Creating Magic Functions in IPython — Part 2

Image for post
Image for post

One of the new features that came in Python in version 3.5 are type hints. Some people like them, some people don’t (which is probably true for every new feature in every programming language). The nice thing about Python type hints is that they are not mandatory. If you don’t like them — don’t use them. For fast prototyping or a project that you are maintaining yourself, you are probably fine without them. But for a large code base, with plenty of legacy code maintained by multiple developers — type hints can be tremendously helpful!

Speaking of type hints — if you are not using them yet and your project is still on Python 2, migration to Python 3 (that you will have to go through soon) is a perfect opportunity to introduce them! There are many resources around the web on how to migrate a Python 2 code base to Python 3, but if you need some help — I can help you and your company. You can contact me to learn more about my experience with migrating large projects to Python 3.

As you are probably starting to guess, our cell magic function will check types for a block of code. Why? Well, with IPython, you can quickly prototype some code, tweak it and save it to a file using the %save or %%writefile magic functions (or simply copy and paste it, if it’s faster for you). But, at the time of writing this article, there is no built-in type checker in Python. The mypy library is a de facto static type checker, but it’s still an external tool that you run from shell (mypy So let’s make a helper that will allow us to type check Python code directly in IPython!

This is how we expect it to work:

To achieve this, we will simply call the run function from mypy.api (as suggested in the documentation) and pass the -c PROGRAM_TEXT parameter that checks a string.

Here is the code for the type checker:

Let’s go through the code, given that there are a few interesting bits:

We start by defining a function called typechecker and registering it as a cell magic function called %%mypy. Why didn’t I just define a function called mypy instead of doing this renaming? Well, if I did that, then our mypy function would shadow the mypy module. In this case, it probably won’t cause any problems. But in general, you should avoid shadowing variables/functions/modules, because one day, it will cause you a lot of headache.

Inside our function, we first try to import the mypy module. If it’s not available, we inform the user that it should be installed, before this magic function can be used. The nice thing about importing mypy in the typechecker function is that the import error will show up only when you run the magic function. If you put the import at the top of the file, then save the file inside IPython startup directory, and you don’t have mypy module installed, you will get the ImportError every time you start IPython. The downside of this approach is that you are running the import code every time you run the typechecker function. This is something that you should avoid doing, if you care about the performance, but in case of our little helper, it’s not a big problem.

If you are using Python 3.6 or higher, you can catch the ModuleNotFoundError error instead of ImportError. ModuleNotFoundError is a new subclass of ImportError thrown when a module can’t be located. I want to keep my code compatible with lower versions of Python 3, so I will stick to the ImportError.

result = run(['-c', cell, *args])

Note that the function used for defining a cell magic must accept both a line and cell parameter. Which is great, because this way, we can actually pass parameters to mypy! So here, we are passing additional arguments from the line parameter to the run function. Here is how you could run our magic function with different settings:

which is equivalent to running the following command in the command line: mypy --ignore-missing-imports --follow-imports error -c 'CODEBLOCK'.

The rest of the code is quite similar to the example from the documentation.

Testing time!

Our cell magic function is ready. Let’s save it in the IPython startup directory (what’s IPython startup directory?), so it will be available next time we start IPython. In my case, I’m saving it in a file called:

Now, let’s fire up IPython and see if it works:

In [2]: %%mypy
...: def greet(name: str) -> str:
...: return f"hello {name}"
...: greet(1)
Type checking report:<string>:3: error: Argument 1 to "greet" has incompatible type "int"; expected "str"Out[2]: 1

Great, it works! It returns 0 (which is a standard UNIX exit code for a successful command) if everything is fine. Otherwise, it reports what problems have been found.

How about passing some additional parameters?

Type checking report:<string>:1: error: No library stub file for module 'flask'
<string>:1: note: (Stub files are from
Out[3]: 1# Ok, this can happen (
# Let's ignore this error
In [4]: %%mypy --ignore-missing-imports
...: import flask
Out[4]: 0

Passing additional parameters also works!

Great, we created a nice little helper function that we can use for checking, if the type hints are correct in a given block of code.

Line and cell magic function

There is one more decorator that we didn’t discuss yet: @register_line_cell_magic. It’s nothing special - especially now that you know how line magic and cell magic works - so there is no need for a separate article. IPython documentation explains this decorator very well:

If you run %lcmagic, this function won’t receive the cell parameter and it will act as a line magic. If you run %%lcmagic, it will receive the cell parameter and - optionally - the line parameter (like in our last example with %%mypy). So you can check for the presence of cell parameter and based on that, control if it should act as a line or cell magic.


Now you know how to make a line magic and a cell magic functions and how to combine them together into a line and magic function. There is still one more feature that IPython offers — the Magics class. It allows you to write more powerful magic functions, as they can, for example, hold state in between calls. So stay tuned for the last part of this article!

Image from: Pexels


[1]: Writing a translator is still a great exercise! I recently followed the Let’s Build A Simple Interpreter series, where you would build a Pascal interpreter in Python, and it was a really fun project for someone who never studied the compilers. So, if you are interested in this type of challenge, that blog can help you get started.

Originally published at on February 8, 2019.

Written by

Python consultant, freelancer, and trainer at Writes about productivity, tools, Python, and programming best practices.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store