Package org.apache.hadoop.hbase.test.util.warc

Class WARCFileWriter

java.lang.Object

org.apache.hadoop.hbase.test.util.warc.WARCFileWriter

public class WARCFileWriter
extends Object

Writes WARCRecords to a WARC file, using Hadoop's filesystem APIs. (This means you can write to HDFS, S3 or any other filesystem supported by Hadoop). This implementation is not tied to the MapReduce APIs -- that link is provided by the mapred com.martinkl.warc.mapred.WARCOutputFormat and the mapreduce com.martinkl.warc.mapreduce.WARCOutputFormat. WARCFileWriter keeps track of how much data it has written (optionally gzip-compressed); when the file becomes larger than some threshold, it is automatically closed and a new segment is started. A segment number is appended to the filename for that purpose. The segment number always starts at 00000, and by default a new segment is started when the file size exceeds 1GB. To change the target size for a segment, you can set the `warc.output.segment.size` key in the Hadoop configuration to the number of bytes. (Files may actually be a bit larger than this threshold, since we finish writing the current record before opening a new file.)

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
private class	WARCFileWriter.CountingOutputStream

Field Summary

Fields

Modifier and Type	Field	Description
private WARCFileWriter.CountingOutputStream	byteStream
private long	bytesWritten
private final org.apache.hadoop.io.compress.CompressionCodec	codec
private final org.apache.hadoop.conf.Configuration	conf
private DataOutputStream	dataStream
static final long	DEFAULT_MAX_SEGMENT_SIZE
private final String	extensionFormat
private static final org.slf4j.Logger	logger
private final long	maxSegmentSize
private final org.apache.hadoop.util.Progressable	progress
private long	segmentsAttempted
private long	segmentsCreated
private final org.apache.hadoop.fs.Path	workOutputPath

Constructor Summary

Constructors

Constructor	Description
WARCFileWriter(org.apache.hadoop.conf.Configuration conf, org.apache.hadoop.io.compress.CompressionCodec codec, org.apache.hadoop.fs.Path workOutputPath)	Creates a WARC file, and opens it for writing.
WARCFileWriter(org.apache.hadoop.conf.Configuration conf, org.apache.hadoop.io.compress.CompressionCodec codec, org.apache.hadoop.fs.Path workOutputPath, org.apache.hadoop.util.Progressable progress)	Creates a WARC file, and opens it for writing.

Method Summary

Modifier and Type	Method	Description
void	close()	Flushes any buffered data and closes the file.
private void	createSegment()	Creates an output segment file and sets up the output streams to point at it.
static org.apache.hadoop.io.compress.CompressionCodec	getGzipCodec(org.apache.hadoop.conf.Configuration conf)	Instantiates a Hadoop codec for compressing and decompressing Gzip files.
void	write(WARCRecord record)	Appends a WARCRecord to the file, in WARC/1.0 format.
void	write(WARCWritable record)	Appends a WARCRecord wrapped in a WARCWritable to the file.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

logger

private static final org.slf4j.Logger logger

DEFAULT_MAX_SEGMENT_SIZE

public static final long DEFAULT_MAX_SEGMENT_SIZE

See Also:

• Constant Field Values

conf

private final org.apache.hadoop.conf.Configuration conf

codec

private final org.apache.hadoop.io.compress.CompressionCodec codec

workOutputPath

private final org.apache.hadoop.fs.Path workOutputPath

progress

private final org.apache.hadoop.util.Progressable progress

extensionFormat

private final String extensionFormat

maxSegmentSize

private final long maxSegmentSize

segmentsCreated

private long segmentsCreated

segmentsAttempted

private long segmentsAttempted

bytesWritten

private long bytesWritten

byteStream

private WARCFileWriter.CountingOutputStream byteStream

dataStream

private DataOutputStream dataStream

Constructor Details

WARCFileWriter

public WARCFileWriter(org.apache.hadoop.conf.Configuration conf,
 org.apache.hadoop.io.compress.CompressionCodec codec,
 org.apache.hadoop.fs.Path workOutputPath)
               throws IOException

Creates a WARC file, and opens it for writing. If a file with the same name already exists, an attempt number in the filename is incremented until we find a file that doesn't already exist.

Parameters:

conf - The Hadoop configuration.

codec - If null, the file is uncompressed. If non-null, this compression codec will be used. The codec's default file extension is appended to the filename.

workOutputPath - The directory and filename prefix to which the data should be written. We append a segment number and filename extensions to it.

Throws:

IOException

WARCFileWriter

public WARCFileWriter(org.apache.hadoop.conf.Configuration conf,
 org.apache.hadoop.io.compress.CompressionCodec codec,
 org.apache.hadoop.fs.Path workOutputPath,
 org.apache.hadoop.util.Progressable progress)
               throws IOException

Creates a WARC file, and opens it for writing. If a file with the same name already exists, it is *overwritten*. Note that this is different behaviour from the other constructor. Yes, this sucks. It will probably change in a future version.

Parameters:

conf - The Hadoop configuration.

codec - If null, the file is uncompressed. If non-null, this compression codec will be used. The codec's default file extension is appended to the filename.

workOutputPath - The directory and filename prefix to which the data should be written. We append a segment number and filename extensions to it.

progress - An object used by the mapred API for tracking a task's progress.

Throws:

IOException

Method Details

getGzipCodec

public static org.apache.hadoop.io.compress.CompressionCodec getGzipCodec(org.apache.hadoop.conf.Configuration conf)

Instantiates a Hadoop codec for compressing and decompressing Gzip files. This is the most common compression applied to WARC files.

Parameters:

conf - The Hadoop configuration.

createSegment

private void createSegment()
                    throws IOException

Creates an output segment file and sets up the output streams to point at it. If the file already exists, retries with a different filename. This is a bit nasty -- after all, FileOutputFormat's work directory concept is supposed to prevent filename clashes -- but it looks like Amazon Elastic MapReduce prevents use of per-task work directories if the output of a job is on S3. TODO: Investigate this and find a better solution.

Throws:

IOException

write

public void write(WARCRecord record)
           throws IOException

Appends a WARCRecord to the file, in WARC/1.0 format.

Parameters:

record - The record to be written.

Throws:

IOException

write

public void write(WARCWritable record)
           throws IOException

Appends a WARCRecord wrapped in a WARCWritable to the file.

Parameters:

record - The wrapper around the record to be written.

Throws:

IOException

close

public void close()
           throws IOException

Flushes any buffered data and closes the file.

Throws:

IOException