Module org.hdfgroup.object
Package hdf.object

Class Dataset

Object
hdf.object.HObject
hdf.object.Dataset
All Implemented Interfaces:
DataFormat, MetaDataContainer, Serializable
Direct Known Subclasses:
Attribute, CompoundDS, ScalarDS
public abstract class Dataset extends HObject implements MetaDataContainer, DataFormat
The abstract class provides general APIs to create and manipulate dataset objects, and retrieve dataset properties, datatype and dimension sizes.

This class provides two convenient functions, read()/write(), to read/write data values. Reading/writing data may take many library calls if we use the library APIs directly. The read() and write functions hide all the details of these calls from users.

For more details on dataset, see HDF5 User's Guide

Version:
1.1 9/4/2007
Author:
Peter X. Cao
See Also:
ScalarDS, CompoundDS, Serialized Form

  • Field Details

    • data

      protected transient Object data
      The memory buffer that holds the raw data array of the dataset.

    • rank

      protected int rank
      The number of dimensions of the dataset.

    • dims

      protected long[] dims
      The current dimension sizes of the dataset

    • maxDims

      protected long[] maxDims
      The max dimension sizes of the dataset

    • selectedDims

      protected long[] selectedDims
      Array that contains the number of data points selected (for read/write) in each dimension.

      The selected size must be less than or equal to the current dimension size. A subset of a rectangle selection is defined by the starting position and selected sizes.

      For example, if a 4 X 5 dataset is as follows: 

           0,  1,  2,  3,  4
    10, 11, 12, 13, 14
    20, 21, 22, 23, 24
    30, 31, 32, 33, 34
 long[] dims = {4, 5};
 long[] startDims = {1, 2};
 long[] selectedDims = {3, 3};
 then the following subset is selected by the startDims and selectedDims above:
     12, 13, 14
     22, 23, 24
     32, 33, 34

    • startDims

      protected long[] startDims
      The starting position of each dimension of a selected subset. With both the starting position and selected sizes, the subset of a rectangle selection is fully defined.

    • selectedIndex

      protected final int[] selectedIndex
      Array that contains the indices of the dimensions selected for display.

      selectedIndex[] is provided for two purposes:

      1. selectedIndex[] is used to indicate the order of dimensions for display, i.e. selectedIndex[0] = row, selectedIndex[1] = column and selectedIndex[2] = depth. For example, for a four dimension dataset, if selectedIndex[] is {1, 2, 3}, then dim[1] is selected as row index, dim[2] is selected as column index and dim[3] is selected as depth index.
      2. selectedIndex[] is also used to select dimensions for display for datasets with three or more dimensions. We assume that applications such as HDFView can only display data up to three dimensions (a 2D spreadsheet/image with a third dimension that the 2D spreadsheet/image is cut from). For datasets with more than three dimensions, we need selectedIndex[] to store which three dimensions are chosen for display. For example, for a four dimension dataset, if selectedIndex[] = {1, 2, 3}, then dim[1] is selected as row index, dim[2] is selected as column index and dim[3] is selected as depth index. dim[0] is not selected. Its location is fixed at 0 by default.

    • selectedStride

      protected long[] selectedStride
      The number of elements to move from the start location in each dimension. For example, if selectedStride[0] = 2, every other data point is selected along dim[0].

    • chunkSize

      protected long[] chunkSize
      The array of dimension sizes for a chunk.

    • compression

      protected StringBuilder compression
      The compression information.

    • COMPRESSION_GZIP_TXT

      public static final String COMPRESSION_GZIP_TXT
      See Also:
      Constant Field Values

    • filters

      protected StringBuilder filters
      The filters information.

    • storageLayout

      protected StringBuilder storageLayout
      The storage layout information.

    • storage

      protected StringBuilder storage
      The storage information.

    • datatype

      protected Datatype datatype
      The datatype object of the dataset.

    • dimNames

      protected String[] dimNames
      Array of strings that represent the dimension names. It is null if dimension names do not exist.

