Functions in Python provide organized, reusable and modular code to perform a set of specific actions. Functions simplify the coding process, prevent redundant logic, and make the code easier to follow. This topic describes the declaration and utilization of functions in Python.
Python has many built-in functions like
len(). Besides built-ins you can also create your own functions to do more specific jobs—these are called user-defined functions.
- def function_name(arg1, ... argN, *args, kw1, kw2=default, ..., **kwargs): statements
- lambda arg1, ... argN, *args, kw1, kw2=default, ..., **kwargs: expression
|arg1, ..., argN||Regular arguments|
|*args||Unnamed positional arguments|
|kw1, ..., kwN||Keyword-only arguments|
|**kwargs||The rest of keyword arguments|
5 basic things you can do with functions:
Assign functions to variables
Define functions within other functions (Nested functions )
Functions can return other functions
Functions can be passed as parameters to other functions
Inner functions have access to the enclosing scope (Closure )
- More on functions and decorators: https://www.thecodeship.com/patterns/guide-to-python-function-decorators/
Defining and calling simple functions
def statement is the most common way to define a function in python. This statement is a so called single clause compound statement with the following syntax:
function_name is known as the identifier of the function. Since a function definition is an executable statement its execution binds the function name to the function object which can be called later on using the identifier.
parameters is an optional list of identifiers that get bound to the values supplied as arguments when the function is called. A function may have an arbitrary number of arguments which are separated by commas.
statement(s) – also known as the function body – are a nonempty sequence of statements executed each time the function is called. This means a function body cannot be empty, just like any indented block.
Here’s an example of a simple function definition which purpose is to print
Hello each time it’s called:
Now let’s call the defined
That’s an other example of a function definition which takes one single argument and displays the passed in value each time the function is called:
After that the
greet_two() function must be called with an argument:
Also you can give a default value to that function argument:
Now you can call the function without giving a value:
You'll notice that unlike many other languages, you do not need to explicitly declare a return type of the function. Python functions can return values of any type via the
return keyword. One function can return any number of different types!
As long as this is handled correctly by the caller, this is perfectly valid Python code.
A function that reaches the end of execution without a return statement will always return
As mentioned previously a function definition must have a function body, a nonempty sequence of statements. Therefore the
pass statement is used as function body, which is a null operation – when it is executed, nothing happens. It does what it means, it skips. It is useful as a placeholder when a statement is required syntactically, but no code needs to be executed.
Argument passing and mutability
First, some terminology:
- argument (actual parameter): the actual variable being passed to a function;
- parameter (formal parameter): the receiving variable that is used in a function.
In Python, arguments are passed by assignment (as opposed to other languages, where arguments can be passed by value/reference/pointer).
Mutating a parameter will mutate the argument (if the argument's type is mutable).
Reassigning the parameter won’t reassign the argument.
In Python, we don’t really assign values to variables, instead we bind (i.e. assign, attach) variables (considered as names) to objects.
- Immutable: Integers, strings, tuples, and so on. All operations make copies.
- Mutable: Lists, dictionaries, sets, and so on. Operations may or may not mutate.
Closures in Python are created by function calls. Here, the call to
makeInc creates a binding for
x that is referenced inside the function
inc. Each call to
makeInc creates a new instance of this function, but each instance has a link to a different binding of
Notice that while in a regular closure the enclosed function fully inherits all variables from its enclosing environment, in this construct the enclosed function has only read access to the inherited variables but cannot make assignments to them
Python 3 offers the
nonlocal statement (Nonlocal Variables ) for realizing a full closure with nested functions.
Defining a function with an arbitrary number of arguments
Arbitrary number of positional arguments:
Defining a function capable of taking an arbitrary number of arguments can be done by prefixing one of the arguments with a
You can't provide a default for
args, for example
func(*args=[1, 2, 3]) will raise a syntax error (won't even compile).
