Fixing lost metadata

As it turns out, function metadata like its name and its docstring can be accessed through certain attributes:

def my_function():
    """This is the docstring of a function."""
    ...

print(my_function.__name__)  # my_function
print(my_function.__doc__)  # This is the docstring of a function.

Therefore, to make it so that a decorator produces a wrapper that looks like the original function, all you have to do is patch up these attributes __name__, __doc__, and others, and the wrapper function will look like the original function!

You could do this by hand in all of your decorators before returning the wrapper:

import time

def timed_factory(f):
    def timed_f(*args, **kwargs):
        start_time = time.perf_counter()
        result = f(*args, **kwargs)
        end_time = time.perf_counter()
        return result, end_time - start_time

    # Patch some metadata:
    timed_f.__name__ = f.__name__
    timed_f.__doc__ = f.__doc__
    return timed_f

And this is a start, indeed. Try applying this version of the decorator timed_factory to the function fibonacci as shown below:

@timed_factory
def fibonacci(n):
    """Computes the recurrence relation F(n) = F(n - 1) + F(n - 2)."""
    if n < 2:
        return n
    return fibonacci(n - 1) + fibonacci(n - 2)

Now, running help(fibonacci) shows a better help screen than before, although it isn’t perfect yet:

Help on function fibonacci in module __main__:

fibonacci(*args, **kwargs)
    Computes the recurrence relation F(n) = F(n - 1) + F(n - 2).
 Help on timed_factory.<locals>.timed_f line 1/4 (END) (press h for help or q to quit)

Note how the signature of fibonacci is shown to be (*args, **kwargs) and how the final line still says you are getting help regarding the function timed_factory.<locals>.timed_f.

The best way to patch all of the metadata is by using another decorator called functools.wraps! This decorator from the standard library must be used inside your decorator, around the wrapper function, and it accepts the original function as an argument, like so:

import functools
import time

def timed_factory(f):
    @functools.wraps(f)  # <--
    def timed_f(*args, **kwargs):
        start_time = time.perf_counter()
        result = f(*args, **kwargs)
        end_time = time.perf_counter()
        return result, end_time - start_time

    return timed_f

The line @functools.wraps(f) says “the function timed_f will replace the function f, so transfer f’s metadata to timed_f”. And this was the last ingredient you needed! Now, you have all the tools you might require to write any decorators you might need.