How to convert a float numpy.ndarray to list? [duplicate]

Question

here is my code

import numpy
a = numpy.arange(0.5, 1.5, 0.1, dtype=numpy.float64)
print(a)
print(a.tolist())

>>>[0.5 0.6 0.7 0.8 0.9 1.  1.1 1.2 1.3 1.4]
>>>[0.5, 0.6, 0.7, 0.7999999999999999, 0.8999999999999999, 0.9999999999999999, 1.0999999999999999, 1.1999999999999997, 1.2999999999999998, 1.4]

When trying to convert numpy array to list getting this problem with value like 0.7999999999999999 in place of 0.8

please help me to convert numpy array to normal python list without losing the decimal value.

Answer 1

import numpy

a = numpy.arange(0.5, 1.5, 0.1, dtype=numpy.float64)

print(a)

a = numpy.round(a.astype(numpy.float64), 1)

print (a)

print (a.tolist())

Output:

[0.5 0.6 0.7 0.8 0.9 1.  1.1 1.2 1.3 1.4]
[0.5 0.6 0.7 0.8 0.9 1.  1.1 1.2 1.3 1.4]
[0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.4]

You can use numpy.round() with astype() method to round your float, getting something consistant upon tolist() conversion.

Answer 2

By the conversion via .tolist() you're not gaining or loosing any precision. You're just converting to another data type which chooses to represent itself differently. You seem to be thinking that by the conversion it turns the 0.8 into 0.7999999999999999, but the situation is, it has never been 0.8. This is due to limited floating point precision. You can verify yourself:

>>> import numpy
>>> a = numpy.arange(0.5, 1.5, 0.1, dtype=numpy.float64)
>>> a
array([0.5, 0.6, 0.7, 0.8, 0.9, 1. , 1.1, 1.2, 1.3, 1.4])
>>> a[3]
0.7999999999999999

The situation is that that every type can decide how its instances are represented via __repr__ or __str__. In this specific case np.ndarray decides to show a rounded version of its elements for more clarity. This can be controlled via numpy.set_printoptions. This function only affects how arrays are displayed:

>>> numpy.set_printoptions(floatmode='unique')
>>> a
array([0.5               , 0.6               , 0.7               ,
       0.7999999999999999, 0.8999999999999999, 0.9999999999999999,
       1.0999999999999999, 1.1999999999999997, 1.2999999999999998,
       1.4               ])

A note on np.arange from the docs:

When using a non-integer step, such as 0.1, the results will often not be consistent. It is better to use numpy.linspace for these cases.

Here are the results for linspace:

>>> np.linspace(0.5, 1.5, 11)
array([0.5               , 0.6               , 0.7               ,
       0.8               , 0.9               , 1.                ,
       1.1               , 1.2000000000000002, 1.3               ,
       1.4               , 1.5               ])