CDM Data Types

This document explains how CDM data types are mapped into Netcdf-Java objects.


An Array contains the actual data for a Variable after it is read from the disk or network. Data access can be done in a general way through the method

 Array data =;

The one exception to this is Sequences, which must be accessed through Sequence.getStructureIterator() which returns StructureData objects.

For Variables that are members of a Structure, data is accessed generally through the method

    Array data = StructureData.getArray( memberName);

When a Variable has a variable length dimension, Variable.isVariableLength() is true and a ArrayObject with the appropriate element type is returned. You cannot subset on the variable length dimension, all of it is always read. For example:

  short levels(acqtime=10, *);

  Variable v = ncfile.findVariable("levels");
  Array data =;
  NCdumpW.printArray(data, "read()",  new PrintWriter( System.out), null);
  // loop over outer dimension
  while (data.hasNext()) {
    Array as = (Array); // inner variable length array of short
    NCdumpW.printArray(as, "",  new PrintWriter( System.out), null);
  // subset ok on outer dimension
  data ="0:9:2,:");
  NCdumpW.printArray(data, "read(0:9:2,:)",  new PrintWriter( System.out), null); // ok
  data = Section().appendRange(0,9,2).appendRange(null));
  NCdumpW.printArray(data, "read(Section)",  new PrintWriter( System.out), null); // ok

You cannot use the older style read(origin, shape) interface:

  // fail
  int[] origin = new int[] {0, 0};
  int[] size = new int[] {3, -1};
  data =, size);

Otherwise the Array is rectangular, and the following is returned from or StructureData.getArray():

DataType Array subclass Array.getElementType
BYTE ArrayByte byte.class
SHORT ArrayShort short.class
INT ArrayInt int.class
LONG ArrayLong long.class
FLOAT ArrayFloat float.class
DOUBLE ArrayDouble double.class
CHAR ArrayChar char.class
STRING ArrayObject String.class
STRUCTURE ArrayStructure StructureData.class
SEQUENCE ArraySequence StructureData.class
ENUM1 ArrayByte byte.class
ENUM2 ArrayShort short.class
ENUM4 ArrayInt int.class
OPAQUE ArrayObject ByteBuffer.class

Integer Types

The CDM integer types are byte, short, int, and long. Each are mapped to the corresponding Java primitive type, which are 1,2,4, and 8 bytes wide, respectively.

When the underlying file format stores data as unsigned integers, an IOSP may decide to widen the type to a signed type, eg

  • unsigned byte > signed short
  • unsigned short > signed int
  • unsigned int > signed long

Otherwise, the the variable is kept as an unsigned integer type and the attribute _Unsigned = “true” is added to the variable. Since Java does not have unsigned integer types, handling unsigned values requires some attention. Unsigned integer data are stored in the corresponding signed primitive types. You can detect this by calling Array.isUnsigned() or Variable.isUnsigned().

You may use static methods in ucar.ma2.DataType to convert one value at a time:

  • static public long unsignedIntToLong(int i);
  • static public int unsignedShortToInt(short s);
  • static public short unsignedByteToShort(byte b);

You may use this static method in ucar.ma2.MAMath to widen all the values in an Array:

 public static Array convertUnsigned( Array unsigned);

Theres not much to do in a general way with unsigned longs, as there is no primitive type that can hold 64 bits of precision.

ArrayByte, ArrayShort and ArrayInt will widen an unsigned value when casting to wider type like float or double. For example, calling


on an unsigned integer type will return the widened value cast to a double.

Floating Point Types

The CDM integer types are float and double. Each are mapped to the corresponding Java primitive type.

Char and String Types

A String is a variable length array of Unicode characters. When reading/writing a String to a file or other external representation, the characters are by default UTF-8 encoded (note that ASCII is a subset of UTF-8). Libraries may use different internal representations, for example the Java library uses UTF-16 encoding.

The char type contains uninterpreted characters, one character per byte. Typically these contain 7-bit ASCII characters.


A Structure is a type of Variable that contains other Variables, analogous to a struct in C, or a row in a relational database. In general, the data in a Structure are physically stored close together on disk, so that it is efficient to retrieve all of the data in a Structure at the same time. A Variable contained in a Structure is a member Variable, and can only be read in the context of its containing Structure.

The member data of a Structure is returned in a StructureData object. Since a Structure may be multidimensional, returns an ArrayStructure, a subclass of Array which contains an array of StructureData objects. Alternatively, one can call Structure.getStructureIterator() and iterate through the StructureData in canonical order. This potentially is a very efficient way to access the data, since the data does not have to all be memory resident at the same time.

or type specific access:

    StructureData.getScalarXXX( memberName);
    StructureData.getJavaArrayXXX( memberName);

For nested Structure and Sequences (that is, Structure members that are themselves Structures or Sequences, use

StructureData data = StructureData.getScalarStructure( memberName);

  ArrayStructure data = StructureData.getArrayStructure( memberName);
  ArraySequence data = StructureData.getArraySequence( memberName);


A Sequence is a variable length, one dimensional Structure whose length is not known until you actually read the data. To access the data in a Sequence, you must use Structure.getStructureIterator(), ie you cannot call Sequences make the most sense as members of a Structure.


An enum type is an mapping of integer values to Strings. The mapping itself is stored in an EnumTypedef object in the Group, and so is shared across all Variables that use that enumeration.

A enumeration Variable will have DataType ENUM1, ENUM2, or ENUM4, depending on whether the the enum value is stored in 1, 2, or 4 bytes. The raw values are returned in a byte, short, or integer array. One can convert these raw values to the corresponding String enumeration values in a way that does not depend on their internal representation, for example:

 if (var.getDataType().isEnum()) {
    Array rawValues =;
    Array enumValues = Array.factory(DataType.STRING, rawValues.getShape());
    IndexIterator ii = enumValues.getIndexIterator();

    // use implicit Array iterator

    while (rawValues.hasNext()) {
      String sval = var.lookupEnumString(rawValues.nextInt());

When wrapping a NetcdfFile with a NetcdfDataset, by default enumerations will be converted to their String values, and the type of the Variable will be changed to String. This conversion is controlled by the enhanceMode parameter on

static public NetcdfDataset openDataset(String location, EnumSet enhanceMode, int buffer_size, CancelTask cancelTask, Object spiObject); or by setting the default enhanceMode:

static public void setDefaultEnhanceMode(EnumSet mode); If you want to turn enum conversion off, create your own EnumSet enhanceMode, for example this leaves the other default enhancements on:

EnumSet myEnhanceMode = EnumSet.of(Enhance.ScaleMissing, Enhance.CoordSystems);


An opaque type stores uninterpreted blobs of bytes. The length of the blob is not known until it is read, and an array of opaque objects may have different lengths for each of the objects. Opaque data is returned as java.nio.ByteBuffer objects wrapped by an ArrayObject.