Skip to content

tiny python decorator library for (A) enforcing class structure and instance behavior on class hierarchies (B) invoking user defined functionality based on object state.

License

Notifications You must be signed in to change notification settings

hraoyama/decorules

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

57 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

decorules

Introduction

decorules is a tiny python decorator library with two objectives:

A. To enforce rules on class structure and instance behavior for classes and class hierarchies through decorators at the point of class declaration. Useful for library developers.

B. To trigger user defined functionality using boolean conditions on an instance of a class1

The decorators employed are:

  1. raise_if_false_on_class will raise an exception should class structure and/or attributes within a class hierarchy not adhere to user defined rules
  2. raise_if_false_on_instance will raise an exception should class instances not adhere to user defined rules2
  3. run_if_false_on_instance will run user supplied functionality should class instances not adhere to user defined criteria2
  4. run_instance_rules will apply the rules from 2. on any member function using this decorator
  5. run_instance_actions will apply the actions from 3. on any member function using this decorator

All rules and actions are specified through the decorators on the class declaration and using the metaclass HasRulesActions from the library.

Enforcement of the rules is done by throwing exceptions (which can be developer specified) when a predicate function fails.

The actions taken when a predicate fails are supplied by the user through functions taking the instance as argument.

Installation

decorules was built using python 3.10. It is available as a package on pypi and can be installed through pip:

pip install decorules

Should you require an installation of pip, follow the instructions on the pip website.

Examples

A worked out example of several types of class hierarchies can be found under src/example, with library_class.py and client_class.py representing the library and client respectively.

Further examples, including interaction with other decorators3, can be found in the source file under the tests directory.

The aim here is to simply walk through some simple examples to demonstrate usage.

Firstly, suppose we wish to enforce that a (base) class or an instance of the class must have an attribute of a certain type. Here are the basic steps:

  1. Create a function that takes a class or an instance and checks whether an attribute exists and is of the correct type. In the example, this function is key_type_enforcer
def key_type_enforcer(instance_or_type,
                      enforced_type: type,
                      enforced_key: str,
                      attrs: dict = None):
    member_object = getattr(instance_or_type, enforced_key, None)
    if member_object is None:
        if attrs is not None:
            member_object = attrs.get(enforced_key, None)
    if member_object is None:
        return False
    else:
        return issubclass(type(member_object), enforced_type)
    pass

In order to guarantee that the class (and its derived classes) implements a function named library_functionality we would implement:

from decorules.has_rules_actions import HasRulesActions
import types
from functools import partial

@raise_if_false_on_class(partial(key_type_enforcer, 
                                 enforced_type=types.FunctionType, 
                                 enforced_key='library_functionality'), 
                         AttributeError)
class HasCorrectMethodClass(metaclass=HasRulesActions):
    def library_functionality(self):
        return 1
  1. For restrictions on instances, the function must be predicate. This means the function takes one argument (the instance) and returns a boolean. Functions can be turned into predicates using different methods, in this example we will use partial from the functools package. For restrictions on classes that do not check the values of attributes predicate functions can be provided. If the rule on the class does make use of such a value (e.g., check if a static float is positive), the function must take 2 arguments and return a boolean. The second argument should always default to None4.
  2. Use the decorator raise_if_false_on_class when enforcing a rule on a class level, or raise_if_false_on_instance when enforcing upon instantiation. Both decorators take 1 compulsory argument (the function from step 2. which returns a True/False value) and 2 optional arguments, the first is the type of the exception to be raised should the rule not hold5 and the second optional argument is a string providing extra information when the exception is raised.
  3. The rules on instances are only applied after the call to __init__. We have the option to add the run_instance_rules decorator to any method of the class, thereby enforcing the instance rules after each method call.

