ByteSequenceTensorBuffer

public class ByteSequenceTensorBuffer

Buffer for storing string tensor data.

The values are stored as an array of TF_TString, internally wrapped with tensorflow::tstring, which is essentially a portable version of std::string.

The data of the buffer must be initialized only once, by calling ERROR(/#init(NdArray, Function)), and the buffer must have been allocated with enough space (use ERROR(/#computeSize(NdArray, Function)) priory to know exactly how many bytes are required to store the data).

After its data has been initialized, the buffer is read-only as it is not possible to change safely a value without reinitializing the whole data.

Public Methods

static <T> long	computeSize(ByteSequenceProvider<?> byteSequenceProvider) Computes how many bytes are required to store the given data in a string buffer.
DataBuffer<byte[]>	copyTo(DataBuffer<byte[]> dst, long size) Write the references of the objects in the source array into this buffer.
byte[]	getObject(long index)
<T> void	init(ByteSequenceProvider<T> byteSequenceProvider) Initialize the data of this buffer.
boolean	isReadOnly()
DataBuffer<byte[]>	setObject(byte[] values, long index)
long	size()
DataBuffer<byte[]>	slice(long index, long size)

Inherited Methods

From class org.tensorflow.ndarray.impl.buffer.AbstractDataBuffer

DataBuffer<byte[]>	copyTo(DataBuffer<byte[]> dst, long size) Write the references of the objects in the source array into this buffer.
boolean	equals(Object obj)
int	hashCode()
DataBuffer<byte[]>	read(T[] dst, int offset, int length) Read the references of the objects in this buffer into the destination array.
DataBuffer<byte[]>	write(T[] src, int offset, int length) Bulk put method, using int arrays.

From class java.lang.Object

boolean	equals(Object arg0)
final Class<?>	getClass()
int	hashCode()
final void	notify()
final void	notifyAll()
String	toString()
final void	wait(long arg0, int arg1)
final void	wait(long arg0)
final void	wait()

From interface org.tensorflow.ndarray.buffer.DataBuffer

abstract <R> R	accept(DataStorageVisitor<R> visitor) Visits the backing storage of this buffer.
abstract DataBuffer<byte[]>	copyTo(DataBuffer<byte[]> dst, long size) Write the references of the objects in the source array into this buffer.
abstract boolean	equals(Object obj) Checks equality between data buffers.
abstract byte[]	getObject(long index) Reads the value at the given index.
abstract boolean	isReadOnly() Tells whether or not this buffer is backed by an accessible array.
abstract DataBuffer<byte[]>	narrow(long size) Creates a new buffer whose content is a shared subsequence of this buffer's content, whose size is set to the given value.
abstract DataBuffer<byte[]>	offset(long index) Creates a new buffer whose content is a shared subsequence of this buffer's content, starting at the given index.
abstract DataBuffer<byte[]>	read(T[] dst) Read the references of the objects in this buffer into the destination array.
abstract DataBuffer<byte[]>	read(T[] dst, int offset, int length) Read the references of the objects in this buffer into the destination array.
abstract DataBuffer<byte[]>	setObject(byte[] value, long index) Writes the given value into this buffer at the given index.
abstract long	size() Size of the buffer, in elements.
abstract DataBuffer<byte[]>	slice(long index, long size) Creates a new buffer whose content is a shared subsequence of this buffer's content, starting at the given index and of the given size.
abstract DataBufferWindow<? extends DataBuffer<T>>	window(long size) Creates a DataBufferWindow that provides a partial view of this buffer.
abstract DataBuffer<byte[]>	write(T[] src) Write the references of the objects in the source array into this buffer.
abstract DataBuffer<byte[]>	write(T[] src, int offset, int length) Bulk put method, using int arrays.

Public Methods

public static long computeSize (ByteSequenceProvider<?> byteSequenceProvider)

Computes how many bytes are required to store the given data in a string buffer.

Parameters

byteSequenceProvider	produces sequences of bytes

Returns

• number of bytes required to store the data.

public DataBuffer<byte[]> copyTo (DataBuffer<byte[]> dst, long size)

Write the references of the objects in the source array into this buffer.

If there are more values to copy than the destination buffer size, i.e. size > dst.size(), then no values are transferred and a BufferOverflowException is thrown. On the other hand, if there are more values to copy that the source buffer size, i.e. > src.size(), then a BufferUnderfloatException is thrown.

Otherwise, this method copies n = size values from this buffer into the destination buffer.

Parameters

dst	the destination buffer into which values are copied; must not be this buffer
size	number of values to copy to the destination buffer

Returns

• this buffer

public byte[] getObject (long index)

public void init (ByteSequenceProvider<T> byteSequenceProvider)

Initialize the data of this buffer.

While it is not enforced programmatically, it is mandatory that this method is called only once after the creation of the buffer. The buffer must have been allocated according to the same set of data, calling ERROR(/#computeSize(NdArray, Function)) priory to make sure there is enough space to store it.

Parameters

byteSequenceProvider	produces sequences of bytes to use as the tensor data

public boolean isReadOnly ()

public DataBuffer<byte[]> setObject (byte[] values, long index)

public long size ()

public DataBuffer<byte[]> slice (long index, long size)