Importing numpy statement in script

Question

I am a matlab coder and have dabbled in and out of Python for various things. One very basic concept/thing in python which I have not got to the bottom of relates to the from and import statements.

When i import bitarray from the bitarray module (am I correct saying this?) using example 1) I just call it's methods when I want to use them in the script that follows. However, when i import numpy I cannot use the code shown in 2) as it throws an error, i have to use the statement in 3) then type np. before using any of its methods, which makes sense. I can also import bitarray using similer syntax which i have shown in 4) below....... why does 2) not work when 1) does? I am praying for a simple explanation :) .....

1)

from bitarray import bitarray

2)

from numpy import numpy

3)

import numpy as np

4)

import bitarray as pp

Answer 1

This is due to the different package structures between bitarray and numpy ; specifically, one contains an object with the same name as the module itself, the other doesn't.

The bitarray module contains a class called bitarray , (similarly, for the example, the datetime module in the standard library contains the datetime class) . Hence the following are equivalent:

from bitarray import bitarray
ba = bitarray(...)

and

import bitarray
ba = bitarray.bitarray(...)

In the former we import the class bitarray directly into the current namespace from the module; in the latter we import the whole module as bitarray and need to access the class by specifying the module namespace.

By contrast, numpy does not contain an object named numpy , so

from numpy import numpy 

doesn't work.

---

The as is separate; that simply allows you to provide an alias for whatever you're importing, so any of the following will work fine:

import numpy as np
a = np.array(...)

from bitarray import bitarray as pp
b = pp(...)

import bitarray as pp
b = pp.bitarray(...)

---

As mauve comments above, you can use the * "wildcard import" to import everything, so:

from numpy import *
from bitarray import *

will import every top-level name in both modules into the current namespace. Afterwards, eg

a = array(...)
b = bitarray(...)

will both work fine. However, this is not recommended as it is generally a bad idea - for example, numpy contains a definition of all which will shadow the built-in function of the same name if you import everything.

Also, it makes it much more difficult for anyone reading your code (including a later version of you!) to figure out where the classes and functions being used have come from. Compare:

from numpy import *
from random import *

a = choice([1, 2, 3]) # which module did choice come from?
b = all(x == y for x in l) # is this numpy's all or the built-in?

with the much clearer:

import numpy as np
import random

a = random.choice([1, 2, 3]) # clearly from random
b = all(x == y for x in l) # must be the built-in, not np.all()

Answer 2

1) If bitarray is a function or class name in the library bitarray then it is perfectly valid.

2) This doesn't make sense because numpy doesn't have a function named numpy .

3) This is fine because you are importing a library and assigning it an alias.

4) This is fine as well because again you are simply importing under an alias.