Compare commits

...
4 Commits
Author SHA1 Message Date
Dominique Eifländer 1ca02f72c8 Merge branch 'RED-8627-4.0' into 'release/0.89.x'
RED-8627: Fixed scrambled text after sorting

See merge request fforesight/layout-parser!121
2024-03-19 11:26:49 +01:00
Dominique Eifländer 350513a699 RED-8627: Fixed scrambled text after sorting 2024-03-19 11:16:07 +01:00
Dominique Eifländer ab7b2cf0d5 Merge branch 'RED-7384' into 'release/0.89.x'
RED-7384: Fixes for ClassCastException

See merge request fforesight/layout-parser!110
2024-03-08 12:53:14 +01:00
Kilian Schuettler 007cbfd1ee RED-7384: Fixes for ClassCastException
* changed save -> incrementalSave
* always use origin file instead of reusing viewerdoc
* Sometimes the viewer document is corrupted after saving and missing the contentstreams on a random page, for the files we viewed it did not seem to happen with incrementalSave.might only be a timing issue though
2024-03-08 12:42:40 +01:00
5 changed files with 32 additions and 25 deletions
@@ -94,8 +94,10 @@ public class LayoutParsingPipeline {
log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());
File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId())
.orElse(originFile);
// File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId())
// .orElse(originFile);
File viewerDocumentFile = originFile;
ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
if (layoutParsingRequest.imagesFileStorageId()
@@ -1,7 +1,6 @@
package com.knecon.fforesight.service.layoutparser.processor.model.text;
import org.apache.pdfbox.text.TextPosition;
import org.springframework.beans.BeanUtils;
import com.fasterxml.jackson.annotation.JsonIgnore;
@@ -50,7 +49,13 @@ public class RedTextPosition {
public static RedTextPosition fromTextPosition(TextPosition textPosition) {
var pos = new RedTextPosition();
BeanUtils.copyProperties(textPosition, pos);
pos.setRotation(textPosition.getRotation());
pos.setPageHeight(textPosition.getPageHeight());
pos.setPageWidth(textPosition.getPageWidth());
pos.setUnicode(textPosition.getUnicode());
pos.setDir(textPosition.getDir());
pos.setWidthOfSpace(textPosition.getWidthOfSpace());
pos.setFontSizeInPt(textPosition.getFontSizeInPt());
pos.setFontName(textPosition.getFont().getName());
pos.setFontSizeInPt(textPosition.getFontSizeInPt());
@@ -190,6 +190,12 @@ public class TextPositionSequence implements CharSequence {
}
public float getTextHeightNoPadding() {
return textPositions.get(0).getHeightDir();
}
public float getTextHeight() {
return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
@@ -28,15 +28,13 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPosit
*
* @author Ben Litchfield
*/
public class TextPositionSequenceComparator implements Comparator<TextPositionSequence>
{
public class TextPositionSequenceComparator implements Comparator<TextPositionSequence> {
@Override
public int compare(TextPositionSequence pos1, TextPositionSequence pos2)
{
public int compare(TextPositionSequence pos1, TextPositionSequence pos2) {
// only compare text that is in the same direction
int cmp1 = Float.compare(pos1.getDir().getDegrees(), pos2.getDir().getDegrees());
if (cmp1 != 0)
{
if (cmp1 != 0) {
return cmp1;
}
@@ -48,25 +46,19 @@ public class TextPositionSequenceComparator implements Comparator<TextPositionSe
float pos2YBottom = pos2.getMaxYDirAdj();
// note that the coordinates have been adjusted so 0,0 is in upper left
float pos1YTop = pos1YBottom - pos1.getTextHeight();
float pos2YTop = pos2YBottom - pos2.getTextHeight();
float pos1YTop = pos1YBottom - pos1.getTextHeightNoPadding();
float pos2YTop = pos2YBottom - pos2.getTextHeightNoPadding();
float yDifference = Math.abs(pos1YBottom - pos2YBottom);
// we will do a simple tolerance comparison
if (yDifference < .1 ||
pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom ||
pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom)
{
if (yDifference < .1 || pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom || pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom) {
return Float.compare(x1, x2);
}
else if (pos1YBottom < pos2YBottom)
{
} else if (pos1YBottom < pos2YBottom) {
return -1;
}
else
{
} else {
return 1;
}
}
}
@@ -126,8 +126,8 @@ public class ViewerDocumentService {
pdDocument = openPDDocument(tmpFile.toFile());
}
}
observedIncrementalSave(pdDocument, destinationFile);
observedIncrementalSave(pdDocument, destinationFile);
pdDocument.close();
assert tmpFile.toFile().delete();
}
@@ -282,10 +282,12 @@ public class ViewerDocumentService {
@SneakyThrows
private void observedIncrementalSave(PDDocument pdDocument, File outputFile) {
/*
Sometimes the viewer document is corrupted after saving and missing the content streams on a random page, for the files we viewed it did not seem to happen with incrementalSave. It might only be a timing issue though
*/
Observation.createNotStarted("ViewerDocumentService", registry).contextualName("incremental-save").observe(() -> {
try (var out = new FileOutputStream(outputFile)) {
pdDocument.save(out);
pdDocument.saveIncremental(out);
} catch (IOException e) {
throw new RuntimeException(e);
}