Pointers can be a complicated concept in programming languages. C and C++ are known for having pointers, but what about Python? It does! Let’s explore some exciting things about pointers in Python.
Table Of Contents
What are Pointers?
Pointers in Python?
Objects
3.1. Immutable Objects
3.2. Mutable Objects
Pythons’ Object Model
4.1. C Variables
4.2. Python Names
Faking Pointers in Python
5.1. Using Mutable Objects
Pointers with ctypes
Pointers are variables which store the address of other variables. It’s a data type which stores the address of other data types. If you are familiar with C or C++, then you are familiar with what pointers are?
Whenever we create a variable or an object in a programming language, it’s stored in a particular CPU address. Whenever we output the data, it pulls from that address. Pointers are used to store the addresses and for memory management. But, at times pointers can crash our programs. Let’s get into the details.
Python doesn’t have any pointers concept. Why doesn’t Python speak about pointers? The reason is unknown. Maybe because of the difficulty level of pointers which is against the Zen of Python. Python focuses on its simplicity instead of speed. We can implement pointers in Python with the help of other objects.
We have to first understand some concepts in Python before diving into the pointers. Let’s first see what objects are? And types in it.
Everything in Python is an object.
If you are new to Python, run the following programs as proof for the above statement.
## int
print(f'int:- {isinstance(int, object)}')
## str
print(f'str:- {isinstance(str, object)}')
## bool
print(f'bool:- {isinstance(False, object)}')
## list
print(f'list:- {isinstance(list(), object)}')
## function
def sample():
pass
print(f'function:- {isinstance(sample, object)}')
int:- True
str:- True
bool:- True
list:- True
function:- True
As you see, everything in Python is an object. Each object in Python consists of three parts.
It deals with the memory in the CPU. It represents the number of Python variables referring to a memory location.
It refers to the kind of object like int, float, string, etc…,
It’s the actual value of an object stored in the memory.
Objects are of two types Immutable and Mutable. Knowing about these objects will clear the first step of our pointers. Let’s first see the immutable and mutable.
We can’t change the immutable objects once we create them. Most of the commonly used data types in Python are immutable. Let’s see what that means.
We can test whether an object is immutable or mutable by using id() and is.
int is an immutable object. Let’s see with an example.
a = 7
id(a)
1744268544
We have assigned 7 to the a. We can’t modify the value of x in 1744268544 memory. If we try to change it, it will create a new object. Let’s see by adding one to the a.
a += 1
id(a)
1744268576
The address of a is changed. That means the object we created first, is now referring to the new address.
b = a b is a
True
When we assign a to b, Python doesn’t create a new object. It merely made reference of b to a value. It saves memory.
We can change the mutable objects even after creation. Python doesn’t create new objects when we modify a mutable object. Let’s see with some examples.
## list is a mutable object
nums = [1, 2, 3, 4, 5]
print("---------Before Modifying------------")
print(id(nums))
print()
## modifying element
nums[0] += 1
print("-----------After Modifying Element-------------")
print(id(nums))
print()
## appending an element
nums.append(6)
print("-----------After Appending Element-------------")
print(id(nums))
---------Before Modifying------------
2197288429320
-----------After Modifying Element-------------
2197288429320
-----------After Appending Element-------------
2197288429320
Even after performing some operations on the list, the address doesn’t change. Because the list is a mutable object. The same thing occurs when we perform other mutable objects like set or dict.
Now you know the difference between immutable and mutable objects. This makes the upcoming concepts easy to understand.
Python variables are different from C. First, we will learn how variables work in C for better understanding of pointers.
C variables memory management is entirely different from Python. Let’s see how to define a variable in C and the steps it goes through when executed.
// C syntax not Python
int a = 918
Execution steps of the above statement.
If we illustrate the memory, it may look like the following.
Here, a has a memory location of 0x531. If we update the value of a. The address location of a doesn’t change.
a = 849
If we see the location of a didn’t change in C programming language, the variable is not just a name for the value. It’s a memory location itself. So, we are overwriting the value of a memory location directly. It’s completely different from Python.
If we want to assign a to another variable, then C creates a new memory location, unlike Python. The following code assigns a to a new variable b.
int b = a
Notice that the address of b has changed. It’s because the C creates a new memory location for every variable we create. C creates a new memory location for b and copies the values of a and assigns them to b.