You can't provide these by name when calling the function, for example
func(*args=[1, 2, 3]) will raise a
But if you already have your arguments in an array (or any other
Iterable), you can invoke your function like this:
These arguments (
*args) can be accessed by index, for example
args will return the first argument
Arbitrary number of keyword arguments
You can take an arbitrary number of arguments with a name by defining an argument in the definition with two
* in front of it:
You can't provide these without names, for example
func(1, 2, 3) will raise a
kwargs is a plain native python dictionary. For example,
args['value1'] will give the value for argument
value1. Be sure to check beforehand that there is such an argument or a
KeyError will be raised.
You can mix these with other optional and required arguments but the order inside the definition matters.
The positional/keyword arguments come first. (Required arguments).
Then comes the arbitrary
*arg arguments. (Optional).
Then keyword-only arguments come next. (Required).
Finally the arbitrary keyword
**kwargs come. (Optional).
arg1must be given, otherwise a
TypeErroris raised. It can be given as positional (
func(10)) or keyword argument (
kwarg1must also be given, but it can only be provided as keyword-argument:
kwarg2are optional. If the value is to be changed the same rules as for
arg1(either positional or keyword) and
kwarg1(only keyword) apply.
*argscatches additional positional parameters. But note, that
arg2must be provided as positional arguments to pass arguments to
func(1, 1, 1, 1).
**kwargscatches all additional keyword parameters. In this case any parameter that is not
kwarg2. For example:
- In Python 3, you can use
*alone to indicate that all subsequent arguments must be specified as keywords. For instance the
math.isclosefunction in Python 3.5 and higher is defined using
def math.isclose (a, b, *, rel_tol=1e-09, abs_tol=0.0), which means the first two arguments can be supplied positionally but the optional third and fourth parameters can only be supplied as keyword arguments.
Python 2.x doesn't support keyword-only parameters. This behavior can be emulated with
Note on Naming
The convention of naming optional positional arguments
args and optional keyword arguments
kwargs is just a convention you can use any names you like but it is useful to follow the convention so that others know what you are doing, or even yourself later so please do.
Note on Uniqueness
Any function can be defined with none or one
*args and none or one
**kwargs but not with more than one of each. Also
*args must be the last positional argument and
**kwargs must be the last parameter. Attempting to use more than one of either will result in a Syntax Error exception.
Note on Nesting Functions with Optional Arguments
It is possible to nest such functions and the usual convention is to remove the items that the code has already handled but if you are passing down the parameters you need to pass optional positional args with a
* prefix and optional keyword args with a
** prefix, otherwise args with be passed as a list or tuple and kwargs as a single dictionary. e.g.:
Defining a function with arguments
Arguments are defined in parentheses after the function name:
The function name and its list of arguments are called the signature of the function. Each named argument is effectively a local variable of the function.
When calling the function, give values for the arguments by listing them in order
or specify them in any order using the names from the function definition:
Defining a function with multiple arguments
One can give a function as many arguments as one wants, the only fixed rules are that each argument name must be unique and that optional arguments must be after the not-optional ones:
When calling the function you can either give each keyword without the name but then the order matters:
Or combine giving the arguments with name and without. Then the ones with name must follow those without but the order of the ones with name doesn't matter:
Defining a function with optional arguments
Optional arguments can be defined by assigning (using
=) a default value to the argument-name:
Calling this function is possible in 3 different ways:
Mutable types (
set, etc.) should be treated with care when given as default attribute. Any mutation of the default argument will change it permanently. See Defining a function with optional mutable arguments.
Defining a function with optional mutable arguments
There is a problem when using optional arguments with a mutable default type (described in Defining a function with optional arguments), which can potentially lead to unexpected behaviour.
This problem arises because a function's default arguments are initialised once, at the point when the function is defined, and not (like many other languages) when the function is called. The default values are stored inside the function object's
__defaults__ member variable.
For immutable types (see Argument passing and mutability) this is not a problem because there is no way to mutate the variable; it can only ever be reassigned, leaving the original value unchanged. Hence, subsequent are guaranteed to have the same default value. However, for a mutable type, the original value can mutate, by making calls to its various member functions. Therefore, successive calls to the function are not guaranteed to have the initial default value.
