std.algorithm.iteration

auto cache(Range)(Range range)
if (isInputRange!Range);

auto cacheBidirectional(Range)(Range range)
if (isBidirectionalRange!Range);

cache eagerly evaluates front of range on each construction or call to popFront, to store the result in a cache. The result is then directly returned when front is called, rather than re-evaluated.

This can be a useful function to place in a chain, after functions that have expensive evaluation, as a lazy alternative to std.array.array. In particular, it can be placed after a call to map, or before a call to filter.

cache may provide bidirectional iteration if needed, but since this comes at an increased cost, it must be explicitly requested via the call to cacheBidirectional. Furthermore, a bidirectional cache will evaluate the "center" element twice, when there is only one element left in the range.

cache does not provide random access primitives, as cache would be unable to cache the random accesses. If Range provides slicing primitives, then cache will provide the same slicing primitives, but hasSlicing!Cache will not yield true (as the std.range.primitives.hasSlicing trait also checks for random access).

Parameters:

Range range

an input range

Returns:

an input range with the cached values of range

Examples:

import std.algorithm.comparison : equal;
import std.stdio, std.range;
import std.typecons : tuple;

ulong counter = 0;
double fun(int x)
{
    ++counter;
    // http://en.wikipedia.org/wiki/Quartic_function
    return ( (x + 4.0) * (x + 1.0) * (x - 1.0) * (x - 3.0) ) / 14.0 + 0.5;
}
// Without cache, with array (greedy)
auto result1 = iota(-4, 5).map!(a =>tuple(a, fun(a)))()
                         .filter!(a => a[1] < 0)()
                         .map!(a => a[0])()
                         .array();

// the values of x that have a negative y are:
assert(equal(result1, [-3, -2, 2]));

// Check how many times fun was evaluated.
// As many times as the number of items in both source and result.
assert(counter == iota(-4, 5).length + result1.length);

counter = 0;
// Without array, with cache (lazy)
auto result2 = iota(-4, 5).map!(a =>tuple(a, fun(a)))()
                         .cache()
                         .filter!(a => a[1] < 0)()
                         .map!(a => a[0])();

// the values of x that have a negative y are:
assert(equal(result2, [-3, -2, 2]));

// Check how many times fun was evaluated.
// Only as many times as the number of items in source.
assert(counter == iota(-4, 5).length);

Examples:

Tip: cache is eager when evaluating elements. If calling front on the underlying range has a side effect, it will be observeable before calling front on the actual cached range.

Furthermore, care should be taken composing cache with std.range.take. By placing take before cache, then cache will be "aware" of when the range ends, and correctly stop caching elements when needed. If calling front has no side effect though, placing take after cache may yield a faster range.

Either way, the resulting ranges will be equivalent, but maybe not at the same cost or side effects.

import std.algorithm.comparison : equal;
import std.range;
int i = 0;

auto r = iota(0, 4).tee!((a){i = a;}, No.pipeOnPop);
auto r1 = r.take(3).cache();
auto r2 = r.cache().take(3);

assert(equal(r1, [0, 1, 2]));
assert(i == 2); //The last "seen" element was 2. The data in cache has been cleared.

assert(equal(r2, [0, 1, 2]));
assert(i == 3); //cache has accessed 3. It is still stored internally by cache.

template map(fun...) if (fun.length >= 1)

auto map(Range)(Range r) if (isInputRange!(Unqual!Range));

Implements the homonym function (also known as transform) present in many languages of functional flavor. The call map!(fun)(range) returns a range of which elements are obtained by applying fun(a) left to right for all elements a in range. The original ranges are not changed. Evaluation is done lazily.

Parameters:

fun	one or more functions
r	an input range

Returns:

a range with each fun applied to all the elements. If there is more than one fun, the element type will be Tuple containing one element for each fun.

See Also:

Map (higher-order function)

Examples:

import std.algorithm.comparison : equal;
import std.range : chain;
int[] arr1 = [ 1, 2, 3, 4 ];
int[] arr2 = [ 5, 6 ];
auto squares = map!(a => a * a)(chain(arr1, arr2));
assert(equal(squares, [ 1, 4, 9, 16, 25, 36 ]));

Examples:

Multiple functions can be passed to map. In that case, the element type of map is a tuple containing one element for each function.

auto sums = [2, 4, 6, 8];
auto products = [1, 4, 9, 16];

size_t i = 0;
foreach (result; [ 1, 2, 3, 4 ].map!("a + a", "a * a"))
{
    assert(result[0] == sums[i]);
    assert(result[1] == products[i]);
    ++i;
}

Examples:

You may alias map with some function(s) to a symbol and use it separately:

import std.algorithm.comparison : equal;
import std.conv : to;

alias stringize = map!(to!string);
assert(equal(stringize([ 1, 2, 3, 4 ]), [ "1", "2", "3", "4" ]));

template each(alias pred = "a")

Eagerly iterates over r and calls pred over each element.