    • convertByteToString

      protected boolean convertByteToString
      Flag to indicate if the byte[] array is converted to strings

    • isDataLoaded

      protected boolean isDataLoaded
      Flag to indicate if data values are loaded into memory.

    • inited

      protected boolean inited
      Flag to indicate if this dataset has been initialized

    • nPoints

      protected long nPoints
      The number of data points in the memory buffer.

    • originalBuf

      protected transient Object originalBuf
      The data buffer that contains the raw data directly reading from file (before any data conversion).

    • convertedBuf

      protected transient Object convertedBuf
      The array that holds the converted data of unsigned C-type integers.

      For example, Suppose that the original data is an array of unsigned 16-bit short integers. Since Java does not support unsigned integer, the data is converted to an array of 32-bit singed integer. In that case, the converted buffer is the array of 32-bit singed integer.

  • Constructor Details

    • Dataset

      public Dataset​(FileFormat theFile, String dsName, String dsPath)
      Constructs a Dataset object with a given file, name and path.
      Parameters:
      theFile - the file that contains the dataset.
      dsName - the name of the Dataset, e.g. "dset1".
      dsPath - the full group path of this Dataset, e.g. "/arrays/".

    • Dataset

      @Deprecated public Dataset​(FileFormat theFile, String dsName, String dsPath, long[] oid)
      Deprecated.
      Not for public use in the future.
      Using Dataset(FileFormat, String, String)
      Parameters:
      theFile - the file that contains the dataset.
      dsName - the name of the Dataset, e.g. "dset1".
      dsPath - the full group path of this Dataset, e.g. "/arrays/".
      oid - the oid of this Dataset.

  • Method Details

    • clear

      public void clear()
      Clears memory held by the dataset, such as the data buffer.

    • getRank

      public final int getRank()
      Returns the rank (number of dimensions) of the dataset.
      Specified by:
      getRank in interface DataFormat
      Returns:
      the number of dimensions of the dataset.

    • getDims

      public final long[] getDims()
      Returns the array that contains the dimension sizes of the dataset.
      Specified by:
      getDims in interface DataFormat
      Returns:
      the dimension sizes of the dataset.

    • getMaxDims

      public final long[] getMaxDims()
      Returns the array that contains the max dimension sizes of the dataset.
      Returns:
      the max dimension sizes of the dataset.

    • getSelectedDims

      public final long[] getSelectedDims()
      Returns the dimension sizes of the selected subset.

      The SelectedDims is the number of data points of the selected subset. Applications can use this array to change the size of selected subset. The selected size must be less than or equal to the current dimension size. Combined with the starting position, selected sizes and stride, the subset of a rectangle selection is fully defined.

      For example, if a 4 X 5 dataset is as follows: 

           0,  1,  2,  3,  4
    10, 11, 12, 13, 14
    20, 21, 22, 23, 24
    30, 31, 32, 33, 34
 long[] dims = {4, 5};
 long[] startDims = {1, 2};
 long[] selectedDims = {3, 3};
 long[] selectedStride = {1, 1};
 then the following subset is selected by the startDims and selectedDims
     12, 13, 14
     22, 23, 24
     32, 33, 34
      Specified by:
      getSelectedDims in interface DataFormat
      Returns:
      the dimension sizes of the selected subset.

    • getStartDims

      public final long[] getStartDims()
      Returns the starting position of a selected subset.

      Applications can use this array to change the starting position of a selection. Combined with the selected dimensions, selected sizes and stride, the subset of a rectangle selection is fully defined.

      For example, if a 4 X 5 dataset is as follows: 

           0,  1,  2,  3,  4
    10, 11, 12, 13, 14
    20, 21, 22, 23, 24
    30, 31, 32, 33, 34
 long[] dims = {4, 5};
 long[] startDims = {1, 2};
 long[] selectedDims = {3, 3};
 long[] selectedStride = {1, 1};
 then the following subset is selected by the startDims and selectedDims
     12, 13, 14
     22, 23, 24
     32, 33, 34
      Specified by:
      getStartDims in interface DataFormat
      Returns:
      the starting position of a selected subset.

    • getStride

      public final long[] getStride()
      Returns the selectedStride of the selected dataset.

