source: gs3-extensions/atlas-src/trunk/src/org/greenstone/atlas/server/HTMLParser.java@ 23934

package org.greenstone.atlas.server;

import java.io.BufferedWriter;
import java.io.FileWriter;
import java.util.ArrayList;

public class HTMLParser
{
    protected char[] _html = null;
    protected int _index = 0;
    
    ArrayList<String> _fullWordList = new ArrayList<String>();
    ArrayList<String> _fullBetweenList = new ArrayList<String>();
    
    public HTMLParser(String html)
    {       
        _html = html.toCharArray();
        
        String other = getNextCharactersBetweenWords();
        String word = getNextWord();
        while(word != null || other != null)
        {
            if(word != null)
            {
                _fullWordList.add(word);
            }
            
            if(other != null)
            {
                _fullBetweenList.add(other);
            }
            
            other = getNextCharactersBetweenWords();
            word = getNextWord();
        }
    }
    
    public String getNextCharactersBetweenWords()
    {
        if(_index >= _html.length)
        {
            return null;
        }
        
        StringBuilder chars = new StringBuilder();
        
        while(true)
        {
            if(_index >= _html.length)
            {
                break;
            }
            
            if(_html[_index] == '<')
            {
                chars.append(_html[_index]);
                int bracketCount = 1;
                while(_index < _html.length && bracketCount > 0)
                {
                    _index++;
                    if(_html[_index] == '>')
                    {
                        bracketCount--;
                    }
                    else if(_html[_index] == '<')
                    {
                        bracketCount++;
                    }
                    chars.append(_html[_index]);    
                }
            }
            else if(!Character.isLetter(_html[_index]))
            {
                chars.append(_html[_index]);
            }
            else
            {
                break;
            }
            _index++;
        }
        
        if(chars.length() == 0)
        {
            return null;
        }
        return chars.toString();
    }
    
    public String getNextWord()
    {
        if(_index >= _html.length)
        {
            return null;
        }
        
        StringBuilder newWord = new StringBuilder();
        
        while(true)
        {
            char currentCharacter = _html[_index];
            if(!Character.isLetter(currentCharacter))
            {
                break;
            }
            else
            {
                newWord.append(currentCharacter);
            }
            _index++;
        }
        return newWord.toString();
    }
    
    public ArrayList<String> getFullHTMLWordList()
    {
        return _fullWordList;
    }
    
    public ArrayList<String> getFullBetweenWordList()
    {
        return _fullBetweenList;
    }
    
    public static String removeTags(String htmlString)
    {
        StringBuilder cleanText = new StringBuilder();
        
        int inTag = 0;
        for(char c : htmlString.toCharArray())
        {
            if(c == '<')
            {
                inTag++;
            }
            else if(c == '>' && inTag > 0)
            {
                inTag--;
            }
            else if(inTag == 0)
            {
                cleanText.append(c);
            }
        }
        
        return cleanText.toString();
    }
}