If no predicate is specified, each will default to doing nothing but consuming the entire range. .front will be evaluated, but this can be avoided by explicitly specifying a predicate lambda with a lazy parameter.

each also supports opApply-based iterators, so it will work with e.g. std.parallelism.parallel.

Parameters:

pred	predicate to apply to each element of the range
r	range or iterable over which each iterates

See Also:

std.range.tee

Examples:

import std.range : iota;

long[] arr;
iota(5).each!(n => arr ~= n);
assert(arr == [0, 1, 2, 3, 4]);

// If the range supports it, the value can be mutated in place
arr.each!((ref n) => n++);
assert(arr == [1, 2, 3, 4, 5]);

arr.each!"a++";
assert(arr == [2, 3, 4, 5, 6]);

// by-ref lambdas are not allowed for non-ref ranges
static assert(!is(typeof(arr.map!(n => n).each!((ref n) => n++))));

// The default predicate consumes the range
auto m = arr.map!(n => n);
(&m).each();
assert(m.empty);

// Indexes are also available for in-place mutations
arr[] = 0;
arr.each!"a=i"();
assert(arr == [0, 1, 2, 3, 4]);

// opApply iterators work as well
static assert(is(typeof({
    import std.parallelism;
    arr.parallel.each!"a++";
})));

template filter(alias predicate) if (is(typeof(unaryFun!predicate)))

auto filter(Range)(Range rs) if (isInputRange!(Unqual!Range));

Implements the higher order filter function. The predicate is passed to std.functional.unaryFun, and can either accept a string, or any callable that can be executed via pred(element).

Parameters:

predicate	Function to apply to each element of range
range	Input range of elements

Returns:

filter!(predicate)(range) returns a new range containing only elements x in range for which predicate(x) returns true.

See Also:

Filter (higher-order function)

Examples:

import std.algorithm.comparison : equal;
import std.math : approxEqual;
import std.range;

int[] arr = [ 1, 2, 3, 4, 5 ];

// Sum all elements
auto small = filter!(a => a < 3)(arr);
assert(equal(small, [ 1, 2 ]));

// Sum again, but with Uniform Function Call Syntax (UFCS)
auto sum = arr.filter!(a => a < 3);
assert(equal(sum, [ 1, 2 ]));

// In combination with chain() to span multiple ranges
int[] a = [ 3, -2, 400 ];
int[] b = [ 100, -101, 102 ];
auto r = chain(a, b).filter!(a => a > 0);
assert(equal(r, [ 3, 400, 100, 102 ]));

// Mixing convertible types is fair game, too
double[] c = [ 2.5, 3.0 ];
auto r1 = chain(c, a, b).filter!(a => cast(int) a != a);
assert(approxEqual(r1, [ 2.5 ]));

template filterBidirectional(alias pred)

auto filterBidirectional(Range)(Range r) if (isBidirectionalRange!(Unqual!Range));

Similar to filter, except it defines a bidirectional range. There is a speed disadvantage - the constructor spends time finding the last element in the range that satisfies the filtering condition (in addition to finding the first one). The advantage is that the filtered range can be spanned from both directions. Also, std.range.retro can be applied against the filtered range.

The predicate is passed to std.functional.unaryFun, and can either accept a string, or any callable that can be executed via pred(element).

Parameters:

pred	Function to apply to each element of range
r	Bidirectional range of elements

Returns:

a new range containing only the elements in r for which pred returns true.

Examples:

import std.algorithm.comparison : equal;
import std.range;

int[] arr = [ 1, 2, 3, 4, 5 ];
auto small = filterBidirectional!("a < 3")(arr);
static assert(isBidirectionalRange!(typeof(small)));
assert(small.back == 2);
assert(equal(small, [ 1, 2 ]));
assert(equal(retro(small), [ 2, 1 ]));
// In combination with chain() to span multiple ranges
int[] a = [ 3, -2, 400 ];
int[] b = [ 100, -101, 102 ];
auto r = filterBidirectional!("a > 0")(chain(a, b));
assert(r.back == 102);

Group!(pred, Range) group(alias pred = "a == b", Range)(Range r);

Groups consecutively equivalent elements into a single tuple of the element and the number of its repetitions.

Similarly to uniq, group produces a range that iterates over unique consecutive elements of the given range. Each element of this range is a tuple of the element and the number of times it is repeated in the original range. Equivalence of elements is assessed by using the predicate pred, which defaults to "a == b". The predicate is passed to std.functional.binaryFun, and can either accept a string, or any callable that can be executed via pred(element, element).

Parameters:

pred	Binary predicate for determining equivalence of two elements.
Range r	The input range to iterate over.

