().stream()//
.map(entry -> EntityPosition.builder().pageNode(entry.getKey()).rectanglePerLine(entry.getValue()).build())//
.toList();
private Map<Page, List<Rectangle2D>> mergeEntityPositionsWithSamePageNode(Map<Page, List<Rectangle2D>> map1, Map<Page, List<Rectangle2D>> map2) {
Map<Page, List<Rectangle2D>> mergedMap = new HashMap<>(map1);
map2.forEach((pageNode, rectangles) -> mergedMap.merge(pageNode, rectangles, (l1, l2) -> Stream.concat(l1.stream(), l2.stream()).toList()));
return mergedMap;
}
@@ -176,4 +183,38 @@ public class ConcatenatedTextBlock implements TextBlock {
return getSearchText();
}
@Override
public List<Boundary> getBoldTextBoundaries() {
return getAtomicTextBlocks().stream().map(AtomicTextBlock::getBoldTextBoundaries).flatMap(Collection::stream).toList();
}
@Override
public List<Boundary> getItalicTextBoundaries() {
return getAtomicTextBlocks().stream().map(AtomicTextBlock::getItalicTextBoundaries).flatMap(Collection::stream).toList();
}
@Override
public String getOrientation() {
if (atomicTextBlocks.isEmpty()) {
return "";
}
return atomicTextBlocks.get(0).getOrientation();
}
@Override
public int getTextDirection() {
if (atomicTextBlocks.isEmpty()) {
return 0;
}
return atomicTextBlocks.get(0).getTextDirection();
}
}
@@ -1,15 +1,17 @@
package com.knecon.fforesight.service.layoutparser.internal.api.graph.textblock;
package com.knecon.fforesight.service.layoutparser.processor.graph.textblock;
import static java.lang.String.format;
import java.awt.geom.Rectangle2D;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.stream.Collectors;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.entity.EntityPosition;
import com.knecon.fforesight.service.layoutparser.internal.api.graph.nodes.PageNode;
import com.knecon.fforesight.service.layoutparser.processor.graph.Boundary;
import com.knecon.fforesight.service.layoutparser.processor.graph.nodes.Page;
public interface TextBlock extends CharSequence {
@@ -19,6 +21,18 @@ public interface TextBlock extends CharSequence {
List<AtomicTextBlock> getAtomicTextBlocks();
List<Boundary> getBoldTextBoundaries();
List<Boundary> getItalicTextBoundaries();
String getOrientation();
int getTextDirection();
Boundary getBoundary();
@@ -37,7 +51,7 @@ public interface TextBlock extends CharSequence {
List<Rectangle2D> getPositions(Boundary stringBoundary);
List<EntityPosition> getEntityPositionsPerPage(Boundary stringBoundary);
Map<Page, List<Rectangle2D>> getPositionsPerPage(Boundary stringBoundary);
int numberOfLines();
@@ -49,12 +63,21 @@ public interface TextBlock extends CharSequence {
}
default Set<PageNode> getPages() {
default Set<Page> getPages() {
return getAtomicTextBlocks().stream().map(AtomicTextBlock::getPage).collect(Collectors.toUnmodifiableSet());
}
default Set<Page> getPages(Boundary boundary) {
return getAtomicTextBlocks().stream()
.filter(atomicTextBlock -> atomicTextBlock.getBoundary().intersects(boundary))
.map(AtomicTextBlock::getPage)
.collect(Collectors.toUnmodifiableSet());
}
default int indexOf(String searchTerm, int startOffset) {
int start = getSearchText().indexOf(searchTerm, startOffset - getBoundary().start());
@@ -74,7 +97,7 @@ public interface TextBlock extends CharSequence {
default boolean containsBoundary(Boundary boundary) {
if (boundary.end() < boundary.start()) {
throw new IllegalArgumentException(String.format("Invalid %s, StartIndex must be smaller than EndIndex", boundary));
throw new IllegalArgumentException(format("Invalid %s, StartIndex must be smaller than EndIndex", boundary));
}
return getBoundary().contains(boundary);
}

Some files were not shown because too many files have changed in this diff Show More