Compress an InputStream with gzip
Here is a version that I wrote that doesn't have the CRC/GZIP Magic cookies in it, because it delegates to a GZIPOutputStream. It is also memory efficient in that it only uses enough memory to buffer the compression (a 42MB file used a 45k buffer). Performance is the same as compressing to memory.
import java.io.IOException;
import java.io.InputStream;
import java.io.OutputStream;
import java.util.zip.GZIPOutputStream;
/**
* Compresses an InputStream in a memory-optimal, on-demand way only compressing enough to fill a buffer.
*
* @author Ben La Monica
*/
public class GZIPCompressingInputStream extends InputStream {
private InputStream in;
private GZIPOutputStream gz;
private OutputStream delegate;
private byte[] buf = new byte[8192];
private byte[] readBuf = new byte[8192];
int read = 0;
int write = 0;
public GZIPCompressingInputStream(InputStream in) throws IOException {
this.in = in;
this.delegate = new OutputStream() {
private void growBufferIfNeeded(int len) {
if ((write + len) >= buf.length) {
// grow the array if we don't have enough space to fulfill the incoming data
byte[] newbuf = new byte[(buf.length + len) * 2];
System.arraycopy(buf, 0, newbuf, 0, buf.length);
buf = newbuf;
}
}
@Override
public void write(byte[] b, int off, int len) throws IOException {
growBufferIfNeeded(len);
System.arraycopy(b, off, buf, write, len);
write += len;
}
@Override
public void write(int b) throws IOException {
growBufferIfNeeded(1);
buf[write++] = (byte) b;
}
};
this.gz = new GZIPOutputStream(delegate);
}
@Override
public int read(byte[] b, int off, int len) throws IOException {
compressStream();
int numBytes = Math.min(len, write-read);
if (numBytes > 0) {
System.arraycopy(buf, read, b, off, numBytes);
read += numBytes;
} else if (len > 0) {
// if bytes were requested, but we have none, then we're at the end of the stream
return -1;
}
return numBytes;
}
private void compressStream() throws IOException {
// if the reader has caught up with the writer, then zero the positions out
if (read == write) {
read = 0;
write = 0;
}
while (write == 0) {
// feed the gzip stream data until it spits out a block
int val = in.read(readBuf);
if (val == -1) {
// nothing left to do, we've hit the end of the stream. finalize and break out
gz.close();
break;
} else if (val > 0) {
gz.write(readBuf, 0, val);
}
}
}
@Override
public int read() throws IOException {
compressStream();
if (write == 0) {
// write should not be 0 if we were able to get data from compress stream, must mean we're at the end
return -1;
} else {
// reading a single byte
return buf[read++] & 0xFF;
}
}
}
DeflaterInputStream is not what you want because it lacks gzip header/trailer and uses a slightly different compression.
If you change from OutputStream (push) to InputStream (pull) you need to do things different.
What GzipOutputStream does is:
- write a static gzip header
- write a deflated stream using DeflaterOutputStream. While the stream is written, a CRC32 checksum is built from the uncompressed data and the number of bytes is count
- write a trailer containing the CRC32 Checksum and the number of bytes.
If you want to do the same with InputStreams, you need a stream that contains:
- the header
- the deflated content
- the trailer
The best way to do this is to provide 3 different streams and combine them to one. Fortunately there is SequenceInputStream that does the combining of the streams for you.
Here's my implementation plus a simple unit test:
import java.io.ByteArrayInputStream;
import java.io.FileInputStream;
import java.io.FilterInputStream;
import java.io.IOException;
import java.io.InputStream;
import java.io.SequenceInputStream;
import java.util.Enumeration;
import java.util.zip.CRC32;
import java.util.zip.Deflater;
import java.util.zip.DeflaterInputStream;
import java.util.zip.DeflaterOutputStream;
/**
* @author mwyraz
* Wraps an input stream and compresses it's contents. Similiar to DeflateInputStream but adds GZIP-header and trailer
* See GzipOutputStream for details.
* LICENSE: Free to use. Contains some lines from GzipOutputStream, so oracle's license might apply as well!
