Branch | Status |
---|---|
Master: | |
Dev: |
Enforce.py
Enforce.py is a Python 3.5+ library for integration testing and data validation through configurable and optional runtime type hint enforcement. It uses the standard type hinting syntax (defined in PEP 484).
NOTICE: Python versions 3.5.2 and earlier (3.5.0-3.5.2) are now deprecated. Only Python versions 3.5.3+ would be supported. Deprecated versions will no longer be officially supported in Enforce.py version 0.4.x.
Overview
- Supports most of simple and nested types
- Supports Callables, TypeVars and Generics
- Supports invariant, covariant, contravariant and bivariant type checking
- Default mode is invariant - the type has to match exactly, which is better suitable for testing but differs from Python's normal covariant type checking (a subclass can be used wherever a parent class is expected).
- Can be applied to both functions and classes (in this case it will be applied to all methods of the class)
- Highly configurable
- Global on/off switch
- Group configuration
- Local override of groups
- Type checking mode selection
- Dynamic reconfiguration
Installation
Stable 0.3.x - Stable and ready for every day use version
pip install enforce
Dev current - "Bleeding edge" features that, while are fairly consistent, may change.
pip install git+https://github.com/RussBaz/enforce.git@dev
Usage
Type enforcement is done using decorators around functions that you desire to be checked. By default, this decorator will ensure that any variables passed into the function call matches its declaration (invariantly by default). This includes integers, strings, etc. as well as lists, dictionaries, and more complex objects. Currently, the type checking is eager.
Note, eager means that for a large nested structure, every item in that structure will be checked. This may be a nightmare for performance! See caveats for more details.
You can also apply the runtime_validation
decorator around a class, and it
will enforce the types of every method in that class.
Note: this is a development feature and is not as thoroughly tested as the function decorators.
Features
Basic type hint enforcement
>>> import enforce
>>>
>>> @enforce.runtime_validation
... def foo(text: str) -> None:
... print(text)
>>>
>>> foo('Hello World')
Hello World
>>>
>>> foo(5)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/william/.local/lib/python3.5/site-packages/enforce/decorators.py", line 106, in universal
_args, _kwargs = enforcer.validate_inputs(parameters)
File "/home/william/.local/lib/python3.5/site-packages/enforce/enforcers.py", line 69, in validate_inputs
raise RuntimeTypeError(exception_text)
enforce.exceptions.RuntimeTypeError:
The following runtime type errors were encountered:
Argument 'text' was not of type <class 'str'>. Actual type was <class 'int'>.
>>>
Callable Support
@runtime_validation
def foo(a: typing.Callable[[int, int], str]) -> str:
return a(5, 6)
def bar(a: int, b: int) -> str:
return str(a * b)
class Baz:
def __call__(self, a: int, b: int) -> str:
return bar(a, b)
foo(bar)
foo(Baz())
TypeVar and Generics
T = typing.TypeVar('T', int, str)
@runtime_validation
class Sample(typing.Generic[T]):
def get(self, data: T) -> T:
return data
@runtime_validation
def foo(data: Sample[int], arg: int) -> int:
return data.get(arg)
@runtime_validation
def bar(data: T, arg: int) -> T:
return arg
sample_good = Sample[int]()
sample_bad = Sample()
with self.assertRaises(TypeError):
sample = Sample[list]()
foo(sample_good, 1)
with self.assertRaises(RuntimeTypeError):
foo(sample_bad, 1)
bar(1, 1)
with self.assertRaises(RuntimeTypeError):
bar('str', 1)
Class Decorator
Applying this decorator to a class will automatically apply the decorator to every method in the class.
@runtime_validation
class DoTheThing(object):
def __init__(self):
self.do_the_stuff(5, 6.0)
def do_the_stuff(self, a: int, b: float) -> str:
return str(a * b)
NamedTuple
Enforce.py supports typed NamedTuples.
MyNamedTuple = typing.NamedTuple('MyNamedTuple', [('param', int)])
# Optionally making a NamedTuple typed
# It will now enforce its type signature
# and will throw exceptions if there is a type mismatch
# MyNamedTuple(param='str') will now throw an exception
MyNamedTuple = runtime_validation(MyNamedTuple)
# This function now accepts only NamedTuple arguments
@runtime_validation
def foo(data: MyNamedTuple):
return data.param
Configuration
You can assign functions to groups, and apply options on the group level.
'None' leaves previous value unchanged.
All available global settings:
default_options = {
# Global enforce.py on/off switch
'enabled': None,
# Group related settings
'groups': {
# Dictionary of type {<name: str>: <status: bool>}
# Sets the status of specified groups
# Enable - True, disabled - False, do not change - None
'set': {},
# Sets the status of all groups to False before updating
'disable_previous': False,
# Sets the status of all groups to True before updating
'enable_previous': False,
# Deletes all the existing groups before updating
'clear_previous': False,
# Updating the default group status - default group is not affected by other settings
'default': None
},
# Sets the type checking mode
# Available options: 'invariant', 'covariant', 'contravariant', 'bivariant' and None
'mode': None
}
# Basic Example
@runtime_validation(group='best_group')
def foo(a: List[str]):
pass
foo(1) # No exception as the 'best_group' was not explicitly enabled
# Group Configuration
enforce.config({'groups': {'set': {'best_group': True}}}) # Enabling group 'best_group'
with self.assertRaises(RuntimeTypeError):
foo(1)
enforce.config({
'groups': {
'set': {
'foo': True
},
'disable_previous': True,
'default': False
}
}) # Disable everything but the 'foo' group
# Using foo's settings
@runtime_validation(group='foo')
def test1(a: str): return a
# Using foo's settings but locally overriding it to stay constantly enabled
@runtime_validation(group='foo', enabled=False)
def test2(a: str): return a
# Using bar's settings - deactivated group -> no type checking is performed
@runtime_validation(group='bar')
def test3(a: str): return a
# Using bar's settings but overriding locally -> type checking enabled
@runtime_validation(group='bar', enabled=True)
def test4(a: str): return a
with self.assertRaises(RuntimeTypeError):
test1(1)
test2(1)
test3(1)
with self.assertRaises(RuntimeTypeError):
test4(1)
foo(1)
enforce.config({'enabled': False}) # Disables enforce.py
test1(1)
test2(1)
test3(1)
test4(1)
foo(1)
enforce.config({'enabled': True}) # Re-enables enforce.py
enforce.config(reset=True) # Resets global settings to their default state
Caveats
Currently, iterators, generators and coroutines type checks are not supported (mostly). However, it is still possible to check if an object is iterable.
We are still working on the best approach for lazy type checking (checking list items only when accessed) and lazy type evaluation (accepting strings as type hints).
Currently, the type checker will examine every object in a list. This means that for large structures performance can be a nightmare.
Class decorators are not as well tested, and you may encounter a bug or two. Please report an issue if you do find one and we'll try to fix it as quickly as possible.
Changelog
0.3.4 - 11.06.2017
- Further improved exception messages and their consistency
- General bug fixes
0.3.3 - 23.04.2017
- Improved support for Dictionaries
- Fixed some thread safety issues
0.3.2 - 29.01.2017
- Added support for Python 3.5.3 and 3.6.0
- Added support for NamedTuple
- Added support for Set
- New exception message generation system
- Fixed failing nested lists type checking
0.3.1 - 17.09.2016
- Added support for Callable classes (classes with __call__ method are now treated like any other Callable object)
- Fixed bugs in processing callables without specified return type
Contributing
Please check out our active issues on our Github page to see what work needs to be done, and feel free to create a new issue if you find a bug.
Actual development is done in the 'dev' branch, which is merged to master at milestones.