Programming in D – Tutorial and Reference
Ali Çehreli

Other D Resources

foreach Loop

One of the most common statements in D is the foreach loop. It is used for applying the same operations to every element of a container (or a range).

Operations that are applied to elements of containers are very common in programming. We have seen in the for Loop chapter that elements of an array are accessed in a for loop by an index value that is incremented at each iteration: 

    for (int i = 0; i != array.length; ++i) {
        writeln(array[i]);
    }

The following steps are involved in iterating over all the elements:

foreach has essentially the same behavior but it simplifies the code by handling those steps automatically: 

    foreach (element; array) {
        writeln(element);
    }

Part of the power of foreach comes from the fact that it can be used the same way regardless of the type of the container. As we have seen in the previous chapter, one way of iterating over the values of an associative array in a for loop is by first calling the array's .values property: 

    auto values = aa.values;
    for (int i = 0; i != values.length; ++i) {
        writeln(values[i]);
    }

foreach does not require anything special for associative arrays; it is used exactly the same as with arrays: 

    foreach (value; aa) {
        writeln(value);
    }
The foreach syntax

foreach consists of three sections: 

    foreach (names; container_or_range) {
        operations
    }
continue and break

These keywords have the same meaning as they do for the for loop: continue moves to the next iteration before completing the rest of the operations for the current element, and break terminates the loop altogether.

foreach with arrays

When using foreach with plain arrays and there is a single name specified in the names section, that name represents the value of the element at each iteration: 

    foreach (element; array) {
        writeln(element);
    }

When two names are specified in the names section, they represent an automatic counter and the value of the element, respectively: 

    foreach (i, element; array) {
        writeln(i, ": ", element);
    }

The counter is incremented automatically by foreach. Although it can be named anything else, i is a very common name for the automatic counter.

foreach with strings and std.range.stride

Since strings are arrays of characters, foreach works with strings the same way as it does with arrays: A single name refers to the character, two names refer to the counter and the character, respectively: 

    foreach (c; "hello") {
        writeln(c);
    }

    foreach (i, c; "hello") {
        writeln(i, ": ", c);
    }

However, being UTF code units, char and wchar iterate over UTF code units, not Unicode code points: 

    foreach (i, code; "abcçd") {
        writeln(i, ": ", code);
    }

The two UTF-8 code units that make up ç would be accessed as separate elements: 

0: a
1: b
2: c
3: 
4: �
5: d

One way of iterating over Unicode characters of strings in a foreach loop is stride from the std.range module. stride presents the string as a container that consists of Unicode characters. Its second parameter is the number of steps that it should take as it strides over the characters: 

import std.range;

// ...

    foreach (c; stride("abcçd", 1)) {
        writeln(c);
    }

Regardless of the character type of the string, stride always presents its elements as Unicode characters: 

a
b
c
ç
d

I will explain below why this loop could not include an automatic counter.

foreach with associative arrays

When using foreach with associative arrays, a single name refers to the value, while two names refer to the key and the value, respectively: 

    foreach (value; aa) {
        writeln(value);
    }

    foreach (key, value; aa) {
        writeln(key, ": ", value);
    }

Associative arrays can provide their keys and values as ranges as well. We will see ranges in a later chapter. .byKey, .byValue, and .byKeyValue return efficient range objects that are useful in contexts other than foreach loops as well.

.byValue does not bring any benefit in foreach loops over the regular value iteration above. On the other hand, .byKey is the only efficient way of iterating over just the keys of an associative array: 

    foreach (key; aa.byKey) {
        writeln(key);
    }

.byKeyValue provides each key-value element through a variable that is similar to a tuple. The key and the value are accessed separately through the .key and .value properties of that variable: 

    foreach (element; aa.byKeyValue) {
        writefln("The value for key %s is %s",
                 element.key, element.value);
    }
foreach with number ranges

We have seen number ranges before, in the Slices and Other Array Features chapter. It is possible to specify a number range in the container_or_range section: 

    foreach (number; 10..15) {
        writeln(number);
    }

Remember that 10 would be included in the range but 15 would not be.

foreach with structs, classes, and ranges

foreach can also be used with objects of user-defined types that define their own iteration in foreach loops. Structs and classes provide support for foreach iteration either by their opApply() member functions, or by a set of range member functions. We will see these features in later chapters.

The counter is automatic only for arrays

The automatic counter is provided only when iterating over arrays. There are two options for other containers

Such a variable is needed when counting a specific condition as well. For example, the following code counts only the values that are divisible by 10: 

import std.stdio;

void main() {
    auto numbers = [ 1, 0, 15, 10, 3, 5, 20, 30 ];

    size_t count = 0;
    foreach (number; numbers) {
        if ((number % 10) == 0) {
            ++count;
            write(count);

        } else {
            write(' ');
        }

        writeln(": ", number);
    }
}

The output: 

 : 1
1: 0
 : 15
2: 10
 : 3
 : 5
3: 20
4: 30
The copy of the element, not the element itself

The foreach loop normally provides a copy of the element, not the actual element that is stored in the container. This may be a cause of bugs.

To see an example of this, let's have a look at the following program that is trying to double the values of the elements of an array: 

import std.stdio;

void main() {
    double[] numbers = [ 1.2, 3.4, 5.6 ];

    writefln("Before: %s", numbers);

    foreach (number; numbers) {
        number *= 2;
    }

    writefln("After : %s", numbers);
}

The output of the program indicates that the assignment made to each element inside the foreach body does not have any effect on the elements of the container: 

Before: [1.2, 3.4, 5.6]
After : [1.2, 3.4, 5.6]

That is because number is not an actual element of the array, but a copy of each element. When the actual elements need to be operated on, the name must be defined as a reference of the actual element, by using the ref keyword: 

    foreach (ref number; numbers) {
        number *= 2;
    }

The new output shows that the assignments now modify the actual elements of the array: 

Before: [1.2, 3.4, 5.6]
After : [2.4, 6.8, 11.2]

The ref keyword makes number an alias of the actual element at each iteration. As a result, the modifications through number modify that actual element of the container.

The integrity of the container must be preserved

Although it is fine to modify the elements of a container through ref variables, the structure of a container must not be changed during its iteration. For example, elements must not be removed nor added to the container during a foreach loop.

Such modifications may confuse the inner workings of the loop iteration and result in incorrect program states.

foreach_reverse to iterate in the reverse direction

foreach_reverse works the same way as foreach except it iterates in the reverse direction: 

    auto container = [ 1, 2, 3 ];

    foreach_reverse (element; container) {
        writefln("%s ", element);
    }

The output: 

3 
2 
1

The use of foreach_reverse is not common because the range function retro() achieves the same goal. We will see retro() in a later chapter.

Exercise

We know that associative arrays provide a mapping from keys to values. This mapping is unidirectional: values are accessed by keys but not the other way around.

Assume that there is already the following associative array: 

    string[int] names = [ 1:"one", 7:"seven", 20:"twenty" ];

Use that associative array and a foreach loop to fill another associative array named values. The new associative array should provide values that correspond to names. For example, the following line should print 20: 

    writeln(values["twenty"]);

... the solution