Data type objects (dtype)

A data type object (an instance of numpy.dtype class) describes how the bytes in the fixed-size block of memory corresponding to an array item should be interpreted. It describes the following aspects of the data:

  1. Type of the data (integer, float, Python object, etc.)
  2. Size of the data (how many bytes is in e.g. the integer)
  3. Byte order of the data (little-endian or big-endian)
  4. If the data type is a record, an aggregate of other data types, (e.g., describing an array item consisting of an integer and a float),
    1. what are the names of the “fields” of the record, by which they can be accessed,
    2. what is the data-type of each field, and
    3. which part of the memory block each field takes.
  5. If the data is a sub-array, what is its shape and data type.

To describe the type of scalar data, there are several built-in scalar types in Numpy for various precision of integers, floating-point numbers, etc. An item extracted from an array, e.g., by indexing, will be a Python object whose type is the scalar type associated with the data type of the array.

Note that the scalar types are not dtype objects, even though they can be used in place of one whenever a data type specification is needed in Numpy.

Record data types are formed by creating a data type whose fields contain other data types. Each field has a name by which it can be accessed. The parent data type should be of sufficient size to contain all its fields; the parent can for example be based on the void type which allows an arbitrary item size. Record data types may also contain other record types and fixed-size sub-array data types in their fields.

Finally, a data type can describe items that are themselves arrays of items of another data type. These sub-arrays must, however, be of a fixed size. If an array is created using a data-type describing a sub-array, the dimensions of the sub-array are appended to the shape of the array when the array is created. Sub-arrays in a field of a record behave differently, see Record Access.


A simple data type containing a 32-bit big-endian integer: (see Specifying and constructing data types for details on construction)

>>> dt = np.dtype('>i4')
>>> dt.byteorder
>>> dt.itemsize
>>> dt.type is np.int32

The corresponding array scalar type is int32.


A record data type containing a 16-character string (in field ‘name’) and a sub-array of two 64-bit floating-point number (in field ‘grades’):

>>> dt = np.dtype([('name', np.str_, 16), ('grades', np.float64, (2,))])
>>> dt['name']
>>> dt['grades']

Items of an array of this data type are wrapped in an array scalar type that also has two fields:

>>> x = np.array([('Sarah', (8.0, 7.0)), ('John', (6.0, 7.0))], dtype=dt)
>>> x[1]
('John', [6.0, 7.0])
>>> x[1]['grades']
array([ 6.,  7.])
>>> type(x[1])
<type 'numpy.void'>
>>> type(x[1]['grades'])
<type 'numpy.ndarray'>

Specifying and constructing data types

Whenever a data-type is required in a NumPy function or method, either a dtype object or something that can be converted to one can be supplied. Such conversions are done by the dtype constructor:

dtype Create a data type object.

What can be converted to a data-type object is described below:

dtype object

Used as-is.


The default data type: float_.

Array-scalar types

The 24 built-in array scalar type objects all convert to an associated data-type object. This is true for their sub-classes as well.

Note that not all data-type information can be supplied with a type-object: for example, flexible data-types have a default itemsize of 0, and require an explicitly given size to be useful.


>>> dt = np.dtype(np.int32)      # 32-bit integer
>>> dt = np.dtype(np.complex128) # 128-bit complex floating-point number

Generic types

The generic hierarchical type objects convert to corresponding type objects according to the associations:

number, inexact, floating float
complexfloating cfloat
integer, signedinteger int_
unsignedinteger uint
character string
generic, flexible void

Built-in Python types

Several python types are equivalent to a corresponding array scalar when used to generate a dtype object:

int int_
bool bool_
float float_
complex cfloat
str string
unicode unicode_
buffer void
(all others) object_


>>> dt = np.dtype(float)   # Python-compatible floating-point number
>>> dt = np.dtype(int)     # Python-compatible integer
>>> dt = np.dtype(object)  # Python object

Types with .dtype

Any type object with a dtype attribute: The attribute will be accessed and used directly. The attribute must return something that is convertible into a dtype object.

Several kinds of strings can be converted. Recognized strings can be prepended with '>' (big-endian), '<' (little-endian), or '=' (hardware-native, the default), to specify the byte order.