If in addition, we ensure that an int member named x existed after every instantiation:

@raise_if_false_on_instance(partial(key_type_enforcer, enforced_type=int, enforced_key='x'), AttributeError)  
@raise_if_false_on_class(partial(key_type_enforcer, enforced_type=types.FunctionType, enforced_key='library_functionality'), AttributeError)
class HasCorrectMethodAndInstanceVarClass(metaclass=HasRulesActions):
    def __init__(self, value=20):
        self.x = value
    def library_functionality(self):
        return 1

Should the __init__ implementation not set self.x or remove it using del self.x, all of the following calls would throw an AttributeError:

a = HasCorrectMethodAndInstanceVarClass()
b = HasCorrectMethodAndInstanceVarClass(25)
c = HasCorrectMethodAndInstanceVarClass(5)

For forcing the member x to be larger than 10:

@raise_if_false_on_instance(lambda ins: ins.x > 10, ValueError, "Check x-member>10")  
@raise_if_false_on_instance(partial(key_type_enforcer, enforced_type=int, enforced_key='x'), AttributeError)  
@raise_if_false_on_class(partial(key_type_enforcer, enforced_type=types.FunctionType, enforced_key='library_functionality'), AttributeError)
class HasCorrectMethodAndInstanceVarCheckClass(metaclass=HasRulesActions):
    def __init__(self, value=20):
        self.x = value
    def library_functionality(self):
        return 1

Note the third argument in the decorator, this will be prepended to the message of the exception. For the implementation above, only the third line would raise an exception:

a = HasCorrectMethodAndInstanceVarCheckClass()
b = HasCorrectMethodAndInstanceVarCheckClass(25)
c = HasCorrectMethodAndInstanceVarCheckClass(5) # a ValueError is raised

Because the key-type + comparison paradigm is expected to be widely used for classes and instances, decorules provides a utility for this called member_enforcer6. The previous snippet could have been simplified using:

import operator
from decorules.utils import member_enforcer

@raise_if_false_on_instance(member_enforcer('x',int, 10, operator.gt), ValueError, "Check x-member>10")
@raise_if_false_on_class(member_enforcer('library_functionality', types.FunctionType), AttributeError)
class HasCorrectMethodAndInstanceVarCheckClass(metaclass=HasRulesActions):
    def __init__(self, value=20):
        self.x = value
    def library_functionality(self):
        return 1

If we wanted to ensure that a static set had a minimum number of instances of each type (e.g., 1 string, 2 int and 1 float):

from collections import Counter
from collections.abc import Iterable

def min_list_type_counter(instance_or_type,
                          list_name: str,
                          min_counter: Counter,
                          attrs: dict = None):
    member_object = getattr(instance_or_type, list_name, None)
    if member_object is None:
        if attrs is not None:
            member_object = attrs.get(list_name, None)
    if member_object is None:
        return False
    else:
        if isinstance(member_object, Iterable):
            return Counter(type(x) for x in member_object) >= min_counter
        else:
            return False


@raise_if_false_on_class(partial(min_list_type_counter, 
                                 list_name='STATIC_SET', 
                                 min_counter = Counter({str: 1, int: 2, float:1})), 
                         AttributeError)
class HasClassLevelMemberTypeCheckClass(metaclass=HasRulesActions):
    STATIC_SET = ("Test", 10, 40, 50, 45.5, 60.0, '3', 'i', BaseException())

If we wanted to raise an exception as soon as a member value reaches the value 10 during the course of the process:

@raise_if_false_on_instance(lambda x: x.y<10, ValueError)
class HasMethodCheckedAndFailsAfterCall(metaclass=HasRulesActions):
    def __init__(self, value=20):
        self.y = value
    @run_instance_rules
    def add(self, value=0):
        self.y += value

a = HasMethodCheckedAndFailsAfterCall(0)
a.add(1)
a.add(1)
a.add(1)
a.add(10)  # will raise a ValueError

To illustrate the triggering of functionality we create the following contrived example: a ProducerClass manages an integer resource that has be >=0 and <100. Every time a value larger than or equal to 20 is produced, it passes the value to an instance of LargeNumberProcessor. If the latter is passed a value larger than or equal to 50 it raises an exception. If the average of unprocessed values in its list is larger than or equal to 30, all of the values to process will get halved.

