Numpy (data type)

Common data types

Python There are relatively few native data types , bool、int、float、str etc. . This is convenient in applications that do not need to care about all the ways data is represented in a computer . However , For scientific computing , More control is often required . To distinguish numpy At the end of the names of these types “_”.

The following table lists the common numpy Basic types .

Create data types

numpy The numerical type of is actually dtype Instance of object .

class dtype(object):
    def __init__(self, obj, align=False, copy=False):
        pass

Each built-in type has a unique character code that defines it , as follows ：

【 example 】

import numpy as np

a = np.dtype('b1')
print(a.type)  # <class 'numpy.bool_'>
print(a.itemsize)  # 1

a = np.dtype('i1')
print(a.type)  # <class 'numpy.int8'>
print(a.itemsize)  # 1
a = np.dtype('i2')
print(a.type)  # <class 'numpy.int16'>
print(a.itemsize)  # 2
a = np.dtype('i4')
print(a.type)  # <class 'numpy.int32'>
print(a.itemsize)  # 4
a = np.dtype('i8')
print(a.type)  # <class 'numpy.int64'>
print(a.itemsize)  # 8

a = np.dtype('u1')
print(a.type)  # <class 'numpy.uint8'>
print(a.itemsize)  # 1
a = np.dtype('u2')
print(a.type)  # <class 'numpy.uint16'>
print(a.itemsize)  # 2
a = np.dtype('u4')
print(a.type)  # <class 'numpy.uint32'>
print(a.itemsize)  # 4
a = np.dtype('u8')
print(a.type)  # <class 'numpy.uint64'>
print(a.itemsize)  # 8

a = np.dtype('f2')
print(a.type)  # <class 'numpy.float16'>
print(a.itemsize)  # 2
a = np.dtype('f4')
print(a.type)  # <class 'numpy.float32'>
print(a.itemsize)  # 4
a = np.dtype('f8')
print(a.type)  # <class 'numpy.float64'>
print(a.itemsize)  # 8

a = np.dtype('S')
print(a.type)  # <class 'numpy.bytes_'>
print(a.itemsize)  # 0
a = np.dtype('S3')
print(a.type)  # <class 'numpy.bytes_'>
print(a.itemsize)  # 3

a = np.dtype('U3')
print(a.type)  # <class 'numpy.str_'>
print(a.itemsize)  # 12

Data type information

Python The floating-point number of is usually 64 Bit floating point , Almost equal to np.float64.

NumPy and Python The behavior of integer types differs significantly in terms of integer overflow , And NumPy Different ,Python Of int It's flexible . It means Python Integers can be extended to accommodate any integer without overflowing .

• Machine limits for integer types .

class iinfo(object):
    def __init__(self, int_type):
        pass
    def min(self):
        pass
    def max(self):
        pass

【 example 】

import numpy as np

ii16 = np.iinfo(np.int16)
print(ii16.min)  # -32768
print(ii16.max)  # 32767

ii32 = np.iinfo(np.int32)
print(ii32.min)  # -2147483648
print(ii32.max)  # 2147483647

• Machine limits for floating point types .

class finfo(object):
    def _init(self, dtype):

【 example 】

import numpy as np

ff16 = np.finfo(np.float16)
print(ff16.bits)  # 16
print(ff16.min)  # -65500.0
print(ff16.max)  # 65500.0
print(ff16.eps)  # 0.000977

ff32 = np.finfo(np.float32)
print(ff32.bits)  # 32
print(ff32.min)  # -3.4028235e+38
print(ff32.max)  # 3.4028235e+38
print(ff32.eps)  # 1.1920929e-07