One-character strings

Each built-in data-type has a character code (the updated Numeric typecodes), that uniquely identifies it.


>>> dt = np.dtype('b')  # byte, native byte order
>>> dt = np.dtype('>H') # big-endian unsigned short
>>> dt = np.dtype('<f') # little-endian single-precision float
>>> dt = np.dtype('d')  # double-precision floating-point number

Array-protocol type strings (see The Array Interface)

The first character specifies the kind of data and the remaining characters specify how many bytes of data. The supported kinds are

'b' Boolean
'i' (signed) integer
'u' unsigned integer
'f' floating-point
'c' complex-floating point
'S', 'a' string
'U' unicode
'V' anything (void)


>>> dt = np.dtype('i4')   # 32-bit signed integer
>>> dt = np.dtype('f8')   # 64-bit floating-point number
>>> dt = np.dtype('c16')  # 128-bit complex floating-point number
>>> dt = np.dtype('a25')  # 25-character string

String with comma-separated fields

Numarray introduced a short-hand notation for specifying the format of a record as a comma-separated string of basic formats.

A basic format in this context is an optional shape specifier followed by an array-protocol type string. Parenthesis are required on the shape if it is greater than 1-d. NumPy allows a modification on the format in that any string that can uniquely identify the type can be used to specify the data-type in a field. The generated data-type fields are named 'f0', 'f1', ..., 'f<N-1>' where N (>1) is the number of comma-separated basic formats in the string. If the optional shape specifier is provided, then the data-type for the corresponding field describes a sub-array.


  • field named f0 containing a 32-bit integer
  • field named f1 containing a 2 x 3 sub-array of 64-bit floating-point numbers
  • field named f2 containing a 32-bit floating-point number
>>> dt = np.dtype("i4, (2,3)f8, f4")
  • field named f0 containing a 3-character string
  • field named f1 containing a sub-array of shape (3,) containing 64-bit unsigned integers
  • field named f2 containing a 3 x 4 sub-array containing 10-character strings
>>> dt = np.dtype("a3, 3u8, (3,4)a10")

Type strings

Any string in numpy.sctypeDict.keys():


>>> dt = np.dtype('uint32')   # 32-bit unsigned integer
>>> dt = np.dtype('Float64')  # 64-bit floating-point number

(flexible_dtype, itemsize)

The first argument must be an object that is converted to a flexible data-type object (one whose element size is 0), the second argument is an integer providing the desired itemsize.


>>> dt = np.dtype((void, 10))  # 10-byte wide data block
>>> dt = np.dtype((str, 35))   # 35-character string
>>> dt = np.dtype(('U', 10))   # 10-character unicode string

(fixed_dtype, shape)

The first argument is any object that can be converted into a fixed-size data-type object. The second argument is the desired shape of this type. If the shape parameter is 1, then the data-type object is equivalent to fixed dtype. If shape is a tuple, then the new dtype defines a sub-array of the given shape.


>>> dt = np.dtype((np.int32, (2,2)))          # 2 x 2 integer sub-array
>>> dt = np.dtype(('S10', 1))                 # 10-character string
>>> dt = np.dtype(('i4, (2,3)f8, f4', (2,3))) # 2 x 3 record sub-array

(base_dtype, new_dtype)

Both arguments must be convertible to data-type objects in this case. The base_dtype is the data-type object that the new data-type builds on. This is how you could assign named fields to any built-in data-type object.


32-bit integer, whose first two bytes are interpreted as an integer via field real, and the following two bytes via field imag.

>>> dt = np.dtype((np.int32,{'real':(np.int16, 0),'imag':(np.int16, 2)})

32-bit integer, which is interpreted as consisting of a sub-array of shape (4,) containing 8-bit integers:

>>> dt = np.dtype((np.int32, (np.int8, 4)))

32-bit integer, containing fields r, g, b, a that interpret the 4 bytes in the integer as four unsigned integers:

>>> dt = np.dtype(('i4', [('r','u1'),('g','u1'),('b','u1'),('a','u1')]))

[(field_name, field_dtype, field_shape), ...]

obj should be a list of fields where each field is described by a tuple of length 2 or 3. (Equivalent to the descr item in the __array_interface__ attribute.)