      Applications can use this array to change how many elements to move in each dimension. Combined with the starting position and selected sizes, the subset of a rectangle selection is defined.

      For example, if a 4 X 5 dataset is as follows: 

           0,  1,  2,  3,  4
    10, 11, 12, 13, 14
    20, 21, 22, 23, 24
    30, 31, 32, 33, 34
 long[] dims = {4, 5};
 long[] startDims = {0, 0};
 long[] selectedDims = {2, 2};
 long[] selectedStride = {2, 3};
 then the following subset is selected by the startDims and selectedDims
     0,   3
     20, 23
      Specified by:
      getStride in interface DataFormat
      Returns:
      the selectedStride of the selected dataset.

    • setConvertByteToString

      public final void setConvertByteToString​(boolean b)
      Sets the flag that indicates if a byte array is converted to a string array.

      In a string dataset, the raw data from file is stored in a byte array. By default, this byte array is converted to an array of strings. For a large dataset (e.g. more than one million strings), the conversion takes a long time and requires a lot of memory space to store the strings. In some applications, such a conversion can be delayed. For example, A GUI application may convert only the part of the strings that is visible to the users, not the entire data array.

      setConvertByteToString(boolean b) allows users to set the flag so that applications can choose to perform the byte-to-string conversion or not. If the flag is set to false, the getData() returns an array of byte instead of an array of strings.

      Parameters:
      b - convert bytes to strings if b is true; otherwise, if false, do not convert bytes to strings.

    • getConvertByteToString

      public final boolean getConvertByteToString()
      Returns the flag that indicates if a byte array is converted to a string array.
      Returns:
      true if byte array is converted to string; otherwise, returns false if there is no conversion.

    • readBytes

      public abstract byte[] readBytes() throws Exception
      Reads the raw data of the dataset from file to a byte array.

      readBytes() reads raw data to an array of bytes instead of array of its datatype. For example, for a one-dimension 32-bit integer dataset of size 5, readBytes() returns a byte array of size 20 instead of an int array of 5.

      readBytes() can be used to copy data from one dataset to another efficiently because the raw data is not converted to its native type, it saves memory space and CPU time.

      Returns:
      the byte array of the raw data.
      Throws:
      Exception - if data can not be read

    • write

      public final void write() throws Exception
      Writes the memory buffer of this dataset to file.
      Specified by:
      write in interface DataFormat
      Throws:
      Exception - if buffer can not be written

    • copy

      public abstract Dataset copy​(Group pgroup, String name, long[] dims, Object data) throws Exception
      Creates a new dataset and writes the data buffer to the new dataset.

      This function allows applications to create a new dataset for a given data buffer. For example, users can select a specific interesting part from a large image and create a new image with the selection.

      The new dataset retains the datatype and dataset creation properties of this dataset.

      Parameters:
      pgroup - the group which the dataset is copied to.
      name - the name of the new dataset.
      dims - the dimension sizes of the the new dataset.
      data - the data values of the subset to be copied.
      Returns:
      the new dataset.
      Throws:
      Exception - if dataset can not be copied

    • isInited

      public final boolean isInited()
      Specified by:
      isInited in interface DataFormat

    • getData

      public final Object getData() throws Exception, OutOfMemoryError
      Returns the data buffer of the dataset in memory.

      If data is already loaded into memory, returns the data; otherwise, calls read() to read data from file into a memory buffer and returns the memory buffer.

      By default, the whole dataset is read into memory. Users can also select a subset to read. Subsetting is done in an implicit way.

      How to Select a Subset

      A selection is specified by three arrays: start, stride and count.

      1. start: offset of a selection
      2. stride: determines how many elements to move in each dimension
      3. count: number of elements to select in each dimension
      getStartDims(), getStride() and getSelectedDims() returns the start, stride and count arrays respectively. Applications can make a selection by changing the values of the arrays.

