多个线程从iText pdf中提取文本

时间:2013-08-06 19:31:37

标签: java multithreading pdf itext text-extraction

我一直在尝试加快从PDF文件中提取文本。我试过了SnowTide的PDFTextStream。它在我的电脑上比iText慢。

我最近的尝试是使用多个线程。

我使用字节数组而不是文件名初始化PdfReader。   这样我就可以通过将所有内容都拉入内存来避免IO延迟。据我所知,iText可以根据需要从文件中获取数据。使用字节数组时,它不应该有这个问题。

我使用线程提取多个页面范围。

然而,这使它变得更慢。使用一个线程与五个线程更快。我不知道为什么会这样。据我所知,瓶颈正在提取页面本身PdfTextExtractor.getTextFromPage(mReader, i)。我定时使用for循环来连接线程的文本,这不会影响整个时间。

我的问题是,为什么它会变慢,我怎样才能让它变得更快?

访问同一个PdfReader是否会造成瓶颈? 我应该使用相同的字节数组初始化多个PdfReader吗? 我的任何研究都错了吗? 如果有人对全新的开源PDF提取器或其他建议有任何建议,那就太棒了!

import java.io.*;

import com.itextpdf.text.pdf.parser.PdfTextExtractor; 
import com.itextpdf.text.pdf.PdfReader;
import java.nio.channels.FileChannel;
import java.nio.MappedByteBuffer;

public class PDFParser
{
    private PdfReader mReader;
    private int numPages;
    private String bodyText;

    public PDFParser(String pdf)
    throws IOException
    {
    /*
      scrapped this
               mReader = new PdfReader(pdf);
      and replaced with the following. 
      From what I can tell Pdfreader with the above constructor 
      probably uses a stream to get the pages one at a time.
      I want to load it all into memory to increase the speed.
    */

        final FileChannel channel = new FileInputStream(pdf).getChannel();
        MappedByteBuffer buffer = channel.map(FileChannel.MapMode.READ_ONLY, 0, channel.size());
        byte[] byteArray = new byte[buffer.remaining()];
        buffer.get(byteArray);
        channel.close();

        mReader = new PdfReader(byteArray);

        numPages = mReader.getNumberOfPages();
        bodyText = "";
    }

    public String parse()
    {
        int numThreads = 5;
        int offset = (numPages / numThreads) - 1;
        int currentFirstPage = 1;
        ParserThread[] threads = new ParserThread[numThreads];
        int forEnd = numThreads - 1;
        for (int i = 0; i < forEnd; i++)
        {
            threads[i] = new ParserThread(mReader, currentFirstPage, currentFirstPage + offset);
            threads[i].start();
            currentFirstPage += offset + 1;
        }
        threads[forEnd] = new ParserThread(mReader, currentFirstPage, numPages);
        threads[forEnd].start();

        try
        {
            long before = System.currentTimeMillis();
            for(int i = 0; i < numThreads; i++)
            threads[i].join();
            System.out.println(System.currentTimeMillis() - before);
        }
        catch(InterruptedException ie){ie.printStackTrace();}
        mReader.close(); 

        StringBuilder sb = new StringBuilder();
        for (int i = 0; i < numThreads;i++)
        {
            sb.append(threads[i].getText());
        }

        bodyText = sb.toString();
        if (bodyText != "")
            return bodyText;
        else
            return "error";
    }

    public static void main(final String args[]) throws IOException
    {
        for (int i = 0; i < args.length; i++) 
        {
            String text = new PDFParser(args[i]).parse();
            //      System.out.println(text);
        }
    }

}

class ParserThread extends Thread
{
    private PdfReader mReader;
    private int mFirstPage;
    private int mLastPage;
    private String text;

    ParserThread(PdfReader reader, int firstPage, int lastPage)
    {
        mReader = reader;
        mFirstPage = firstPage;
        mLastPage = lastPage;
        text = "";
    }

    @Override
    public void run()
    {
        StringBuilder sb = new StringBuilder();
        try
        {
            for (int i = mFirstPage; i <= mLastPage; i++)
            sb.append(PdfTextExtractor.getTextFromPage(mReader, i));
        }
        catch (IOException ioe)
        {
            ioe.printStackTrace();
        }
        text = sb.toString();
   }

   public String getText()
   {
       return text;
   }
}

0 个答案:

没有答案