NumPy astype()

The astype() method converts an array to a specified data type.

Example

import numpy as np

# original array of integers
integerArray = np.array([1, 2, 3, 4, 5])

# convert array to floating-point numbers floatArray = integerArray.astype(float)
print(floatArray) # Output: [1. 2. 3. 4. 5.]

astype() Syntax

The syntax of astype() is:

ndarray.astype(dtype, order = 'K', casting = 'unsafe', subok = True, copy = True)

astype() Arguments

The astype() method takes five arguments:

  • dtype - desired data type for the new array
  • order (optional) - memory layout order of the returned array
  • casting (optional) - casting behavior when converting data types
  • subok (optional) - determines whether to subclass the output array if the data type is changed or to return a base-class array
  • copy (optional) - creates a copy if True, modifies the original array if False

astype() Return Value

The astype() method returns the modified array:

  • If the copy argument is True, a new array is returned.
  • If the copy argument is False, the original array is modified.

Example 1: Convert an Integer Array to Different Data Types

import numpy as np

# create a 1D array
array = np.array([0, 1, 2, 3, 4, 5])

# convert to different data types
floatArray = array.astype(float)
complexArray = array.astype(complex)
boolArray = array.astype(bool)
stringArray = array.astype(str)

print("Original Array:", array)
print("Float Array:", floatArray)
print("Complex Array:", complexArray)
print("Boolean Array:", boolArray)
print("String Array:", stringArray)

Output

Original Array: [0 1 2 3 4 5]
Float Array: [0. 1. 2. 3. 4. 5.]
Complex Array: [0.+0.j 1.+0.j 2.+0.j 3.+0.j 4.+0.j 5.+0.j]
Boolean Array: [False  True  True  True  True  True]
String Array: ['0' '1' '2' '3' '4' '5']

Using Optional order Argument in astype()

The order argument specifies the order in which the array elements are stored in memory.

The order can be:

  • 'C' - flattens the elements row-wise (in C-style order)
  • 'F' - flattens the elements column-wise (in Fortran-style order)
  • 'A' - tries to preserve the original array's order, otherwise defaults to C-order.
  • 'K' - flattens the elements in the order they occur in memory, and uses C-order by default.

Let's see an example.

import numpy as np

array1 = np.array([[1, 2, 3], [4, 5, 6]])

# list of orders
orders = ['C', 'F', 'A', 'K']

# convert integer array to floating point in different orders
for item in orders:
floatArray = array1.astype(float, order=item)
print(item, 'order Array:\n', floatArray) print('C_CONTIGUOUS :', floatArray.flags['C_CONTIGUOUS']) print('F_CONTIGUOUS :', floatArray.flags['F_CONTIGUOUS'],'\n')

Output

C order Array:
[[1. 2. 3.]
 [4. 5. 6.]]
C_CONTIGUOUS : True
F_CONTIGUOUS : False 

F order Array:
 [[1. 2. 3.]
 [4. 5. 6.]]
C_CONTIGUOUS : False
F_CONTIGUOUS : True 

A order Array:
 [[1. 2. 3.]
 [4. 5. 6.]]
C_CONTIGUOUS : True
F_CONTIGUOUS : False 

K order Array:
 [[1. 2. 3.]
 [4. 5. 6.]]
C_CONTIGUOUS : True
F_CONTIGUOUS : False 

Using Optional casting Argument in astype()

The casting argument specifies the casting behavior when converting data types.

The casting can be:

  • 'no' - data types should not be cast at all
  • 'equiv' - only byte-order changes are allowed
  • 'safe' - only casts which can preserve values are allowed
  • 'same_kind' - only safe casts or casts within a kind are allowed
  • 'unsafe' - any data conversions may be done

Let's see an example.

import numpy as np

# original array of integers with big-endian byte order
array1 = np.array([1, 2, 3, 4, 5], dtype='>i4')

# casting with 'no' doesn't allow casting to any other data type array2 = array1.astype(array1.dtype, casting='no') # casting with 'equiv' allows casting to equivalent data types array3 = array1.astype('<i4', casting='equiv') #cCasting with 'safe' allows casting to safe data types preserving precision array4 = array1.astype(np.float64, casting='safe') # casting with 'same_kind' allows casting to data types of the same kind array5 = array1.astype(np.int32, casting='same_kind') # casting with 'unsafe' allows casting to any data type without checks array6 = array1.astype(str, casting='unsafe')
print("Array with 'no' casting:", array2) print("Array with 'equiv' casting:", array3) print("Array with 'safe' casting:", array4) print("Array with 'same_kind' casting:", array5) print("Array with 'unsafe' casting:", array6)

Output

Array with 'no' casting: [1 2 3 4 5]
Array with 'equiv' casting: [1 2 3 4 5]
Array with 'safe' casting: [1. 2. 3. 4. 5.]
Array with 'same_kind' casting: [1 2 3 4 5]
Array with 'unsafe' casting: ['1' '2' '3' '4' '5']

Using Optional subok Argument in astype()

The subok argument specifies whether to use subclass instances, if available, for the returned array.

The subok can be:

  • True- resulting array maintains the subclass
  • False - resulting array doesn't maintain the subclass
import numpy as np

# define a custom subclass of ndarray
class CustomArray(np.ndarray):
    pass

# create a custom subclass array
array = np.array([1, 2, 3]).view(CustomArray)

# convert the array to float, preserving the subclass floatArray1 = array.astype(float, subok=True) # convert the array to float, without preserving the subclass floatArray2 = array.astype(float, subok=False)
print("Original Array Type:", type(array)) print("Float Array1 Type:", type(floatArray1)) print("Float Array2 Type:", type(floatArray2))

Output

Original Array Type: <class '__main__.CustomArray'>
Float Array1 Type: <class '__main__.CustomArray'>
Float Array2 Type: <class 'numpy.ndarray'>