*/
public class GzipCompressingInputStream extends SequenceInputStream
{
public GzipCompressingInputStream(InputStream in) throws IOException
{
this(in,512);
}
public GzipCompressingInputStream(InputStream in, int bufferSize) throws IOException
{
super(new StatefullGzipStreamEnumerator(in,bufferSize));
}
static enum StreamState
{
HEADER,
CONTENT,
TRAILER
}
protected static class StatefullGzipStreamEnumerator implements Enumeration<InputStream>
{
protected final InputStream in;
protected final int bufferSize;
protected StreamState state;
public StatefullGzipStreamEnumerator(InputStream in, int bufferSize)
{
this.in=in;
this.bufferSize=bufferSize;
state=StreamState.HEADER;
}
public boolean hasMoreElements()
{
return state!=null;
}
public InputStream nextElement()
{
switch (state)
{
case HEADER:
state=StreamState.CONTENT;
return createHeaderStream();
case CONTENT:
state=StreamState.TRAILER;
return createContentStream();
case TRAILER:
state=null;
return createTrailerStream();
}
return null;
}
static final int GZIP_MAGIC = 0x8b1f;
static final byte[] GZIP_HEADER=new byte[] {
(byte) GZIP_MAGIC, // Magic number (short)
(byte)(GZIP_MAGIC >> 8), // Magic number (short)
Deflater.DEFLATED, // Compression method (CM)
0, // Flags (FLG)
0, // Modification time MTIME (int)
0, // Modification time MTIME (int)
0, // Modification time MTIME (int)
0, // Modification time MTIME (int)
0, // Extra flags (XFLG)
0 // Operating system (OS)
};
protected InputStream createHeaderStream()
{
return new ByteArrayInputStream(GZIP_HEADER);
}
protected InternalGzipCompressingInputStream contentStream;
protected InputStream createContentStream()
{
contentStream=new InternalGzipCompressingInputStream(new CRC32InputStream(in), bufferSize);
return contentStream;
}
protected InputStream createTrailerStream()
{
return new ByteArrayInputStream(contentStream.createTrailer());
}
}
/**
* Internal stream without header/trailer
*/
protected static class CRC32InputStream extends FilterInputStream
{
protected CRC32 crc = new CRC32();
protected long byteCount;
public CRC32InputStream(InputStream in)
{
super(in);
}
@Override
public int read() throws IOException
{
int val=super.read();
if (val>=0)
{
crc.update(val);
byteCount++;
}
return val;
}
@Override
public int read(byte[] b, int off, int len) throws IOException
{
len=super.read(b, off, len);
if (len>=0)
{
crc.update(b,off,len);
byteCount+=len;
}
return len;
}
public long getCrcValue()
{
return crc.getValue();
}
public long getByteCount()
{
return byteCount;
}
}
/**
* Internal stream without header/trailer
*/
protected static class InternalGzipCompressingInputStream extends DeflaterInputStream
{
protected final CRC32InputStream crcIn;
public InternalGzipCompressingInputStream(CRC32InputStream in, int bufferSize)
{
super(in, new Deflater(Deflater.DEFAULT_COMPRESSION, true),bufferSize);
crcIn=in;
}
public void close() throws IOException
{
if (in != null)
{
try
{
def.end();
in.close();
}
finally
{
in = null;
}
}
}
protected final static int TRAILER_SIZE = 8;
public byte[] createTrailer()
{
byte[] trailer= new byte[TRAILER_SIZE];
writeTrailer(trailer, 0);
return trailer;
}
/*
* Writes GZIP member trailer to a byte array, starting at a given
* offset.
*/
private void writeTrailer(byte[] buf, int offset)
{
writeInt((int)crcIn.getCrcValue(), buf, offset); // CRC-32 of uncompr. data
writeInt((int)crcIn.getByteCount(), buf, offset + 4); // Number of uncompr. bytes
}
/*
* Writes integer in Intel byte order to a byte array, starting at a
* given offset.
*/
private void writeInt(int i, byte[] buf, int offset)
{
writeShort(i & 0xffff, buf, offset);
writeShort((i >> 16) & 0xffff, buf, offset + 2);
}
/*
* Writes short integer in Intel byte order to a byte array, starting
* at a given offset
*/
private void writeShort(int s, byte[] buf, int offset)
{
buf[offset] = (byte)(s & 0xff);
buf[offset + 1] = (byte)((s >> 8) & 0xff);
}
}
}
import static org.junit.Assert.*;
import java.io.ByteArrayInputStream;
import java.io.IOException;
import java.io.InputStream;
import java.util.Arrays;
import java.util.zip.CRC32;
import java.util.zip.GZIPInputStream;
import org.junit.Test;
public class TestGzipCompressingInputStream
{
@Test
public void test() throws Exception
{
testCompressor("test1 test2 test3");
testCompressor("1MB binary data",createTestPattern(1024*1024));
for (int i=0;i<4096;i++)
{
testCompressor(i+" bytes of binary data",createTestPattern(i));
}
}
protected byte[] createTestPattern(int size)
{
byte[] data=new byte[size];
byte pattern=0;
for (int i=0;i<size;i++)
{
data[i]=pattern++;
}
return data;
}
protected void testCompressor(String data) throws IOException
{
testCompressor("String: "+data,data.getBytes());
}
protected void testCompressor(String dataInfo, byte[] data) throws IOException
{
InputStream uncompressedIn=new ByteArrayInputStream(data);
InputStream compressedIn=new GzipCompressingInputStream(uncompressedIn);
InputStream uncompressedOut=new GZIPInputStream(compressedIn);
byte[] result=StreamHelper.readBinaryStream(uncompressedOut);
assertTrue("Test failed for: "+dataInfo,Arrays.equals(data,result));
}
}
A working example of a compressing input stream can be found in the popular open source ESB Mule: GZIPCompressorInputStream.
It uses the DeflaterInputStream provided by the JRE for compression, prepends the gzip header and appends the gzip trailer (aka footer).
Unfortunately, it is under CPA License, which does not seem to be very common. In addition, there seems to be no unit test.