ensure: Literate assertions in Python

ensure is a set of simple assertion helpers that let you write more expressive, literate, concise, and readable Pythonic code for validating conditions. It's inspired by should.js, expect.js, and builds on top of the unittest/JUnit assert helpers.

If you use Python 3, you can use ensure to enforce your function signature annotations: see PEP 3107 and the @ensure_annotations decorator below.

Because ensure is fast, is a standalone library (not part of a test framework), doesn't monkey-patch anything or use DSLs, and doesn't use the assert statement (which is liable to be turned off with the -O flag), it can be used to validate conditions in production code, not just for testing (though it certainly works as a BDD test utility library).

Aside from better looking code, a big reason to use ensure is that it provides more consistent, readable, and informative error messages when things go wrong. See Motivation and Goals for more.

Installation

pip install ensure

Synopsis

from ensure import ensure

ensure(1).is_an(int)
ensure({1: {2: 3}}).equals({1: {2: 3}}).also.contains(1)
ensure({1: "a"}).has_key(1).whose_value.has_length(1)
ensure.each_of([{1: 2}, {3: 4}]).is_a(dict).of(int).to(int)
ensure(int).called_with("1100101", base=2).returns(101)
ensure(dict).called_with(1, 2).raises(TypeError)
check(1).is_a(float).or_raise(Exception, "An error happened: {msg}. See http://example.com for more information.")

In Python 3:

from ensure import ensure_annotations

@ensure_annotations
def f(x: int, y: float) -> float:
    return x+y

See More examples below.

Notes

The ensure module exports the Ensure class and its convenience instance ensure. Instances of the class are callable, and the call will reset the contents that the instance is inspecting, so you can reuse it for many checks (as seen above).

The class raises EnsureError (a subclass of AssertionError) by default.

There are several ways to chain clauses, depending on the grammatical context: .also, .which, and .whose_value are available per examples below.

Raising custom exceptions

You can pass a callable or exception class as the error_factory keyword argument to Ensure(), or you can use the Check class or its convenience instance check(). This class behaves like Ensure, but does not raise errors immediately. It saves them and chains the methods otherwise(), or_raise() and or_call() to the end of the clauses.

from ensure import check

check("w00t").is_an(int).or_raise(Exception)
check(1).is_a(float).or_raise(Exception, "An error happened: {msg}. See http://example.com for more information.")
check("w00t").is_an(int).or_raise(MyException, 1, 2, x=3, y=4)

def build_fancy_exception(original_exception):
    return MyException(original_exception)

check("w00t").is_an(int).otherwise(build_fancy_exception)
check("w00t").is_an(int).or_call(build_fancy_exception, *args, **kwargs)

More examples

ensure({1: {2: 3}}).is_not_equal_to({1: {2: 4}})
ensure(True).does_not_equal(False)
ensure(1).is_in(range(10))
ensure(True).is_a(bool)
ensure(True).is_(True)
ensure(True).is_not(False)

ensure(["train", "boat"]).contains_one_of(["train"])
ensure(range(8)).contains(5)
ensure(["spam"]).contains_none_of(["eggs", "ham"])
ensure("abcdef").contains_some_of("abcxyz")
ensure("abcdef").contains_one_or_more_of("abcxyz")
ensure("abcdef").contains_all_of("acf")
ensure("abcd").contains_only("dcba")
ensure("abc").does_not_contain("xyz")
ensure([1, 2, 3]).contains_no(float)
ensure(1).is_in(range(10))
ensure("z").is_not_in("abc")
ensure(None).is_not_in([])
ensure(dict).has_attribute('__contains__').which.is_callable()
ensure({1: "a", 2: "b", 3: "c"}).has_keys([1, 2])
ensure({1: "a", 2: "b"}).has_only_keys([1, 2])

ensure(1).is_true()
ensure(0).is_false()
ensure(None).is_none()
ensure(1).is_not_none()
ensure("").is_empty()
ensure([1, 2]).is_nonempty().also.has_length(2)
ensure(1.1).is_a(float).which.equals(1.10)
ensure(KeyError()).is_an(Exception)
ensure({x: str(x) for x in range(5)}).is_a_nonempty(dict).of(int).to(str)
ensure({}).is_an_empty(dict)
ensure(None).is_not_a(list)

import re
ensure("abc").matches("A", flags=re.IGNORECASE)
ensure([1, 2, 3]).is_an_iterable_of(int)
ensure([1, 2, 3]).is_a_list_of(int)
ensure({1, 2, 3}).is_a_set_of(int)
ensure({1: 2, 3: 4}).is_a_mapping_of(int).to(int)
ensure({1: 2, 3: 4}).is_a_dict_of(int).to(int)
ensure({1: 2, 3: 4}).is_a(dict).of(int).to(int)
ensure(10**100).is_numeric()
ensure(lambda: 1).is_callable()
ensure("abc").has_length(3)
ensure("abc").has_length(min=3, max=8)
ensure(1).is_greater_than(0)
ensure(1).exceeds(0)
ensure(0).is_less_than(1)
ensure(1).is_greater_than_or_equal_to(1)
ensure(0).is_less_than_or_equal_to(0)
ensure(1).is_positive()
ensure(1.1).is_a_positive(float)
ensure(-1).is_negative()
ensure(-1).is_a_negative(int)
ensure(0).is_nonnegative()
ensure(0).is_a_nonnegative(int)
ensure([1,2,3]).is_sorted()

ensure("{x} {y}".format).called_with(x=1, y=2).equals("1 2")
ensure(int).called_with("1100101", base=2).returns(101)
ensure("{x} {y}".format).with_args(x=1, y=2).is_a(str)
with ensure().raises(ZeroDivisionError):
    1/0
with ensure().raises_regex(NameError, "'w00t' is not defined"):
    w00t

See complete API documentation.

Enforcing function annotations

Use the @ensure_annotations decorator to enforce function signature annotations:

from ensure import ensure_annotations

@ensure_annotations
def f(x: int, y: float) -> float:
    return x+y

f(1, 2.3)

>>> 3.3

f(1, 2)

>>> ensure.EnsureError: Argument y to <function f at 0x109b7c710> does not match annotation type <class 'float'>

Compare this runtime type checking to compile-time checking in Mypy and type hinting in PEP 484/Python 3.5+.

Motivation and goals

Many BDD assertion libraries suffer from an excess of magic, or end up having to construct statements that don't parse as English easily. ensure is deliberately kept simple to avoid succumbing to either issue. The source is easy to read and extend.

Work remains to make error messages raised by ensure even more readable, informative, and consistent. Going forward, ability to introspect exceptions to extract structured error information will be a major development focus. You will be in control of how much information is presented in each error, which context it's thrown from, and what introspection capabilities the exception object will have.

The original use case for ensure is as an I/O validation helper for API endpoints, where the client needs to be sent a very clear message about what went wrong, some structured information (such as an HTTP error code and machine-readable reference to a failing element) may need to be added, and some information may need to be hidden from the client. To further improve on that, we will work on better error translation, marshalling, message formatting, and schema validation helpers.

Authors

Andrey Kislyuk
Harrison Metzger

License

Licensed under the terms of the Apache License, Version 2.0.

kislyuk/ensure

kislyuk

Reviews

Repository Details