Headline

Delay evaluation of an expression until its value is needed

Illustration

Lazy by definition

Lazy evaluation is either supported by the underlying programming language or it needs to be encoded by the programs. Let's start with illustrations in Haskell; this language's semantics is lazy by definition.

Consider the following expression and its evaluation:

> repeat 42
[42,42,42,42,42,42,42,42,42,42,42,42,42,42,...

That is, 42 is to be repeated an infinite number of times and all those 42s are to be collected in one list. It is not surprising that the evaluation of this expression never stops as witnessed by printing the infinite result forever. Laziness comes into play when such expressions are used in a way that they do not need to be fully evaluated.

For instance, let us compute the head of an infinite list:

> head $ repeat 42
42

Thus, the list of repeated 42s is never materialized; rather the infinite list is only computed up to the point needed for returning the result, i.e., the head of the list. Here is another example for exploiting laziness to compute on 'infinite' data:

> length $ take 42 $ repeat 42
42

That is, we compute the length of the list that holds the first 42 elements of the earlier infinite list of 42s. Here is yet another example:

> [1..] !! 41
42

That is, we retrieve the 42nd element (the 41st index) of the earlier list.

Lazy conditionals

Most languages are readily lazy in terms of the semantics of their conditionals such that the 'then' and 'else' branches are only evaluated or executed, if necessary. This specific form of laziness is obviously important for programming, regardless of whether we face a language with lazy or strict evaluation. For instance, consider the following definition of factorial in Haskell:

-- A straightforward definition of factorial
factorial :: Integer -> Integer
factorial x =
  if x < 0 
    then error "factorial arg error"
    else if x <= 1
      then 1
      else x * factorial (x-1)

Regardless of language, such a definition should not evaluate the recursive case, except when honored by the value of the condition. Thus, this style of recursive definition even works in a programming language with strict evaluation, .e.g, in Python:

# A straightforward definition of factorial
def factorial(x):
    if not isinstance(x, (int, long)) or x<0:
        raise RuntimeError('factorial arg error')
    else:
        if x <= 1:
            return 1
        else:
            return x * factorial(x-1)

The difference between lazy and eager evaluation becomes quite clear, when we attempt a definition of 'if' as a function. In Haskell, we can actually define a function to mimic 'if' and use it in revising the recursive definition of factorial:

-- A re-definition of "if"
ifThenElse :: Bool -> x -> x -> x
ifThenElse True x   = x
ifThenElse False   x = x

-- Factorial re-defined to use user-defined if
factorial' :: Integer -> Integer
factorial' x =
  ifThenElse (x < 0)
    (error "factorial arg error")
    (ifThenElse (x <= 1)
      1
      (x * factorial' (x-1)))

The fact that this definition works depends on the lazy evaluation semantics of Haskell. The arguments of the function ifThenElse are only evaluated, when they are really needed. Let us attempt the same experiment in a language with eager evaluation semantics, e.g., Python:

# A troubled re-definition of "if"
def troubledIf(b,x1,x2):
    if b:
        return x1
    else:
        return x2

# Factorial re-defined to use user-defined if
def troubledFactorial(x):
    if not isinstance(x, (int, long)) or x<0:
        raise RuntimeError('factorial arg error')
    else:
        return troubledIf(x<=1,1,x * troubledFactorial(x-1))

When exercising this definition, we get this sort of runtime error:

>>> troubledFactorial(5)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "program.py", line 23, in troubledFactorial
    return troubledIf(x<=1,1,x * troubledFactorial(x-1))
  File "program.py", line 23, in troubledFactorial
    return troubledIf(x<=1,1,x * troubledFactorial(x-1))
  File "program.py", line 23, in troubledFactorial
    return troubledIf(x<=1,1,x * troubledFactorial(x-1))
  File "program.py", line 23, in troubledFactorial
    return troubledIf(x<=1,1,x * troubledFactorial(x-1))
  File "program.py", line 23, in troubledFactorial
    return troubledIf(x<=1,1,x * troubledFactorial(x-1))
  File "program.py", line 23, in troubledFactorial
    return troubledIf(x<=1,1,x * troubledFactorial(x-1))
  File "program.py", line 21, in troubledFactorial
    raise RuntimeError('factorial arg error')
RuntimeError: factorial arg error

A quick analysis suggests that this runtime error arises from the fact that an application of the function 'generates' an infinite chain of recursive applications, thereby eventually leading to the application of the function to a negative number, which is intercepted by the precondition test of the function. Thus, the function troubledIf is clearly not lazy and it cannot be used in defining the factorial function.

Encoding laziness

One may encode laziness in a language with eager evaluation. To this end, each expressions, for which evaluation should be deferred, can be turned into a degenerated closure (lambda abstraction) such that the evaluation can be requested explicitly by a trivial application. Consider the following attempt at a user-defined 'if' in Python and its use in another attempt at the factorial function:

# A (properly) lazy re-definition of "if"
def lazyIf(b,x1,x2):
    if b:
        return x1(())
    else:
        return x2(())

# A definition of factorial using lazyIf
def lazyFactorial(x):
    if not isinstance(x, (int, long)) or x<0:
        raise RuntimeError('factorial arg error')
    else:
        return lazyIf(x<=1,lambda  : 1, lambda  : x * lazyFactorial(x-1))

Thus, evaluation is requested explicitly by passing "()" (i.e., the empty tuple) to a "deferred" expression. When constructing a deferred expression, then we use a lambda abstraction with a superfluous variable.

See Document:Okasaki96 for a profound discussion of data structures in a functional programming language while leveraging laziness for the benefit of efficiency.

Relationships

See the related concept of eager evaluation.

Synonyms (in a broad sense):

Metadata


There are no revisions for this page.

User contributions

    This user never has never made submissions.

    User edits

    Syntax for editing wiki

    For you are available next options:

    will make text bold.

    will make text italic.

    will make text underlined.

    will make text striked.

    will allow you to paste code headline into the page.

    will allow you to link into the page.

    will allow you to paste code with syntax highlight into the page. You will need to define used programming language.

    will allow you to paste image into the page.

    is list with bullets.

    is list with numbers.

    will allow your to insert slideshare presentation into the page. You need to copy link to presentation and insert it as parameter in this tag.

    will allow your to insert youtube video into the page. You need to copy link to youtube page with video and insert it as parameter in this tag.

    will allow your to insert code snippets from @worker.

    Syntax for editing wiki

    For you are available next options:

    will make text bold.

    will make text italic.

    will make text underlined.

    will make text striked.

    will allow you to paste code headline into the page.

    will allow you to link into the page.

    will allow you to paste code with syntax highlight into the page. You will need to define used programming language.

    will allow you to paste image into the page.

    is list with bullets.

    is list with numbers.

    will allow your to insert slideshare presentation into the page. You need to copy link to presentation and insert it as parameter in this tag.

    will allow your to insert youtube video into the page. You need to copy link to youtube page with video and insert it as parameter in this tag.

    will allow your to insert code snippets from @worker.