Returns:

A range of elements of type Tuple!(ElementType!R, uint), representing each consecutively unique element and its respective number of occurrences in that run. This will be an input range if R is an input range, and a forward range in all other cases.

Examples:

import std.algorithm.comparison : equal;
import std.typecons : tuple, Tuple;

int[] arr = [ 1, 2, 2, 2, 2, 3, 4, 4, 4, 5 ];
assert(equal(group(arr), [ tuple(1, 1u), tuple(2, 4u), tuple(3, 1u),
    tuple(4, 3u), tuple(5, 1u) ][]));

auto chunkBy(alias pred, Range)(Range r)
if (isInputRange!Range);

Chunks an input range into subranges of equivalent adjacent elements.

Equivalence is defined by the predicate pred, which can be either binary, which is passed to std.functional.binaryFun, or unary, which is passed to std.functional.unaryFun. In the binary form, two range elements a and b are considered equivalent if pred(a,b) is true. In unary form, two elements are considered equivalent if pred(a) == pred(b) is true.

This predicate must be an equivalence relation, that is, it must be reflexive (pred(x,x) is always true), symmetric (pred(x,y) == pred(y,x)), and transitive (pred(x,y) && pred(y,z) implies pred(x,z)). If this is not the case, the range returned by chunkBy may assert at runtime or behave erratically.

Parameters:

pred	Predicate for determining equivalence.
Range r	The range to be chunked.

Returns:

With a binary predicate, a range of ranges is returned in which all elements in a given subrange are equivalent under the given predicate. With a unary predicate, a range of tuples is returned, with the tuple consisting of the result of the unary predicate for each subrange, and the subrange itself.

Notes: Equivalent elements separated by an intervening non-equivalent element will appear in separate subranges; this function only considers adjacent equivalence. Elements in the subranges will always appear in the same order they appear in the original range.

See Also:

group, which collapses adjacent equivalent elements into a single element.

Examples:

Showing usage with binary predicate:

import std.algorithm.comparison : equal;

// Grouping by particular attribute of each element:
auto data = [
    [1, 1],
    [1, 2],
    [2, 2],
    [2, 3]
];

auto r1 = data.chunkBy!((a,b) => a[0] == b[0]);
assert(r1.equal!equal([
    [[1, 1], [1, 2]],
    [[2, 2], [2, 3]]
]));

auto r2 = data.chunkBy!((a,b) => a[1] == b[1]);
assert(r2.equal!equal([
    [[1, 1]],
    [[1, 2], [2, 2]],
    [[2, 3]]
]));

Examples:

Showing usage with unary predicate:

import std.algorithm.comparison : equal;
import std.typecons : tuple;

// Grouping by particular attribute of each element:
auto range =
[
    [1, 1],
    [1, 1],
    [1, 2],
    [2, 2],
    [2, 3],
    [2, 3],
    [3, 3]
];

auto byX = chunkBy!(a => a[0])(range);
auto expected1 =
[
    tuple(1, [[1, 1], [1, 1], [1, 2]]),
    tuple(2, [[2, 2], [2, 3], [2, 3]]),
    tuple(3, [[3, 3]])
];
foreach (e; byX)
{
    assert(!expected1.empty);
    assert(e[0] == expected1.front[0]);
    assert(e[1].equal(expected1.front[1]));
    expected1.popFront();
}

auto byY = chunkBy!(a => a[1])(range);
auto expected2 =
[
    tuple(1, [[1, 1], [1, 1]]),
    tuple(2, [[1, 2], [2, 2]]),
    tuple(3, [[2, 3], [2, 3], [3, 3]])
];
foreach (e; byY)
{
    assert(!expected2.empty);
    assert(e[0] == expected2.front[0]);
    assert(e[1].equal(expected2.front[1]));
    expected2.popFront();
}

auto joiner(RoR, Separator)(RoR r, Separator sep)
if (isInputRange!RoR && isInputRange!(ElementType!RoR) && isForwardRange!Separator && is(ElementType!Separator : ElementType!(ElementType!RoR)));

auto joiner(RoR)(RoR r)
if (isInputRange!RoR && isInputRange!(ElementType!RoR));

Lazily joins a range of ranges with a separator. The separator itself is a range. If a separator is not provided, then the ranges are joined directly without anything in between them (often called flatten in other languages).

Parameters:

RoR r	An input range of input ranges to be joined.
Separator sep	A forward range of element(s) to serve as separators in the joined range.

Returns:

A range of elements in the joined range. This will be a forward range if both outer and inner ranges of RoR are forward ranges; otherwise it will be only an input range.

See Also:

std.range.chain, which chains a sequence of ranges with compatible elements into a single range.