The first element, field_name, is the field name (if this is '' then a standard field name, 'f#', is assigned). The field name may also be a 2-tuple of strings where the first string is either a “title” (which may be any string or unicode string) or meta-data for the field which can be any object, and the second string is the “name” which must be a valid Python identifier.

The second element, field_dtype, can be anything that can be interpreted as a data-type.

The optional third element field_shape contains the shape if this field represents an array of the data-type in the second element. Note that a 3-tuple with a third argument equal to 1 is equivalent to a 2-tuple.

This style does not accept align in the dtype constructor as it is assumed that all of the memory is accounted for by the array interface description.


Data-type with fields big (big-endian 32-bit integer) and little (little-endian 32-bit integer):

>>> dt = np.dtype([('big', '>i4'), ('little', '<i4')])

Data-type with fields R, G, B, A, each being an unsigned 8-bit integer:

>>> dt = np.dtype([('R','u1'), ('G','u1'), ('B','u1'), ('A','u1')])

{'names': ..., 'formats': ..., 'offsets': ..., 'titles': ...}

This style has two required and two optional keys. The names and formats keys are required. Their respective values are equal-length lists with the field names and the field formats. The field names must be strings and the field formats can be any object accepted by dtype constructor.

The optional keys in the dictionary are offsets and titles and their values must each be lists of the same length as the names and formats lists. The offsets value is a list of byte offsets (integers) for each field, while the titles value is a list of titles for each field (None can be used if no title is desired for that field). The titles can be any string or unicode object and will add another entry to the fields dictionary keyed by the title and referencing the same field tuple which will contain the title as an additional tuple member.


Data type with fields r, g, b, a, each being a 8-bit unsigned integer:

>>> dt = np.dtype({'names': ['r','g','b','a'],
...                'formats': [uint8, uint8, uint8, uint8]})

Data type with fields r and b (with the given titles), both being 8-bit unsigned integers, the first at byte position 0 from the start of the field and the second at position 2:

>>> dt = np.dtype({'names': ['r','b'], 'formats': ['u1', 'u1'],
...                'offsets': [0, 2],
...                'titles': ['Red pixel', 'Blue pixel']})

{'field1': ..., 'field2': ..., ...}

This style allows passing in the fields attribute of a data-type object.

obj should contain string or unicode keys that refer to (data-type, offset) or (data-type, offset, title) tuples.


Data type containing field col1 (10-character string at byte position 0), col2 (32-bit float at byte position 10), and col3 (integers at byte position 14):

>>> dt = np.dtype({'col1': ('S10', 0), 'col2': (float32, 10),
    'col3': (int, 14)})


Numpy data type descriptions are instances of the dtype class.


The type of the data is described by the following dtype attributes:

dtype.type The type object used to instantiate a scalar of this data-type.
dtype.kind A character code (one of ‘biufcSUV’) identifying the general kind of data.
dtype.char A unique character code for each of the 21 different built-in types.
dtype.num A unique number for each of the 21 different built-in types.
dtype.str The array-protocol typestring of this data-type object.

Size of the data is in turn described by: A bit-width name for this data-type.
dtype.itemsize The element size of this data-type object.

Endianness of this data:

dtype.byteorder A character indicating the byte-order of this data-type object.

Information about sub-data-types in a record:

dtype.fields Dictionary of named fields defined for this data type, or None.
dtype.names Ordered list of field names, or None if there are no fields.

For data types that describe sub-arrays:

dtype.subdtype Tuple (item_dtype, shape) if this dtype describes a sub-array, and
dtype.shape Shape tuple of the sub-array if this data type describes a sub-array,

Attributes providing additional information:

dtype.hasobject Boolean indicating whether this dtype contains any reference-counted objects in any fields or sub-dtypes.
dtype.flags Bit-flags describing how this data type is to be interpreted.
dtype.isbuiltin Integer indicating how this dtype relates to the built-in dtypes.
dtype.isnative Boolean indicating whether the byte order of this dtype is native
dtype.descr Array-interface compliant full description of the data-type.
dtype.alignment The required alignment (bytes) of this data-type according to the compiler.


Data types have the following method for changing the byte order:

dtype.newbyteorder(new_order=) Return a new dtype with a different byte order.

The following methods implement the pickle protocol:


Table Of Contents

Previous topic


Next topic


This Page