def is_m_positive_and_lt_100(instance):
    return (instance.m >= 0) & (instance.m < 100)

def is_m_lt_20(instance):
    return instance.m < 20

def is_last_entry_lt_50(instance):
    if instance.to_process_list:
        return instance.to_process_list[-1] < 50
    else:
        return True  # still empty list

def is_mean_entry_lt_30(instance):
    if instance.to_process_list:
        return sum(instance.to_process_list) / len(instance.to_process_list) < 30.0
    else:
        return True  # still empty list

def halve_list(instance):
    if instance.to_process_list:
        instance.to_process_list = [int(x*0.5) for x in instance.to_process_list]

@run_if_false_on_instance(is_mean_entry_lt_30, halve_list)
@raise_if_false_on_instance(is_last_entry_lt_50, ValueError, "Refuse to accept value >50")
class LargeNumberProcessor(metaclass=HasRulesActions):
    def __init__(self):
        self.to_process_list = []

    @run_instance_actions
    @run_instance_rules
    def append_number(self, value: int):
        self.to_process_list.append(value)

    @run_instance_actions
    @run_instance_rules
    def process_front_number(self):
        if self.to_process_list:
            return self.to_process_list.pop(0)
        else:
            return None

NUMBER_PROCESSOR = LargeNumberProcessor()

def add_to_LNP(instance):
    NUMBER_PROCESSOR.append_number(instance.m)

@run_if_false_on_instance(is_m_lt_20, add_to_LNP)
@raise_if_false_on_instance(is_m_positive_and_lt_100, AttributeError)
class ProducerClass(metaclass=HasRulesActions):
    def __init__(self, value: int = 0):
        self.m = value

    @run_instance_actions
    @run_instance_rules
    def add(self, other: int):
        self.m += other

An example run would then be:

k = ProducerClass()
k.add(5)
k.add(5)
k.add(5)
k.add(5)
assert len(NUMBER_PROCESSOR.to_process_list) == 1  # first value of 20 is sent
k.add(5)
assert NUMBER_PROCESSOR.to_process_list[-1] == 25 
k.add(-14)
assert NUMBER_PROCESSOR.to_process_list[-1] == 25 # we dropped below 20 so no new value was passed
k.add(11)
NUMBER_PROCESSOR.process_front_number() # 20 is gone from the list
k.add(8)
assert NUMBER_PROCESSOR.to_process_list == [25, 22, 30]
NUMBER_PROCESSOR.process_front_number()
assert NUMBER_PROCESSOR.to_process_list == [22, 30]
NUMBER_PROCESSOR.process_front_number()  # because the list average is now [30], its values will get halved
assert NUMBER_PROCESSOR.to_process_list == [15]
k.add(-9) # the managed int goes from 30 to 21
assert NUMBER_PROCESSOR.to_process_list == [15, 21]
k.add(40)  # will raise a ValueError as we try to pass 61>=50 to the LargeNumberProcessor

Note that run_if_false_on_instance only takes 2 arguments: a predicate function taking the instance as an argument and the function that will be executed should the predicate be false. The latter takes the instance as an argument7.

When using multiple decorators in general, one must be aware that the order of decorator matters with decorator closest to the function/class applied first. With multiple decorator we must also avoid clashes between decorators.

Though not intended for this use, the enforced rules and actions (both through predicate functions) are available through the EnforcedFunctions static class and can thus be retrieved, applied and transferred at any point in the code.

Footnotes

  1. The functionality itself is up to the user. Possible suggestions could be callback mechanisms, logging, asynchronous tasks, etc.

  2. By default, rules and actions on instances are enforced after creation of an instance only. It is possible use these rules and actions after any member function call by using the run_instance_-style decorator on the method. 2

  3. Here we refer to interactions with the dataclasses and property decorators

  4. The second argument will be used to examine class attributes when required. Note that by always providing a second argument and defaulting it to None (as was done in key_type_enforcer), the function can be used both on instances and class declarations.

  5. Note that this is an exception type and not an instance. For rules on classes this defaults to AttributeError, for rules of instantiation this defaults to ValueError. Other exceptions or classes (including user defined ones) can be supplied, provided instances can be constructed from a string

  6. member_enforcer has 2 compulsory arguments: the enforced_key (a string with the attribute name) and the enforced_type (the type of the attribute) and 2 optional arguments: the comparison_value and the operator_used, the latter defaults to the boolean equality operator and is only applied if a value is provided.

  7. additional arguments can be bound using methods like partial