      The following example shows how to make a subset. In the example, the dataset is a 4-dimensional array of [200][100][50][10], i.e. dims[0]=200; dims[1]=100; dims[2]=50; dims[3]=10;
      We want to select every other data point in dims[1] and dims[2] 

       int rank = dataset.getRank(); // number of dimensions of the dataset
 long[] dims = dataset.getDims(); // the dimension sizes of the dataset
 long[] selected = dataset.getSelectedDims(); // the selected size of the dataet
 long[] start = dataset.getStartDims(); // the offset of the selection
 long[] stride = dataset.getStride(); // the stride of the dataset
 int[] selectedIndex = dataset.getSelectedIndex(); // the selected dimensions for display

 // select dim1 and dim2 as 2D data for display,and slice through dim0
 selectedIndex[0] = 1;
 selectedIndex[1] = 2;
 selectedIndex[1] = 0;

 // reset the selection arrays
 for (int i = 0; i < rank; i++) {
     start[i] = 0;
     selected[i] = 1;
     stride[i] = 1;
 }

 // set stride to 2 on dim1 and dim2 so that every other data point is
 // selected.
 stride[1] = 2;
 stride[2] = 2;

 // set the selection size of dim1 and dim2
 selected[1] = dims[1] / stride[1];
 selected[2] = dims[1] / stride[2];

 // when dataset.getData() is called, the selection above will be used since
 // the dimension arrays are passed by reference. Changes of these arrays
 // outside the dataset object directly change the values of these array
 // in the dataset object.

      For ScalarDS, the memory data buffer is a one-dimensional array of byte, short, int, float, double or String type based on the datatype of the dataset.

      For CompoundDS, the memory data object is an java.util.List object. Each element of the list is a data array that corresponds to a compound field.

      For example, if compound dataset "comp" has the following nested structure, and member datatypes 

       comp --> m01 (int)
 comp --> m02 (float)
 comp --> nest1 --> m11 (char)
 comp --> nest1 --> m12 (String)
 comp --> nest1 --> nest2 --> m21 (long)
 comp --> nest1 --> nest2 --> m22 (double)
      getData() returns a list of six arrays: {int[], float[], char[], String[], long[] and double[]}.
      Specified by:
      getData in interface DataFormat
      Returns:
      the memory buffer of the dataset.
      Throws:
      Exception - if object can not be read
      OutOfMemoryError - if memory is exhausted

    • setData

      public final void setData​(Object d)
      Not for public use in the future.

      setData() is not safe to use because it changes memory buffer of the dataset object. Dataset operations such as write/read will fail if the buffer type or size is changed.

      Specified by:
      setData in interface DataFormat
      Parameters:
      d - the object data -must be an array of Objects

    • clearData

      public void clearData()
      Clears the current data buffer in memory and forces the next read() to load the data from file.

      The function read() loads data from file into memory only if the data is not read. If data is already in memory, read() just returns the memory buffer. Sometimes we want to force read() to re-read data from file. For example, when the selection is changed, we need to re-read the data.

      Specified by:
      clearData in interface DataFormat
      See Also:
      getData(), DataFormat.read()

    • getHeight

      public final long getHeight()
      Returns the dimension size of the vertical axis.

      This function is used by GUI applications such as HDFView. GUI applications display a dataset in a 2D table or 2D image. The display order is specified by the index array of selectedIndex as follow:

      selectedIndex[0] -- height
      The vertical axis
      selectedIndex[1] -- width
      The horizontal axis
      selectedIndex[2] -- depth
      The depth axis is used for 3 or more dimensional datasets.
      Applications can use getSelectedIndex() to access and change the display order. For example, in a 2D dataset of 200x50 (dim0=200, dim1=50), the following code will set the height=200 and width=50. 
       int[] selectedIndex = dataset.getSelectedIndex();
 selectedIndex[0] = 0;
 selectedIndex[1] = 1;
      Specified by:
      getHeight in interface DataFormat
      Returns:
      the size of dimension of the vertical axis.
      See Also:
      getSelectedIndex(), getWidth()

    • getWidth

      public final long getWidth()
      Returns the dimension size of the horizontal axis.

