Locking Class Attributes

Most object oriented languages (C++, Java, Dart, Kotlin, Swift) include visibiliy modifiers. This enables encapsulation where for example the inner workings of a class can be detached from the outside world and thus protected from direct modification.

Python on the other hand does not have a language-backed concept of privacy. Instead functions or variables with an identifier that starts with an underscore are deemed private and should not be modified or otherwise relied upon since they may change in a future version of the module.

In some cases, certain attributes may be crucial for the correct working a class and the programmer might want to pervent any inadvertent modification.

The package lockattrs provides a decorator that can be used with the method __setattr__ to lock certain attributes or all attributes.

Note that despite the name similarity lockattrs is not related to the package attrs providing a concise way of creating and validating data classes.

Installation

To install the package lockattrs use the command:

$ pip install lockattrs

Usage

This package provides the decorator function protect which can be used to prevent modification of attributes after they have been initially set.

1. Locking Class Attributes

The intended use-case is demonstrated below. Locking the instance attributes of a meta-class is equivalent to locking the class attributes of the class (the meta-class instance).

Using the decorator protect involves the following steps:

1. Declare a meta-class.
2. Override the method __setattr__.
3. Decorate __setattr__ with the function protect.
4. Optionally: Specify which attributes should be locked and what type of error should be raised during an attribute modification attempt.

from lockattrs import protect

class AMeta(type):
    """
    Meta class of A.
    """
    @protect(('data','id'), )
    def __setattr__(self, name: str, value: Any) -> None:
        return super().__setattr__(name, value)

class A(metaclass=AMeta):
    id = 'a01'
    pass

A.id = 'b02'            # Raises an error. Attribute 'id' is set and locked.

A.data = 'initial-data' # First initiation is OK. Attribute 'data' is now locked.
A.data = 'new-data'     # Raises an error (default type: ProtectedAttributeError).

A.name = 'A'
A.name = 'A1'           # OK, since the attribute 'name' is not locked.

2. Locking Instance Attributes

The code below demonstrates how to use the decorator function @protect to lock certain attributes of a class instance.

from lockattrs import protect

class B():
    """
    Sample class with locked attributes.
    """
    id = 57

    @protect(('data','id'), ) # To lock all attributes use: @protect()
    def __setattr__(self, name: str, value: Any) -> None:
        return super().__setattr__(name, value)



B.id = 28               # OK. Class attributes are not locked.
                        # To lock class attributes see section above.

# Creating an instance of B.
b = B()

b.id = 77               # Modification of the attribute 'id' via 'self' raises
                        # an error since the annotated method `__setattr__` is
                        # called. 

b.data = 'initial-data' # First initiation is OK. Attribute 'data' is now locked.
b.data = 'new-data'     # Raises an error (default type: ProtectedAttributeError).

b.name = 'b'
b.name = 'b1'           # OK, since the attribute 'name' is not locked.

Performance

Note: Locking certain attributes may be prohibitively costly in terms of computational time when used with objects that are instantiated often (for example in a loop) and where attributes are set/modified frequently.

The benchmarks below were produced using the package pytest-benchmark on a PC with 32GB RAM and an Intel Core i5-6260U CPU running at 1.80GHz. As the mean runtimes show, setting an attribute of class A takes approximately 40 times as long compared to a standard class (without an annotated __setattr__ method).

--------------------------------- benchmark: 2 tests -----------------------------------
Name (time in ns)                   Mean              StdDev          Rounds  Iterations
----------------------------------------------------------------------------------------
test_benchmark_set_attrs        348.8611 (1.0)       66.8829 (1.0)         4       20000
test_benchmark_set_attrs_A   13,496.0524 (38.69)    912.2178 (13.64)       4       20000
----------------------------------------------------------------------------------------

Features and bugs

Please file feature requests and bugs at the issue tracker. Contributions are welcome.