This how C variables work. Now, let’s move on to find out how Python variables work.
Generally, variables in the Python are called names.
## code in python
a = 918
The above code will undergo the following steps during execution.
In memory, a may look like the following.
a reference memory location illustration
a refers to the above PyObject in the memory. It’s completely different from C variables. a is not a memory location as it was in C variables.
a = 98
The above statement undergoes the following steps during execution.
In memory a refers to the new PyObject. So, the old PyObject refer count will be 0.
New PyObject
a refers to the immediate below PyObject
Old PyObject
The old PyObject reference count is 0. Because the variable a refers to the new PyObject.
What happens when we assign a to a new variable b? Let’s see…
b = a
Python doesn’t create a new PyObject for the variable b. Python makes the reference count of a’s PyObject to two. Those two variables are a and b.
a and b refers the same PyObject
To see whether a and b referred to the same memory location or not, run the following code.
a is b
True
is returns True both variables refer to the same PyObject. If we modify the value of b Python creates a new PyObject for b because integers are immutable in Python. Now, you are familiar with the Pythons’ object model.
Developers of Python didn’t include pointers in Python. But still, we can stimulate the pointers in Python using different methods.
Let’s write a small function in C, which takes a pointer as an argument and increments the value of variables present in that memory location.
// C code
void increment(int *p) {
*p += 1;
}
The increment function takes a pointer and increments the value of the pointer referring variable. Let’s write the main function.
// C code main function
void main() {
int a = 759;
printf("%d\n", a);
increment(&a); // & operator used to extract the adress of the variable
printf("%d\n", a);
}
// Output
759
769
Now, we are going to implement the same behavior in Python using mutable objects. Let’s see…
We can replicate the above program using mutable objects in Python. Let’s see how we can do this.
## function
def increment(p):
p[0] += 1
if __name__ == "__main__":
a = [759]
increment(a)
print(a[0])
760
We have achieved the same result without changing the memory location of the variable. The function increment takes a list and increments the first element. That’s it. We have achieved the same because lists are mutable. If we try to pass a tuple as an argument to the increment function, we will get an error.
if __name__ == "__main__":
a = (759,)
increment(a)
print(a[0])
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-32-c3bce5df94da> in <module>()
1 if __name__ == "__main__":
2 a = (759,)
----> 3 increment(a)
4 print(a[0])
<ipython-input-30-042f4c577fd9> in increment(p)
1 ## function
2 def increment(p):
----> 3 p[0] += 1
4
5 if __name__ == "__main__":
TypeError: 'tuple' object does not support item assignment
These are not real pointers like C. All we did was replicate the behavior of pointers in Python.
Using ctypes module, we can create real pointers in Python. First, we will compile a .c file containing functions which use pointers and store it. Let’s write the following function into a .c file
void increment(int *p) {
*p += 1;
}
Let’s assume the file name is increment.c and run the following commands.
$ gcc -c -Wall -Werror -fpic increment.c
$ gcc -shared -o libinc.so increment.o
The first command compiles increment.c into an object called increment.o. The second takes the object file and produces libinc.so to work with ctypes.
import ctypes
## make sure the libinc.so present in the same directory as this program
lib = ctypes.CDLL("./libinc.so")
lib.increment
## output
<_FuncPtr object at 0x7f46bf6e0750>
The ctypes.CDLL returns a shared object called libinc.so. We define increment() function in the libinc.so shared object. If we want to pass a pointer to the functions we define in a shared object, then we have to specify it using the ctypes.
inc = lib.increment
## defining the argtypes
inc.argtypes = [ctypes.POINTER(ctypes.c_int)]
Now, if we try to call the function using a different type, we will get an error.
inc(5)
## output
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ctypes.ArgumentError: argument 1: <class 'TypeError'>: expected LP_c_int instance instead of int
We got an error saying that the function wants a pointer. ctypes has a way to pass the C to the functions.
a = ctypes.c_int(5)
a is a C variable. ctypes has a method called byref() which allows you to pass the variable reference.
inc(ctypes.byref(a))
a
## output
c_int(6)
Now, we have an incremented value in the a.
Now, you have a better understanding of Pythons’ objects and pointers. Pointers are not present in Python. But, we implemented the same behavior with mutable objects. The Pointer we implemented with ctypes are real C pointers.
#python