Note: Some IDEs like PyCharm will issue a warning when a mutable type is specified as a default attribute.
If you want to ensure that the default argument is always the one you specify in the function definition, then the solution is to always use an immutable type as your default argument.
A common idiom to achieve this when a mutable type is needed as the default, is to use
None (immutable) as the default argument and then assign the actual default value to the argument variable if it is equal to
Forcing the use of named parameters
All parameters specified after the first asterisk in the function signature are keyword-only.
In Python 3 it's possible to put a single asterisk in the function signature to ensure that the remaining arguments may only be passed using keyword arguments.
Iterable and dictionary unpacking
Functions allow you to specify these types of parameters: positional, named, variable positional, Keyword args (kwargs). Here is a clear and concise use of each type.
Lambda (Inline/Anonymous) Functions
lambda keyword creates an inline function that contains a single expression. The value of this expression is what the function returns when invoked.
Consider the function:
which, when called as:
This can be written as a lambda function as follows:
See note at the bottom of this section regarding the assignment of lambdas to variables. Generally, don't do it.
This creates an inline function with the name
greet_me that returns
Hello. Note that you don't write
return when creating a function with lambda. The value after
: is automatically returned.
Once assigned to a variable, it can be used just like a regular function:
lambdas can take arguments, too:
returns the string:
They can also take arbitrary number of arguments / keyword arguments, like normal functions.
lambdas are commonly used for short functions that are convenient to define at the point where they are called (typically with
For example, this line sorts a list of strings ignoring their case and ignoring whitespace at the beginning and at the end:
Sort list just ignoring whitespaces:
Examples with numerical lists:
One can call other functions (with/without arguments) from inside a lambda function.
This is useful because
lambda may contain only one expression and by using a subsidiary function one can run multiple statements.
Bear in mind that PEP-8 (the official Python style guide) does not recommend assigning lambdas to variables (as we did in the first two examples):
Always use a def statement instead of an assignment statement that binds a lambda expression directly to an identifier.
The first form means that the name of the resulting function object is specifically
finstead of the generic
<lambda>. This is more useful for tracebacks and string representations in general. The use of the assignment statement eliminates the sole benefit a lambda expression can offer over an explicit
defstatement (i.e. that it can be embedded inside a larger expression).
Functions in python are first-class objects. They can be defined in any scope
Functions capture their enclosing scope can be passed around like any other sort of object
There is a limit to the depth of possible recursion, which depends on the Python implementation. When the limit is reached, a RuntimeError exception is raised:
It is possible to change the recursion depth limit by using
sys.setrecursionlimit(limit) and check this limit by
From Python 3.5, the exception is a
RecursionError, which is derived from
A recursive function is a function that calls itself in its definition.
For example the mathematical function, factorial, defined by
factorial(n) = n*(n-1)*(n-2)*...*3*2*1. can be programmed as
the outputs here are:
as expected. Notice that this function is recursive because the second
return factorial(n-1), where the function calls itself in its definition.
Some recursive functions can be implemented using lambda, the factorial function using lambda would be something like this:
The function outputs the same as above.
Recursive Lambda using assigned variable
One method for creating recursive lambda functions involves assigning the function to a variable and then referencing that variable within the function itself. A common example of this is the recursive calculation of the factorial of a number - such as shown in the following code:
Description of code
The lambda function, through its variable assignment, is passed a value (4) which it evaluates and returns 1 if it is 0 or else it returns the current value (
i) * another calculation by the lambda function of the value - 1 (
i-1). This continues until the passed value is decremented to 0 (
return 1). A process which can be visualized as:
Returning values from functions
return a value that you can use directly:
or save the value for later use:
or use the value for any operations:
return is encountered in the function the function will be exited immediately and subsequent operations will not be evaluated:
You can also
return multiple values (in the form of a tuple):
A function with no
return statement implicitly returns
None. Similarly a function with a
return statement, but no return value or variable returns