Pythonic code: the with statement

This post continues a series on “Pythonic” code. Pythonic code is code that fits well with the design of the Python language. Previously, I wrote about list comprehensions as a powerful way to manipulate Python’s list data structure. This post will cover the with statement.

  1. The list comprehension
  2. The with statement
  3. The property decorator
  4. Built-in functions
  5. Using the standard library
  6. Leveraging packages

The with statement

One task that you are likely to encounter while programming Python is the need to open a file. That file might contain tables of data or pictures of kittens. Whatever you find yourself doing, you’ll come across the open function. Let’s work through a thought experiment which can help explain why you should use with.

If you’re brand new to Python, you might open a file like so:

f = open('kitteh.jpg', 'rb')
cat_pic = f.read()
# Do other stuff with the cat picture.

After speaking with a friend with more Python experience than you, you learn that you’re supposed to close files or else the operating system will eventually run into trouble (because it can only track a limited number of open files). You rewrite your code:

f = open('kitteh.jpg', 'rb')
cat_pic = f.read()
# Do other stuff with the cat picture.
f.close()

Then you learn that errors can happen. Being a good developer, you attempt to handle any errors.

try:
    f = open('kitteh.jpg', 'rb')
    cat_pic = f.read()
    # Do other stuff with the cat picture.
    f.close()
except:
    print('oops, something went wrong.')

Your friend tells you that you’ve added a bug. What? How could that be? She tells you that an error can happen before the file is closed. You go read more Python documentation and learn about finally. The code is reworked again to look like:

try:
    f = open('kitteh.jpg', 'rb')
    cat_pic = f.read()
    # Do other stuff with the cat picture.
except:
    print('oops, something went wrong.')
finally:
    f.close()

This code is not stellar. If you had to write 200 lines of extra code for “doing other stuff,” then there is a lot of distance between opening and closing the file. Viewed another way, there is a lot space between setting something up and tearing it down later.

This is a perfect place to use with. Let’s restate the code with the with statement.

with open('kitteh.jpg', 'rb') as f:
    cat_pic = f.read()
try:
    # Do other stuff with the cat picture.
except:
    print('oops, something went wrong.')

At first, you might be suspect of this code. Where did the close call go? The with statement used an extra concept called a context manager. Context managers are designed to handle setup and tear down for anything that needs it. A context manager is some code that implements an __enter__ and __exit__ method.

When open is used with a with statement, a special context manager is called. After the scope of the with block passes (i.e., reading the file content into cat_pic), the interpreter will execute an __exit__ method on the context manager. The open context manager will close the file in __exit__. All of this work is neatly tucked away from the developer. You have the guarantee that the file gets closed and do not have to do that work on your own.

The open context manager is probably the most common usage of the with statement. Other uses of with can include threading locks, timers, or even nicer interfaces for unit testing exceptions. Finally, Python let’s you create your own context managers. Check out contextlib for more info. I’ve included this example to give you a quick idea in action.

>>> from contextlib import contextmanager
>>> @contextmanager
... def praise():
...     print('You can do it.')
...     yield
...     print('You made it.')
...
>>> with praise():
...     print('I am trying to code.')
...
You can do it.
I am trying to code.
You made it.

The with statement is another valuable tool for your Python programmer toolbelt. The key thing to remember is that it can help you clean up any code where you need to set things up or tear things down.