
org.jpedal.io.PageLabels Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of OpenViewerFX Show documentation
Show all versions of OpenViewerFX Show documentation
An Open Source JavaFX PDF Viewer
/*
* ===========================================
* Java Pdf Extraction Decoding Access Library
* ===========================================
*
* Project Info: http://www.idrsolutions.com
* Help section for developers at http://www.idrsolutions.com/support/
*
* (C) Copyright 1997-2016 IDRsolutions and Contributors.
*
* This file is part of JPedal/JPDF2HTML5
*
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* ---------------
* PageLabels.java
* ---------------
*/
package org.jpedal.io;
import java.util.HashMap;
import org.jpedal.objects.raw.PageLabelObject;
import org.jpedal.objects.raw.PdfArrayIterator;
import org.jpedal.objects.raw.PdfDictionary;
import org.jpedal.objects.raw.PdfObject;
/**
* convert names to refs
*/
public class PageLabels extends HashMap{
final PdfFileReader objectReader;
final int pageCount;
static final String symbolLowerCase[]={"m","cm", "d", "cd", "c", "xc", "l", "xl", "x", "ix", "v", "iv", "i"};
static final String symbolUpperCase[]={"M","CM", "D", "CD", "C", "XC", "L", "XL", "X", "IX", "V", "IV", "I"};
static final String lettersLowerCase[]={"a","b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m","n",
"o", "p", "q", "r", "s", "t", "u", "v", "x", "w", "y", "z"};
static final String lettersUpperCase[]={"A","B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M","N",
"O", "P", "Q", "R", "S", "T", "U", "V", "X", "W", "Y", "Z"};
static final int[] power={1000,900, 500, 400, 100, 90, 50, 40, 10, 9, 5, 4, 1};
PageLabels(PdfFileReader objectReader, int pageCount) {
this.objectReader=objectReader;
this.pageCount=pageCount;
}
void readLabels(PdfObject pageObj) {
PdfArrayIterator numList =pageObj.getMixedArray(PdfDictionary.Nums);
if(numList!=null && objectReader!=null){
int startPage,endPage,numbType,ST,pageNum=1;
String convertedPage,pageLabel;
while(numList.hasMoreTokens()){
//read Page
startPage=numList.getNextValueAsInteger()+1;
//read LabelObject
String key=numList.getNextValueAsString(true);
final PageLabelObject labelObj = new PageLabelObject(key);
objectReader.readObject(labelObj);
numbType=labelObj.getNameAsConstant(PdfDictionary.S);
pageLabel=labelObj.getTextStreamValue(PdfDictionary.P);
ST=labelObj.getInt(PdfDictionary.ST);
if(ST>0){
pageNum=ST;
}else{
pageNum=1;
}
if(numList.hasMoreTokens()){
endPage=numList.getNextValueAsInteger(false)+1;
}else{
endPage=pageCount+1;
}
//now decode type of naming and fill range
for(int page=startPage;page0){
pageLabel=pageLabel.concat(getNumberValue(numbType, ST));
ST++;
}
}else{
convertedPage = getNumberValue(numbType, pageNum);
}
this.put(page, convertedPage);
pageNum++;
}
}
}
}
static String getNumberValue(int numbType, int page) {
String convertedPage;
switch(numbType){
case PdfDictionary.a:
convertedPage=convertLetterToNumber(page, lettersLowerCase);
break;
case PdfDictionary.A:
convertedPage=convertLetterToNumber(page, lettersUpperCase);
break;
case PdfDictionary.D:
convertedPage=String.valueOf(page);
break;
case PdfDictionary.R:
convertedPage=convertToRoman(page,symbolUpperCase);
break;
case PdfDictionary.r:
convertedPage=convertToRoman(page,symbolLowerCase);
break;
default:
convertedPage=String.valueOf(page);
}
return convertedPage;
}
static String convertToRoman(int arabicNumber, final String[] symbols){
final StringBuilder romanNumeral=new StringBuilder();
int repeat;
for(int x=0; arabicNumber>0; x++){
repeat=arabicNumber/power[x];
for(int chars=1; chars<=repeat; chars++){
romanNumeral.append(symbols[x]);
}
arabicNumber %= power[x];
}
return romanNumeral.toString();
}
private static String convertLetterToNumber(int page, final String[] letters) {
final StringBuilder finalLetters=new StringBuilder();
int repeat = page/26;
int remainder = page%26;
if (repeat>0) {
for (int x=0; x
© 2015 - 2025 Weber Informatics LLC | Privacy Policy