Iterator


In computer programming, an iterator is an object that enables a programmer to traverse a container, particularly lists. Various types of iterators are often provided via a container's interface. Though the interface and semantics of a given iterator are fixed, iterators are often implemented in terms of the structures underlying a container implementation and are often tightly coupled to the container to enable the operational semantics of the iterator. An iterator performs traversal and also gives access to data elements in a container, but does not itself perform iteration. An iterator is behaviorally similar to a database cursor. Iterators date to the CLU programming language in 1974.

Description

Internal Iterators

Internal iterators are higher order functions such as map, reduce etc., implementing the traversal across a container, applying the given function to every element in turn.

External iterators and the iterator pattern

An external iterator may be thought of as a type of pointer that has two primary operations: referencing one particular element in the object collection, and modifying itself so it points to the next element. There must also be a way to create an iterator so it points to some first element as well as some way to determine when the iterator has exhausted all of the elements in the container. Depending on the language and intended use, iterators may also provide additional operations or exhibit different behaviors.
The primary purpose of an iterator is to allow a user to process every element of a container while isolating the user from the internal structure of the container. This allows the container to store elements in any manner it wishes while allowing the user to treat it as if it were a simple sequence or list. An iterator class is usually designed in tight coordination with the corresponding container class. Usually, the container provides the methods for creating iterators.
A loop counter is sometimes also referred to as a loop iterator. A loop counter, however, only provides the traversal functionality and not the element access functionality.

Generators

One way of implementing iterators is to use a restricted form of coroutine, known as a generator. By contrast with a subroutine, a generator coroutine can yield values to its caller multiple times, instead of returning just once. Most iterators are naturally expressible as generators, but because generators preserve their local state between invocations, they're particularly well-suited for complicated, stateful iterators, such as tree traversers. There are subtle differences and distinctions in the use of the terms "generator" and "iterator", which vary between authors and languages. In Python, a generator is an iterator constructor: a function that returns an iterator. An example of a Python generator returning an iterator for the Fibonacci numbers using Python's yield statement follows:

def fibonacci:
a, b = 0, 1
for _ in range:
yield a
a, b = b, a+b
for number in fibonacci: # The generator constructs an iterator
print

Implicit iterators

Some object-oriented languages such as C#, C++, Delphi, Go, Java, Lua, Perl, Python, Ruby provide an intrinsic way of iterating through the elements of a container object without the introduction of an explicit iterator object. An actual iterator object may exist in reality, but if it does it is not exposed within the source code of the language.
Implicit iterators are often manifested by a "foreach" statement, such as in the following Python example:

for value in iterable:
print

In Python, an iterable is an object which can be converted to an iterator, which is then iterated through during the for loop; this is done implicitly.
Or other times they may be created by the collection object itself, as in this Ruby example:

iterable.each do |value|
puts value
end

This iteration style is sometimes called "internal iteration" because its code fully executes within the context of the iterable object, and the programmer only provides the operation to execute at each step.
Languages that support list comprehensions or similar constructs may also make use of implicit iterators during the construction of the result list, as in Python:

names =

Sometimes the implicit hidden nature is only partial. The C++ language has a few function templates for implicit iteration, such as for_each. These functions still require explicit iterator objects as their initial input, but the subsequent iteration does not expose an iterator object to the user.

Streams

Iterators are a useful abstraction of input streams – they provide a potentially infinite iterable object. Several languages, such as Perl and Python, implement streams as iterators. In Python, iterators are objects representing streams of data. Alternative implementations of stream include data-driven languages, such as AWK and sed.

Contrasting with indexing

In procedural languages it is common to use the subscript operator and a loop counter to loop through all the elements in a sequence such as an array. Although indexing may also be used with some object-oriented containers, the use of iterators may have some advantages:
The ability of a container to be modified while iterating through its elements has become necessary in modern object-oriented programming, where the interrelationships between objects and the effects of operations may not be obvious. By using an iterator one is isolated from these sorts of consequences. This assertion must however be taken with a grain of salt, because more often than not, for efficiency reasons, the iterator implementation is so tightly bound to the container that it does preclude modification of the underlying container without invalidating itself.
For containers that may move around their data in memory, the only way to not invalidate the iterator is, for the container, to somehow keep track of all the currently alive iterators and update them on the fly. Since the number of iterators at a given time may be arbitrarily large in comparison to the size of the tied container, updating them all will drastically impair the complexity guarantee on the container's operations.
An alternative way to keep the number of updates bound relatively to the container size would be to use a kind of handle mechanism, that is a collection of indirect pointers to the container's elements that must be updated with the container, and let the iterators point to these handles instead of directly to the data elements. But this approach will negatively impact the iterator performance, since it must effectuate a double pointer following to access the actual data element. This is usually not desirable, because many algorithms using the iterators invoke the iterators data access operation more often than the advance method. It is therefore especially important to have iterators with very efficient data access.
All in all, this is always a trade-off between security and efficiency. Most of the time, the added security is not worth the efficiency price to pay for it. Using an alternative container would be a better choice if the stability of the iterators is needed.

