/*
*
* This file is part of the iText (R) project.
Copyright (c) 1998-2017 iText Group NV
* Authors: Kevin Day, Bruno Lowagie, et al.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License version 3
* as published by the Free Software Foundation with the addition of the
* following permission added to Section 15 as permitted in Section 7(a):
* FOR ANY PART OF THE COVERED WORK IN WHICH THE COPYRIGHT IS OWNED BY
* ITEXT GROUP. ITEXT GROUP DISCLAIMS THE WARRANTY OF NON INFRINGEMENT
* OF THIRD PARTY RIGHTS.
*
* This program 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 Affero General License for more
* details. You should have received a copy of the GNU Affero General License
* along with this program; if not, see http://www.gnu.org/licenses or write to
* the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
* MA, 02110-1301 USA, or download the license from the following URL:
* http://itextpdf.com/terms-of-use/
*
* The interactive user interfaces in modified source and object code versions
* of this program must display Appropriate Legal Notices, as required under
* Section 5 of the GNU Affero General License.
*
* In accordance with Section 7(b) of the GNU Affero General License, a covered
* work must retain the producer line in every PDF that is created or
* manipulated using iText.
*
* You can be released from the requirements of the license by purchasing a
* commercial license. Buying such a license is mandatory as soon as you develop
* commercial activities involving the iText software without disclosing the
* source code of your own applications. These activities include: offering paid
* services to customers as an ASP, serving PDFs on the fly in a web
* application, shipping iText with a closed source product.
*
* For more information, please contact iText Software Corp. at this address:
* sales@itextpdf.com
*/
package com.itextpdf.text.io;
import java.io.IOException;
import java.nio.channels.FileChannel;
import java.util.Iterator;
import java.util.LinkedList;
/**
* A RandomAccessSource that is based on an underlying {@link FileChannel}. The channel is mapped into memory using a paging scheme to allow for efficient reads of very large files.
* As an implementation detail, we use {@link GroupedRandomAccessSource} functionality, but override to make determination of the underlying
* mapped page more efficient - and to close each page as another is opened
* @since 5.3.5
*/
class PagedChannelRandomAccessSource extends GroupedRandomAccessSource implements RandomAccessSource {
// these values were selected based on parametric testing with extracting text content from a 2.3GB file. These settings resulted in the best improvement over
// the single size MRU case (24% speed improvement)
public static final int DEFAULT_TOTAL_BUFSIZE = 1 << 26;
public static final int DEFAULT_MAX_OPEN_BUFFERS = 16;
/**
* The size of each of the buffers to use when mapping files into memory. This must be greater than 0 and less than {@link Integer#MAX_VALUE}
*/
private final int bufferSize;
/**
* The channel this source is based on
*/
private final FileChannel channel;
/**
* Most recently used list used to hold a number of mapped pages open at a time
*/
private final MRU<RandomAccessSource> mru;
/**
* Constructs a new {@link PagedChannelRandomAccessSource} based on the specified FileChannel, with a default buffer configuration.
* The default buffer configuration is currently 2^26 total paged bytes, spread across a maximum of 16 active buffers. This arrangement
* resulted in a 24% speed improvement over the single buffer case in parametric tests extracting text from a 2.3 GB file.
* @param channel the channel to use as the backing store
* @throws IOException if the channel cannot be opened or mapped
*/
public PagedChannelRandomAccessSource(FileChannel channel) throws IOException {
this(channel, DEFAULT_TOTAL_BUFSIZE, DEFAULT_MAX_OPEN_BUFFERS);
}
/**
* Constructs a new {@link PagedChannelRandomAccessSource} based on the specified FileChannel, with a specific buffer size
* @param channel the channel to use as the backing store
* @param bufferSize the size of the buffers to use
* @throws IOException if the channel cannot be opened or mapped
*/
public PagedChannelRandomAccessSource(final FileChannel channel, final int totalBufferSize, final int maxOpenBuffers) throws IOException {
super(buildSources(channel, totalBufferSize/maxOpenBuffers));
this.channel = channel;
this.bufferSize = totalBufferSize/maxOpenBuffers;
this.mru = new MRU<RandomAccessSource>(maxOpenBuffers);
}
/**
* Constructs a set of {@link MappedChannelRandomAccessSource}s for each page (of size bufferSize) of the underlying channel
* @param channel the underlying channel
* @param bufferSize the size of each page (the last page may be shorter)
* @return a list of sources that represent the pages of the channel
* @throws IOException if IO fails for any reason
*/
private static RandomAccessSource[] buildSources(final FileChannel channel, final int bufferSize) throws IOException{
long size = channel.size();
if (size <= 0)
throw new IOException("File size must be greater than zero");
int bufferCount = (int)(size/bufferSize) + (size % bufferSize == 0 ? 0 : 1);
MappedChannelRandomAccessSource[] sources = new MappedChannelRandomAccessSource[bufferCount];
for (int i = 0; i < bufferCount; i++){
long pageOffset = (long)i*bufferSize;
long pageLength = Math.min(size - pageOffset, bufferSize);
sources[i] = new MappedChannelRandomAccessSource(channel, pageOffset, pageLength);
}
return sources;
}
@Override
/**
* {@inheritDoc}
*/
protected int getStartingSourceIndex(long offset) {
return (int) (offset / bufferSize);
}
@Override
/**
* {@inheritDoc}
* For now, close the source that is no longer being used. In the future, we may implement an MRU that allows multiple pages to be opened at a time
*/
protected void sourceReleased(RandomAccessSource source) throws IOException {
RandomAccessSource old = mru.enqueue(source);
if (old != null)
old.close();
}
@Override
/**
* {@inheritDoc}
* Ensure that the source is mapped. In the future, we may implement an MRU that allows multiple pages to be opened at a time
*/
protected void sourceInUse(RandomAccessSource source) throws IOException {
((MappedChannelRandomAccessSource)source).open();
}
@Override
/**
* {@inheritDoc}
* Cleans the mapped bytebuffers and closes the channel
*/
public void close() throws IOException {
super.close();
channel.close();
}
private static class MRU<E>{
/**
* The maximum number of entries held by this MRU
*/
private final int limit;
/**
* Backing list for managing the MRU
*/
private LinkedList<E> queue = new LinkedList<E>();
/**
* Constructs an MRU with the specified size
* @param limit the limit
*/
public MRU(int limit) {
this.limit = limit;
}
/**
* Adds an element to the MRU. If the element is already in the MRU, it is moved to the top.
* @param newElement the element to add
* @return the element that was removed from the MRU to make room for the new element, or null if no element needed to be removed
*/
public E enqueue(E newElement){
// TODO: this check may not be an effective optimization - the GroupedRandomAccessSource already tracks the 'current' source, so it seems unlikely that we would ever hit this code branch
if (queue.size() > 0 && queue.getFirst() == newElement)
return null;
for(Iterator<E> it = queue.iterator(); it.hasNext();){
E element = it.next();
if (newElement == element){
it.remove();
queue.addFirst(newElement);
return null;
}
}
queue.addFirst(newElement);
if (queue.size() > limit)
return queue.removeLast();
return null;
}
}
}