/** Serializes the given color. */
protected String serializeColor(PdfColor color) {
if (color == null) {
return null;
}
String id = color.getId();
if (id == null) {
return null;
}
List<String> fields = new ArrayList<>();
// Add feature name.
fields.add("color");
// Add id.
fields.add(id);
// Add rgb
fields.add(String.valueOf(color.getR()));
fields.add(String.valueOf(color.getG()));
fields.add(String.valueOf(color.getB()));
return CollectionUtils.join(fields, "\t");
}
/** Serializes the given (non-text) element. */
protected String serializeNonTextElement(PdfElement element) {
if (element == null) {
return null;
}
List<String> fields = new ArrayList<>();
// Add feature name.
PdfFeature feature = element.getFeature();
fields.add(feature != null ? feature.getFieldName() : "");
// Add bounding box.
Rectangle rect = element.getRectangle();
fields.add(rect != null ? String.valueOf(rect.getMinX()) : "");
fields.add(rect != null ? String.valueOf(rect.getMinY()) : "");
fields.add(rect != null ? String.valueOf(rect.getMaxX()) : "");
fields.add(rect != null ? String.valueOf(rect.getMaxY()) : "");
// Add color.
PdfColor color = element.getColor();
fields.add(color != null ? color.getId() : "");
// Add role.
PdfRole role = element.getRole();
fields.add(role != null ? role.name : "");
return CollectionUtils.join(fields, "\t");
}
/** Returns true, if the given character should be considered on extraction. */
public static boolean considerPdfCharacter(PdfCharacter character) {
// Don't consider the character, if it is a diacritic.
if (character.isDiacritic()) {
return false;
}
PdfColor color = character.getColor();
// Don't consider the character, if its color is white.
if (color != null && color.isWhite(0.05f)) {
return false;
}
// Don't consider the character "|" because it is used as PARA_ADDENDUM in
// tex-paragraph-parser.
if ("|".equals(character.getUnicode().trim()) && character.getFontsize() < 6.5f) { // TODO
return false;
}
// Don't consider the character, if its orientation isn't 0.
if (character.getOrientation() != 0) {
return false;
}
// Don't consider the character, if its a proper rectangle.
if (character.getRectangle().getWidth() <= 0 || character.getRectangle().getHeight() <= 0) {
return false;
}
// If the character has a valid encoding, we have to consider it.
if (character.hasEncoding()) {
return true;
}
// Don't consider the character, if its doesn't hold any content.
if (character.getUnicode().trim().isEmpty()) {
return false;
}
return true;
}
/** Returns true, if the given shape should be considered on extraction. */
public static boolean considerPdfShape(PdfShape shape, Rectangle clippingPath) {
if (shape == null) {
return false;
}
PdfColor color = shape.getColor();
if (color == null) {
return false;
}
if (color.isWhite()) {
return false;
}
// If the clipping path is given, it must contain the shape.
if (clippingPath != null) {
return clippingPath.overlaps(shape.getRectangle());
}
return true;
}
/** Serializes the given text element. */
protected String serializeTextElement(PdfTextElement element) {
if (element == null) {
return null;
}
String text =
element.getText(
getSerializePunctuationMarks(), getSerializeSubscripts(), getSerializeSuperscripts());
if (text == null || text.trim().isEmpty()) {
return null;
}
List<String> fields = new ArrayList<>();
// Add feature name.
PdfFeature feature = element.getFeature();
fields.add(feature != null ? feature.getFieldName() : "");
// Add text.
fields.add(text.replaceAll("\t", " "));
// Add page.
PdfPage page = element.getPage();
fields.add(page != null ? String.valueOf(page.getPageNumber()) : "");
// Add bounding box.
Rectangle rect = element.getRectangle();
fields.add(rect != null ? String.valueOf(rect.getMinX()) : "");
fields.add(rect != null ? String.valueOf(rect.getMinY()) : "");
fields.add(rect != null ? String.valueOf(rect.getMaxX()) : "");
fields.add(rect != null ? String.valueOf(rect.getMaxY()) : "");
// Add most common font.
PdfFont font = element.getFont();
fields.add(font != null ? font.getId() : "");
fields.add(String.valueOf(element.getFontsize()));
// Add most common color.
PdfColor color = element.getColor();
fields.add(color != null ? color.getId() : "");
// Add font of first and last character.
PdfFont firstCharacterFont = null;
PdfFont lastCharacterFont = null;
PdfColor firstCharacterColor = null;
PdfColor lastCharacterColor = null;
float firstCharacterFontsize = 0;
float lastCharacterFontsize = 0;
List<PdfCharacter> characters = element.getTextCharacters();
if (characters != null && !characters.isEmpty()) {
Collections.sort(characters, new Comparators.MinXComparator());
PdfCharacter firstCharacter = characters.get(0);
PdfCharacter lastCharacter = characters.get(characters.size() - 1);
if (firstCharacter != null) {
firstCharacterFont = firstCharacter.getFont();
firstCharacterFontsize = firstCharacter.getFontsize();
firstCharacterColor = firstCharacter.getColor();
}
if (lastCharacter != null) {
lastCharacterFont = lastCharacter.getFont();
lastCharacterFontsize = lastCharacter.getFontsize();
lastCharacterColor = lastCharacter.getColor();
}
}
// Append font and color of first character.
fields.add(firstCharacterFont != null ? firstCharacterFont.getId() : "");
fields.add(String.valueOf(firstCharacterFontsize));
fields.add(firstCharacterColor != null ? firstCharacterColor.getId() : "");
// Append font and color of last character.
fields.add(lastCharacterFont != null ? lastCharacterFont.getId() : "");
fields.add(String.valueOf(lastCharacterFontsize));
fields.add(lastCharacterColor != null ? lastCharacterColor.getId() : "");
// Add role.
PdfRole role = element.getRole();
fields.add(role != null ? role.name : "");
return CollectionUtils.join(fields, "\t");
}