Classifying iterators

Iterator categories

Iterators can be categorised according to their functionality. Here is a list of iterator categories:
CategoryLanguages
Bidirectional iteratorC++
Forward iteratorC++
Input iteratorC++
Output iteratorC++
Random access iteratorC++
Trivial iteratorC++

Iterator types

Different languages or libraries used with these languages define iterator types. Some of them are
TypeLanguages
Array iteratorPHP, R
Caching iteratorPHP
Constant iteratorC++, PHP
Directory iteratorPHP, Python
Filter iteratorPHP, R
Limit iteratorPHP
List iteratorJava, R
Recursive array iteratorPHP
XML iteratorPHP

In different programming languages

C# and other .NET languages

Iterators in the.NET Framework are called "enumerators" and represented by the IEnumerator interface. IEnumerator provides a MoveNext method, which advances to the next element and indicates whether the end of the collection has been reached; a Current property, to obtain the value of the element currently being pointed at; and an optional Reset method, to rewind the enumerator back to its initial position. The enumerator initially points to a special value before the first element, so a call to MoveNext is required to begin iterating.
Enumerators are typically obtained by calling the GetEnumerator method of an object implementing the IEnumerable interface. Container classes typically implement this interface. However, the foreach statement in C# can operate on any object providing such a method, even if it doesn't implement IEnumerable. Both interfaces were expanded into generic versions in.NET 2.0.
The following shows a simple use of iterators in C# 2.0:

// explicit version
IEnumerator iter = list.GetEnumerator;
while )
Console.WriteLine;
// implicit version
foreach
Console.WriteLine;

C# 2.0 also supports generators: a method that is declared as returning IEnumerator, but uses the "yield return" statement to produce a sequence of elements instead of returning an object instance, will be transformed by the compiler into a new class implementing the appropriate interface.

C++

The C++ language makes wide use of iterators in its Standard Library, and describes several categories of iterators differing in the repertoire of operations they allow. These include forward iterators, bidirectional iterators, and random access iterators, in order of increasing possibilities. All of the standard container template types provide iterators of one of these categories. Iterators generalize pointers to elements of an array, and their syntax is designed to resemble that of C pointer arithmetic, where the * and -> operators are used to reference the element to which the iterator points, and pointer arithmetic operators like ++ are used modify iterators in traversal of a container.
Traversal using iterators usually involves a single varying iterator, and two fixed iterators that serve to delimit a range to be traversed. The distance between the limiting iterators, in terms of the number of applications of the operator ++ needed to transform the lower limit into the upper one, equals the number of items in the designated range; the number of distinct iterator values involved is one more than that. By convention, the lower limiting iterator "points to" the first element in the range, while the upper limiting iterator does not point to any element in the range, but rather just beyond the end of the range.
For traversal of an entire container, the begin method provides the lower limit, and end the upper limit. The latter does not reference any element of the container at all, but is a valid iterator value that can be compared against.
The following example shows a typical use of an iterator.

std::vector items;
items.push_back; // Append integer value '5' to vector 'items'.
items.push_back; // Append integer value '2' to vector 'items'.
items.push_back; // Append integer value '9' to vector 'items'.
for ; it != items.end; ++it)
// In C++11, the same can be done without using any iterators:
for
// Each loops print "529".

