(All tags.)
ProtocolDuck typing says “if it quacks like a duck, treat it like a duck.” Type checking seems at odds with duck typing: we restrict variables to named types (classes), allowing only that type or subtypes. This seems like it would stop us from passing in arbitrary objects that can “quack the right way”. But since PEP 0544, we can declare “duckish” types with typing.Protocol.
isinstance(), assert, and LiteralType narrowing is the ability for a type checker to infer that inside some branch, a variable has a more specific (narrower) type than its definition. This allows us to perform type-specific operations without any explicit casting.
In a type hint, if we specify a type (class), then we mark the variable as containing an instance of that type. To specify that a variable instead contains a type, we need to use type[Cls] (or the old syntax typing.Type).
__future__.annotationsPython 3.7 added a new “postponed evaluation” mode for type hints. In this post we’ll cover how it changes the behaviour of type hints, how to activate it, and the outstanding problems.
reveal_type()When working with type hints, it’s often useful to debug the types of variables. Type checkers allow you to do this with reveal_type() and reveal_locals().
Circular imports are always annoying when they arise in Python, and type hints make them more common. Thankfully, there’s a trick to add circular imports for type hints without causing ImportErrors.
Using TypedDict is a great way to type code that passes around dictionaries. The default way to create a TypedDict is with its class-based syntax:
*args and **kwargsWhen I started writing type hints, I was a little confused about what to do with Python’s variable argument operators, * and ** (often called *args and **kwargs). Here’s what I figured out.
Take this function:
When starting out with Python type hints, it’s common to overuse typing.Any. This is dangerous, since Any entirely disables type checking for a variable, allowing any operation.
Python 3.9 introduced zoneinfo into the standard library. This module reads your operating system’s copy of the tz database. On older versions of Python, you can install the backports.zoneinfo package to use zoneinfo without upgrading.
functools.partial() in DjangoPython’s functools.partial is a great tool that I feel is underused.
Google has started rolling out FLoC, currently to 0.5% of Chrome users, and some sites are already disabling it. In this post we’ll cover what FLoC is, who’s disabling it, why, and how to do so on a Django site.
heroicons is a free SVG icon set for your websites, from the creators of tailwindcss. SVG icons are great - they’re small, they sit inline in your HTML, and you can scaled and colour them with plain HTML and CSS. And heroicons is a great icon set - minimal, clear, and consistent.
I recently upgraded my client ev.energy to PostgreSQL 13. The first feature listed in this version’s release notes is “Space savings and performance gains from de-duplication of B-tree index entries”. I reindexed all tables after the upgrade to take advantage of this deduplication and saw index storage savings of up to 90%.
TestCase from using setUp() to setUpTestData()Django’s TestCase class provides the setUpTestData() hook for creating your test data. It is faster than using the unittest setUp() hook because it creates the test data only once per test case, rather than per test.
I wrote my library time-machine last year as a way to speed up tests that need to accurately mock the current time.
For all my linting needs these days I use the pre-commit framework. It has integrations with every tool I want to use, and uses Git’s hooks to prevent non-passing code from ever being committed.
I recently optimized a client project’s test suite, and in the process found a test whose runtime had crept up ever since it had been written. The problematic test exercised an import process from a fixed past date until the current day. The test’s runtime therefore grew every day, until it reached over a minute.
In all current releases of the popular WSGI server gunicorn, the Server header reports the complete version of gunicorn. I spotted this on my new project DB Buddy. For example, with httpie to check the response headers:
If you’re testing Python code that relies on the current date or time, you will probably want to mock time to test different scenarios. For example, what happens when you run a certain piece of code on February 29? (A common source of bugs.)
Earlier this week I tweeted about my two part offer on my book Speed Up Your Django Tests for this year’s Cyber Monday.
Sometimes tests need to change environment variables. This is fairly straightforward in pytest, thanks to os.environ quacking like a dict, and the mock.patch.dict decorator/context manager.
Sometimes, tests need to change environment variables. This is straightforward in tests using Python’s unittest, thanks to os.environ quacking like a dict, and the mock.patch.dict decorator/context manager.
When we write custom management commands, it’s easy to write integration tests for them with call_command(). This allows us to invoke the management command as it runs under manage.py, and retrieve the return code, standard output, and standard error. It’s great, but has some overhead, making our tests slower than necessary. If we have logic separated out of the command’s handle() method, it improves both readability and testability, as we can unit test it separately.