lazy-load

ℒazy-ℒoad

.. image:: https://img.shields.io/pypi/v/lazy_load.svg :target: https://pypi.python.org/pypi/lazy_load :alt: Latest PyPI version

.. image:: https://travis-ci.org/kutoga/lazy-load.png :target: https://travis-ci.org/kutoga/lazy-load :alt: Latest Travis CI build status

A minimalistic interface that allows lazy evaluation of expressions and function calls.

Note: This small library is heavily based on python-lazy-object-proxy.

Why using ℒazy-ℒoad? Lazy loading in general may make some software implementations much more efficient. Especially if it is not known if some data has to be loaded or not. Often the resulting code is less efficient, because eager loading is used or the code is not elegant, because one has to program (somehow) lazy loading.

Advantages of this library are that lazy-loading may be used quite elegant and effective.

Examples ^^^^^^^^

In a loop it might happen that a special condition appears once or even more often. If this is the case, an expensive function expensive_function is called and on the resulting object an operation has to be done. If the expensive function had to called more than once, than the result object may be reused.

Possible implementation:

.. code:: python

def expensive_function():
    print("function evaluation")
    ...
    return result

obj = None
for x, y, p in get_coordinates():
    if test_for_something(x, y, p):
        if obj is None:
            obj = expensive_function()
        obj.do_something(x, y)

Given this library, it might be done like this:

.. code:: python

from lazy_load import lazy

def expensive_function():
    print("function evaluation")
    ...
    return result

obj = lazy(expensive_function)
for x, y, p in get_coordinates():
    if test_for_something(x, y, p):
        obj.do_something(x, y)

There are similar situations outside of loops. The implementation without lazy-load might look like this:

.. code:: python

def expensive_function():
    print("function evaluation")
    ...
    return result

obj = None
def get_obj():
    global obj
    if obj is None:
        obj = expensive_function()
    return obj

if condition_a:
    get_obj().xyz()
if condition_b:
    do_something()
if condition_c:
    get_obj().abc()

This code can be realized much easier with lazy-load. Not only is the code shorter, but it is also more readable:

.. code:: python

from lazy_load import lazy

def expensive_function():
    print("function evaluation")
    ...
    return result

obj = lazy(expensive_function)

if condition_a:
    obj.xyz()
if condition_b:
    do_something()
if condition_c:
    obj.abc()

It might be the case that the expensive function is used more often and always a lazy evaluation is done. In this case, a decorator might be used to indicate that all function calls to this function shall be lazily evaluated. This makes it possible to normally use the function. The behaviour is still the same like in the first example:

.. code:: python

from lazy_load import lazy_func

@lazy_func
def expensive_function():
    print("function evaluation")
    ...
    return result

obj = expensive_function()
for x, y, p in get_coordinates():
    if test_for_something(x, y, p):
        obj.do_something(x, y)

A lazy evaluation of functions / methods calls might be done with the @lazy_func decorator of with the lazy-call. This was already shown, therefore the following examples show how to do a one-shot lazy evaluation of a function call:

.. code:: python

from lazy_load import lazy, lz

def expensive_func(x, y):
    print(f"function evaluation with arguments x={x}, y={y}")
    ...
    return result

# Possibility 1: Use `lazy` with a callable
obj = lazy(lambda: expensive_func(a, b))

# Possibility 2: If it doesn't matter if the argument expressions for the expensive-function are eager evaluated, the call may be simplified:
obj = lazy(expensive_func, a, b)

# Possibility 3: `lazy` has a short version / alias: `lz`
obj = lz(expensive_func, a, b)

Python allows it to pass functions around: This is often used for callbacks, but also for other use cases. Assuming an expensive function is passed to an object which calls this function and stores the result of the function call in an attribute. Later it might happen that this attribute is used. Depending on the program flow it also might happen that this attribute is not used. With a lazily evaluated function the expensive function call is only executed if the result is really used. The lazily evaluated version of a function has the exact same signature as the original function.

One might now like to have the possibility to on-the-fly convert a callable to a lazily evaluated callable. This might be done in the following way:

.. code:: python

from lazy_load import lazy_func, lf

def expensive_func(x):
    print(f"function evaluation with argument x={x}")
    ...
    return result

# Possibility 1: Use `lazy_func`.
my_obj.do_something(f=lazy_func(expensive_func))

# Possibility 2: Use `lf` which is an alias of `lazy_func`
my_obj.do_something(f=lf(expensive_func))

# Possibility 3: Use the `ℒ`-"operator"
my_obj.do_something(f=ℒ[expensive_func])

Actually, I want to go deeper into the ℒazy- or ℒ-"operator". This operator converts on-the-fly a function to a lazily evaluated function. Another example:

.. code:: python

from lazy_load import ℒ

def test(name):
    print(f"hey {name}")
    return True

res = test("peter")
# prints "hey peter"

test_l = ℒ[test]
res = test_l("hans")
# prints nothing

if res:
    print("res is True")
# prints "hey hans\nres is True"

It is also possible to convert multiple functions to lazily evaluated functions using ℒ:

.. code:: python

from lazy_load import ℒ

def f1(x):
    print(f"f1 {x}")
    return True

def f2(x):
    print(f"f1 {x}")
    return True

f1_l, f2_l, f3_l = ℒ[f1, f2, lambda x: x == 1]
# This is equal to:
f1_l = ℒ[f1]
f2_l = ℒ[f2]
f3_l = ℒ[lambda x: x == 1]

Finally, one might like to decorate a class in a way that all its public methods which have a return value are lazily evaluated. Public methods are all methods that have a name not starting with _. Methods with a return value are identificated by the given return type hint which must not be None. This behaviour might be done with the @lazy_class-decorator (alias: lc):

.. code:: python

from lazy_load import lazy_class

@lazy_class
class MyClass:
    def __init__(self):
        # Method name starts with "_" => is not public; therefore it is eager evaluated
        pass

    def setX(x) -> None:
        # Method does not return a value => therefore it is eager evaluated
        ...

    def do():
        # Method does not hav a return value type hint =>  therefore it is eager evaluated
        ...

    def compute() -> int:
        # Method will always be lazily evaluated
        ...
        return result

Finally, it is also possible to force the evaluation of a lazy loading object by using force_eval (alias fe). This function can safely to used to non-lazy loading objects: It is then just equal to the identity function.

.. code:: python

from lazy_load import lazy, force_eval

def f1(x):
    print(f"f1 {x}")
    return True

lazy_obj = lazy(f1, 1)

# The following expression prints "f1 1" and returns "True"
force_eval(lazy_obj)

The force_eval function may also be applied to lazy-functions (which are created with lazy_func(x), @lazy_func or with ℒ). This restores the original non-lazy / eager function. For non-lazy functions this call has no effect:

.. code:: python

from lazy_load import lazy_func, force_eval

@lazy_func
def f(x):
    print("hey")
    return x**2

# The following line prints nothing
obj = f(2)

f_eager = force_eval(f)

# The following line prints "hey" and "obj" has immediatly the value "4"
obj = f_eager(2)

Installation

pip install lazy-load

Requirements ^^^^^^^^^^^^

Python 3.6 or Python 3.7.

Licence

MIT

Authors

lazy_load was written by Benjamin Bruno Meier <[e-mail]>_.