Iterator types are separate from the container types they are used with, though the two are often used in concert. The category of the iterator usually depends on the type of container, with for instance arrays or vectors providing random access iterators, but sets only providing bidirectional iterators. A same container type can have more than one associated iterator type; for instance the std::vector<T> container type allows traversal either using pointers to its elements, or values of a special type std::vector<T>::iterator, and yet another type is provided for "reverse iterators", whose operations are defined in such a way that an algorithm performing a usual traversal will actually do traversal in reverse order when called with reverse iterators. Most containers also provide a separate const_iterator type, for which operations that would allow changing the values pointed to are intentionally not defined.
Simple traversal of a container object or a range of its elements can be done using iterators alone. But container types may also provide methods like insert or erase that modify the structure of the container itself; these are methods of the container class, but in addition require one or more iterator values to specify the desired operation. While it is possible to have multiple iterators pointing into the same container simultaneously, structure-modifying operations may invalidate certain iterator values ; using an invalidated iterator is an error that will lead to undefined behvior, and such errors need not be signalled by the run time system.
Implicit iteration is also partially supported by C++ through the use of standard function templates, such as ,

and
.
When used they must be initialized with existing iterators, usually begin and end, that define the range over which iteration occurs. But no explicit iterator object is subsequently exposed as the iteration proceeds. This example shows the use of for_each.

ContainerType c; // Any standard container type of ItemType elements.
void ProcessItem
std::for_each, c.end, ProcessItem); // A for-each iteration loop.

The same can be achieved using std::copy, passing a value as third iterator:

std::copy, c.end, std::ostream_iterator);

Since C++11, lambda function syntax can be used to specify to operation to be iterated inline, avoiding the need to define a named function. Here is an example of for-each iteration using a lambda function:

ContainerType c; // Any standard container type of ItemType elements.
// A for-each iteration loop with a lambda function.
std::for_each, c.end, ;

Java

Introduced in the Java JDK 1.2 release, the interface allows the iteration of container classes. Each Iterator provides a and method, and may optionally support a method. Iterators are created by the corresponding container class, typically by a method named iterator.
The next method advances the iterator and returns the value pointed to by the iterator. The first element is obtained upon the first call to next. To determine when all the elements in the container have been visited the hasNext test method is used. The following example shows a simple use of iterators:

Iterator iter = list.iterator;
//Iterator iter = list.iterator; in J2SE 5.0
while )

To show that hasNext can be called repeatedly, we use it to insert commas between the elements but not after the last element.
This approach does not properly separate the advance operation from the actual data access. If the data element must be used more than once for each advance, it needs to be stored in a temporary variable. When an advance is needed without data access, the access is nonetheless performed, though the returned value is ignored in this case.
For collection types that support it, the remove method of the iterator removes the most recently visited element from the container while keeping the iterator usable. Adding or removing elements by calling the methods of the container makes the iterator unusable. An attempt to get the next element throws the exception. An exception is also thrown if there are no more elements remaining.
Additionally, for there is a with a similar API but that allows forward and backward iteration, provides its current index in the list and allows setting of the list element at its position.
The J2SE 5.0 release of Java introduced the interface to support an enhanced for loop for iterating over collections and arrays. Iterable defines the method that returns an Iterator. Using the enhanced for loop, the preceding example can be rewritten as

for

Some containers also use the older Enumeration class. It provides hasMoreElements and nextElement methods but has no methods to modify the container.

Scala

In Scala, iterators have a rich set of methods similar to collections, and can be used directly in for loops. Indeed, both iterators and collections inherit from a common base trait - scala.collection.TraversableOnce. However, because of the rich set of methods available in the Scala collections library, such as map, collect, filter etc., it is often not necessary to deal with iterators directly when programming in Scala.
Java iterators and collections can be automatically converted into Scala iterators and collections, respectively, simply by adding the single line

import scala.collection.JavaConversions._

to the file. The JavaConversions object provides implicit conversions to do this. Implicit conversions are a feature of Scala: methods that, when visible in the current scope, automatically insert calls to themselves into relevant expressions at the appropriate place to make them typecheck when they otherwise wouldn't.

MATLAB

supports both external and internal implicit iteration using either "native" arrays or cell arrays. In the case of external iteration where the onus is on the user to advance the traversal and request next elements, one can define a set of elements within an array storage structure and traverse the elements using the for-loop construct. For example,

