def __init__(self, path, key_class, value_class, metadata, compress=False, block_compress=False):
if os.path.exists(path):
raise IOError("File %s already exists." % path)
self._key_class = key_class
self._value_class = value_class
self._compress = compress
self._block_compress = block_compress
if not metadata:
metadata = Metadata()
self._metadata = metadata
if self._compress or self._block_compress:
self._codec = CodecPool().getCompressor()
else:
self._codec = None
self._last_sync = 0
self._block = None
self._stream = DataOutputStream(FileOutputStream(path))
# sync is 16 random bytes
self._sync = md5('%s@%d' % (uuid1().bytes, int(time() * 1000))).digest()
self._writeFileHeader()
class Writer(object):
COMPRESSION_BLOCK_SIZE = 1000000
def __init__(self, path, key_class, value_class, metadata, compress=False, block_compress=False):
if os.path.exists(path):
raise IOError("File %s already exists." % path)
self._key_class = key_class
self._value_class = value_class
self._compress = compress
self._block_compress = block_compress
if not metadata:
metadata = Metadata()
self._metadata = metadata
if self._compress or self._block_compress:
self._codec = CodecPool().getCompressor()
else:
self._codec = None
self._last_sync = 0
self._block = None
self._stream = DataOutputStream(FileOutputStream(path))
# sync is 16 random bytes
self._sync = md5('%s@%d' % (uuid1().bytes, int(time() * 1000))).digest()
self._writeFileHeader()
def close(self):
if self._block_compress:
self.sync()
self._stream.close()
def getCompressionCodec(self):
return self._codec
def getKeyClass(self):
return self._key_class
def getKeyClassName(self):
return hadoopClassName(self._key_class)
def getValueClass(self):
return self._value_class
def getValueClassName(self):
return hadoopClassName(self._value_class)
def isBlockCompressed(self):
return self._block_compress
def isCompressed(self):
return self._compress
def getLength(self):
return self._stream.getPos()
def append(self, key, value):
if type(key) != self._key_class:
raise IOError("Wrong key class %s is not %s" % (type(key), self._key_class))
if type(value) != self._value_class:
raise IOError("Wrong Value class %s is not %s" % (type(value), self._value_class))
key_buffer = DataOutputBuffer()
key.write(key_buffer)
value_buffer = DataOutputBuffer()
value.write(value_buffer)
self.appendRaw(key_buffer.toByteArray(), value_buffer.toByteArray())
def appendRaw(self, key, value):
if self._block_compress:
if self._block:
records, keys_len, keys, values_len, values = self._block
else:
keys_len = DataOutputBuffer()
keys = DataOutputBuffer()
values_len = DataOutputBuffer()
values = DataOutputBuffer()
records = 0
writeVInt(keys_len, len(key))
keys.write(key)
writeVInt(values_len, len(value))
values.write(value)
records += 1
self._block = (records, keys_len, keys, values_len, values)
current_block_size = keys.getSize() + values.getSize()
if current_block_size >= self.COMPRESSION_BLOCK_SIZE:
self.sync()
else:
if self._compress:
value = self._codec.compress(value)
key_length = len(key)
value_length = len(value)
self._checkAndWriteSync()
self._stream.writeInt(key_length + value_length)
self._stream.writeInt(key_length)
self._stream.write(key)
self._stream.write(value)
def sync(self):
if self._last_sync != self._stream.getPos():
self._stream.writeInt(SYNC_ESCAPE)
self._stream.write(self._sync)
self._last_sync = self._stream.getPos()
if self._block_compress and self._block:
def _writeBuffer(data_buf):
buf = self._codec.compress(data_buf.toByteArray())
writeVInt(self._stream, len(buf))
self._stream.write(buf)
records, keys_len, keys, values_len, values = self._block
writeVInt(self._stream, records)
_writeBuffer(keys_len)
_writeBuffer(keys)
_writeBuffer(values_len)
_writeBuffer(values)
self._block = None
def _writeFileHeader(self):
self._stream.write(VERSION)
Text.writeString(self._stream, self.getKeyClassName())
Text.writeString(self._stream, self.getValueClassName())
self._stream.writeBoolean(self._compress)
self._stream.writeBoolean(self._block_compress)
if self._codec:
Text.writeString(self._stream, 'org.apache.hadoop.io.compress.DefaultCodec')
self._metadata.write(self._stream)
self._stream.write(self._sync)
def _checkAndWriteSync(self):
if self._stream.getPos() >= (self._last_sync + SYNC_INTERVAL):
self.sync()