      This function is used by GUI applications such as HDFView. GUI applications display a dataset in 2D Table or 2D Image. The display order is specified by the index array of selectedIndex as follow:

      selectedIndex[0] -- height
      The vertical axis
      selectedIndex[1] -- width
      The horizontal axis
      selectedIndex[2] -- depth
      The depth axis, which is used for 3 or more dimension datasets.
      Applications can use getSelectedIndex() to access and change the display order. For example, in a 2D dataset of 200x50 (dim0=200, dim1=50), the following code will set the height=200 and width=100. 
       int[] selectedIndex = dataset.getSelectedIndex();
 selectedIndex[0] = 0;
 selectedIndex[1] = 1;
      Specified by:
      getWidth in interface DataFormat
      Returns:
      the size of dimension of the horizontal axis.
      See Also:
      getSelectedIndex(), getHeight()

    • getSelectedIndex

      public final int[] getSelectedIndex()
      Returns the indices of display order.

      selectedIndex[] is provided for two purposes:

      1. selectedIndex[] is used to indicate the order of dimensions for display. selectedIndex[0] is for the row, selectedIndex[1] is for the column and selectedIndex[2] for the depth.

        For example, for a four dimension dataset, if selectedIndex[] = {1, 2, 3}, then dim[1] is selected as row index, dim[2] is selected as column index and dim[3] is selected as depth index.

      2. selectedIndex[] is also used to select dimensions for display for datasets with three or more dimensions. We assume that applications such as HDFView can only display data values up to three dimensions (2D spreadsheet/image with a third dimension which the 2D spreadsheet/image is selected from). For datasets with more than three dimensions, we need selectedIndex[] to tell applications which three dimensions are chosen for display.
        For example, for a four dimension dataset, if selectedIndex[] = {1, 2, 3}, then dim[1] is selected as row index, dim[2] is selected as column index and dim[3] is selected as depth index. dim[0] is not selected. Its location is fixed at 0 by default.
      Specified by:
      getSelectedIndex in interface DataFormat
      Returns:
      the array of the indices of display order.

    • getCompression

      public final String getCompression()
      Returns the string representation of compression information.

      For example, "SZIP: Pixels per block = 8: H5Z_FILTER_CONFIG_DECODE_ENABLED".

      Specified by:
      getCompression in interface DataFormat
      Returns:
      the string representation of compression information.

    • getFilters

      public final String getFilters()
      Returns the string representation of filter information.
      Returns:
      the string representation of filter information.

    • getStorageLayout

      public final String getStorageLayout()
      Returns the string representation of storage layout information.
      Returns:
      the string representation of storage layout information.

    • getStorage

      public final String getStorage()
      Returns the string representation of storage information.
      Returns:
      the string representation of storage information.

    • getChunkSize

      public final long[] getChunkSize()
      Returns the array that contains the dimension sizes of the chunk of the dataset. Returns null if the dataset is not chunked.
      Returns:
      the array of chunk sizes or returns null if the dataset is not chunked.

    • getDatatype

      public Datatype getDatatype()
      Description copied from interface: DataFormat
      Returns the datatype of the data object.
      Specified by:
      getDatatype in interface DataFormat
      Returns:
      the datatype of the data object.

    • convertFromUnsignedC

      @Deprecated public static Object convertFromUnsignedC​(Object dataIN)
      Deprecated.
      Not for public use in the future.
      Using convertFromUnsignedC(Object, Object)
      Parameters:
      dataIN - the object data
      Returns:
      the converted object

    • convertFromUnsignedC

      public static Object convertFromUnsignedC​(Object dataIN, Object dataOUT)
      Converts one-dimension array of unsigned C-type integers to a new array of appropriate Java integer in memory.

      Since Java does not support unsigned integer, values of unsigned C-type integers must be converted into its appropriate Java integer. Otherwise, the data value will not displayed correctly. For example, if an unsigned C byte, x = 200, is stored into an Java byte y, y will be -56 instead of the correct value of 200.

      Unsigned C integers are upgrade to Java integers according to the following table:

      Mapping Unsigned C Integers to Java Integers
      Unsigned C Integer JAVA Integer
      unsigned byte signed short
      unsigned short signed int
      unsigned int signed long
      unsigned long signed long
      NOTE: this conversion cannot deal with unsigned 64-bit integers. Therefore, the values of unsigned 64-bit datasets may be wrong in Java applications.