% Define an array of integers
myArray = ;
for n = myArray
%... do something with n
disp % Echo integer to Command Window
end

traverses an array of integers using the for keyword.
In the case of internal iteration where the user can supply an operation to the iterator to perform over every element of a collection, many built-in operators and MATLAB functions are overloaded to execute over every element of an array and return a corresponding output array implicitly. Furthermore, the arrayfun and cellfun functions can be leveraged for performing custom or user defined operations over "native" arrays and cell arrays respectively. For example,

function simpleFun
% Define an array of integers
myArray = ;
% Perform a custom operation over each element
myNewArray = arrayfunmyCustomFun;
% Echo resulting array to Command Window
myNewArray
function outScalar = myCustomFun
% Simply multiply by 2
outScalar = 2*inScalar;

defines a primary function simpleFun that implicitly applies custom subfunction myCustomFun to each element of an array using built-in function arrayfun.
Alternatively, it may be desirable to abstract the mechanisms of the array storage container from the user by defining a custom object-oriented MATLAB implementation of the Iterator Pattern. Such an implementation supporting external iteration is demonstrated in MATLAB Central File Exchange item . This is written in the new class-definition syntax introduced with MATLAB software version 7.6
and features a one-dimensional cell array realization of the List Abstract Data Type as the mechanism for storing a heterogeneous set of elements. It provides the functionality for explicit forward List traversal with the hasNext, next and reset methods for use in a while-loop.

PHP

’s foreach loop was introduced in version 4.0 and made compatible with objects as values in 4.0 Beta 4. However, support for iterators was added in PHP 5 through the introduction of the internal Traversable interface. The two main interfaces for implementation in PHP scripts that enable objects to be iterated via the foreach loop are Iterator and IteratorAggregate. The latter does not require the implementing class to declare all required methods, instead it implements an accessor method that returns an instance of Traversable. The Standard PHP Library provides several classes to work with special iterators. PHP also supports Generators since 5.5.
The simplest implementation is by wrapping an array, this can be useful for type hinting and information hiding.

namespace Wikipedia\Iterator;
final class ArrayIterator extends \Iterator

All methods of the example class are used during the execution of a complete foreach loop. The iterator’s methods are executed in the following order:
  1. $iterator->rewind ensures that the internal structure starts from the beginning.
  2. $iterator->valid returns true in this example.
  3. $iterator->current returned value is stored in $value.
  4. $iterator->key returned value is stored in $key.
  5. $iterator->next advances to the next element in the internal structure.
  6. $iterator->valid returns false and the loop is aborted.
The next example illustrates a PHP class that implements the Traversable interface, which could be wrapped in an IteratorIterator class to act upon the data before it is returned to the foreach loop. The usage together with the MYSQLI_USE_RESULT constant allows PHP scripts to iterate result sets with billions of rows with very little memory usage. These features are not exclusive to PHP nor to its MySQL class implementations.

mysqli_report;
$mysqli = new \mysqli;
// The \mysqli_result class that is returned by the method call implements the internal Traversable interface.
foreach

Python

Iterators in Python are a fundamental part of the language and in many cases go unseen as they are implicitly used in the for statement, in list comprehensions, and in generator expressions. All of Python's standard built-in collection types support iteration, as well as many classes that are part of the standard library. The following example shows typical implicit iteration over a sequence:

for value in sequence:
print

Python dictionaries can also be directly iterated over, when the dictionary keys are returned; or the items method of a dictionary can be iterated over where it yields corresponding key,value pairs as a tuple:

for key in dictionary:
value = dictionary
print


for key, value in dictionary.items:
print

Iterators however can be used and defined explicitly. For any iterable sequence type or class, the built-in function iter is used to create an iterator object. The iterator object can then be iterated with the next function, which uses the __next__ method internally, which returns the next element in the container. A StopIteration exception will be raised when no more elements are left. The following example shows an equivalent iteration over a sequence using explicit iterators:

it = iter
while True:
try:
value = it.next # in Python 2.x
value = next # in Python 3.x
except StopIteration:
break
print

Any user-defined class can support standard iteration by defining an __iter__ method that returns an iterator object. The iterator object then needs to define a __next__ method that returns the next element.
Python's generators implement this iteration protocol.

Ruby

Ruby implements iterators quite differently; all iterations are done by means of passing callback closures to container methods - this way Ruby not only implements basic iteration but also several patterns of iteration like function mapping, filters and reducing. Ruby also supports an alternative syntax for the basic iterating method each, the following three examples are equivalent:

.each do |n|
puts n
end

…and…

for n in 0...42
puts n
end

or even shorter

42.times do |n|
puts n
end

Ruby can also iterate over fixed lists by using Enumerators and either calling their #next method or doing a for each on them, as above.

Rust

With Rust one can iterate on element of vectors, or create own iterators.
Each iterator has adapters.

for n in 0..42

Below the fibonacci is a custom iterator.

for i in fibonacci.skip.take