      If memory data of unsigned integers is converted by convertFromUnsignedC(), convertToUnsignedC() must be called to convert the data back to unsigned C before data is written into file.

      Parameters:
      dataIN - the input 1D array of the unsigned C-type integers.
      dataOUT - the output converted (or upgraded) 1D array of Java integers.
      Returns:
      the upgraded 1D array of Java integers.
      See Also:
      convertToUnsignedC(Object, Object)

    • convertToUnsignedC

      @Deprecated public static Object convertToUnsignedC​(Object dataIN)
      Deprecated.
      Not for public use in the future.
      Using convertToUnsignedC(Object, Object)
      Parameters:
      dataIN - the input 1D array of the unsigned C-type integers.
      Returns:
      the upgraded 1D array of Java integers.

    • convertToUnsignedC

      public static Object convertToUnsignedC​(Object dataIN, Object dataOUT)
      Converts the array of converted unsigned integers back to unsigned C-type integer data in memory.

      If memory data of unsigned integers is converted by convertFromUnsignedC(), convertToUnsignedC() must be called to convert the data back to unsigned C before data is written into file.

      Parameters:
      dataIN - the input array of the Java integer.
      dataOUT - the output array of the unsigned C-type integer.
      Returns:
      the converted data of unsigned C-type integer array.
      See Also:
      convertFromUnsignedC(Object, Object)

    • byteToString

      public static final String[] byteToString​(byte[] bytes, int length)
      Converts an array of bytes into an array of Strings for a fixed string dataset.

      A C-string is an array of chars while an Java String is an object. When a string dataset is read into a Java application, the data is stored in an array of Java bytes. byteToString() is used to convert the array of bytes into an array of Java strings so that applications can display and modify the data content.

      For example, the content of a two element C string dataset is {"ABC", "abc"}. Java applications will read the data into a byte array of {65, 66, 67, 97, 98, 99). byteToString(bytes, 3) returns an array of Java String of strs[0]="ABC", and strs[1]="abc".

      If memory data of strings is converted to Java Strings, stringToByte() must be called to convert the memory data back to byte array before data is written to file.

      Parameters:
      bytes - the array of bytes to convert.
      length - the length of string.
      Returns:
      the array of Java String.
      See Also:
      stringToByte(String[], int)

    • stringToByte

      public static final byte[] stringToByte​(String[] strings, int length)
      Converts a string array into an array of bytes for a fixed string dataset.

      If memory data of strings is converted to Java Strings, stringToByte() must be called to convert the memory data back to byte array before data is written to file.

      Parameters:
      strings - the array of string.
      length - the length of string.
      Returns:
      the array of bytes.
      See Also:
      byteToString(byte[] bytes, int length)

    • getDimNames

      public final String[] getDimNames()
      Returns the array of strings that represent the dimension names. Returns null if there is no dimension name.

      Some datasets have pre-defined names for each dimension such as "Latitude" and "Longitude". getDimNames() returns these pre-defined names.

      Returns:
      the names of dimensions, or null if there is no dimension name.

    • isString

      public boolean isString​(long tid)
      Checks if a given datatype is a string. Sub-classes must replace this default implementation.
      Parameters:
      tid - The data type identifier.
      Returns:
      true if the datatype is a string; otherwise returns false.

    • getSize

      public long getSize​(long tid)
      Returns the size in bytes of a given datatype. Sub-classes must replace this default implementation.
      Parameters:
      tid - The data type identifier.
      Returns:
      The size of the datatype

    • getOriginalClass

      public final Class getOriginalClass()
      Get Class of the original data buffer if converted.
      Specified by:
      getOriginalClass in interface DataFormat
      Returns:
      the Class of originalBuf

    • isVirtual

      public boolean isVirtual()

    • getVirtualFilename

      public String getVirtualFilename​(int index)

    • getVirtualMaps

      public int getVirtualMaps()