Examples:

import std.algorithm.comparison : equal;
import std.conv : text;

debug(std_algorithm) scope(success)
    writeln("unittest @", __FILE__, ":", __LINE__, " done.");

static assert(isInputRange!(typeof(joiner([""], ""))));
static assert(isForwardRange!(typeof(joiner([""], ""))));
assert(equal(joiner([""], "xyz"), ""), text(joiner([""], "xyz")));
assert(equal(joiner(["", ""], "xyz"), "xyz"), text(joiner(["", ""], "xyz")));
assert(equal(joiner(["", "abc"], "xyz"), "xyzabc"));
assert(equal(joiner(["abc", ""], "xyz"), "abcxyz"));
assert(equal(joiner(["abc", "def"], "xyz"), "abcxyzdef"));
assert(equal(joiner(["Mary", "has", "a", "little", "lamb"], "..."),
                "Mary...has...a...little...lamb"));
assert(equal(joiner(["abc", "def"]), "abcdef"));

template reduce(fun...) if (fun.length >= 1)

Implements the homonym function (also known as accumulate, compress, inject, or foldl) present in various programming languages of functional flavor. There is also fold which does the same thing but with the opposite parameter order. The call reduce!(fun)(seed, range) first assigns seed to an internal variable result, also called the accumulator. Then, for each element x in range, result = fun(result, x) gets evaluated. Finally, result is returned. The one-argument version reduce!(fun)(range) works similarly, but it uses the first element of the range as the seed (the range must be non-empty).

Returns:

the accumulated result

See Also:

Fold (higher-order function)

fold is functionally equivalent to reduce">reduce with the argument order reversed, and without the need to use tuple for multiple seeds. This makes it easier to use in UFCS chains.

sum is similar to reduce!((a, b) => a + b) that offers precise summing of floating point numbers.

Examples:

Many aggregate range operations turn out to be solved with reduce quickly and easily. The example below illustrates reduce's remarkable power and flexibility.

import std.algorithm.comparison : max, min;
import std.math : approxEqual;
import std.range;

int[] arr = [ 1, 2, 3, 4, 5 ];
// Sum all elements
auto sum = reduce!((a,b) => a + b)(0, arr);
assert(sum == 15);

// Sum again, using a string predicate with "a" and "b"
sum = reduce!"a + b"(0, arr);
assert(sum == 15);

// Compute the maximum of all elements
auto largest = reduce!(max)(arr);
assert(largest == 5);

// Max again, but with Uniform Function Call Syntax (UFCS)
largest = arr.reduce!(max);
assert(largest == 5);

// Compute the number of odd elements
auto odds = reduce!((a,b) => a + (b & 1))(0, arr);
assert(odds == 3);

// Compute the sum of squares
auto ssquares = reduce!((a,b) => a + b * b)(0, arr);
assert(ssquares == 55);

// Chain multiple ranges into seed
int[] a = [ 3, 4 ];
int[] b = [ 100 ];
auto r = reduce!("a + b")(chain(a, b));
assert(r == 107);

// Mixing convertible types is fair game, too
double[] c = [ 2.5, 3.0 ];
auto r1 = reduce!("a + b")(chain(a, b, c));
assert(approxEqual(r1, 112.5));

// To minimize nesting of parentheses, Uniform Function Call Syntax can be used
auto r2 = chain(a, b, c).reduce!("a + b");
assert(approxEqual(r2, 112.5));

Examples:

Sometimes it is very useful to compute multiple aggregates in one pass. One advantage is that the computation is faster because the looping overhead is shared. That's why reduce accepts multiple functions. If two or more functions are passed, reduce returns a std.typecons.Tuple object with one member per passed-in function. The number of seeds must be correspondingly increased.

import std.algorithm.comparison : max, min;
import std.math : approxEqual, sqrt;
import std.typecons : tuple, Tuple;

double[] a = [ 3.0, 4, 7, 11, 3, 2, 5 ];
// Compute minimum and maximum in one pass
auto r = reduce!(min, max)(a);
// The type of r is Tuple!(int, int)
assert(approxEqual(r[0], 2));  // minimum
assert(approxEqual(r[1], 11)); // maximum

// Compute sum and sum of squares in one pass
r = reduce!("a + b", "a + b * b")(tuple(0.0, 0.0), a);
assert(approxEqual(r[0], 35));  // sum
assert(approxEqual(r[1], 233)); // sum of squares
// Compute average and standard deviation from the above
auto avg = r[0] / a.length;
auto stdev = sqrt(r[1] / a.length - avg * avg);

auto reduce(R)(R r)
if (isIterable!R);

No-seed version. The first element of r is used as the seed's value.

For each function f in fun, the corresponding seed type S is Unqual!(typeof(f(e, e))), where e is an element of r: ElementType!R for ranges, and ForeachType!R otherwise.

Once S has been determined, then S s = e; and s = f(s, e); must both be legal.

If r is empty, an Exception is thrown.

Parameters:

fun	one or more functions
R r	an iterable value as defined by isIterable

Returns:

the final result of the accumulator applied to the iterable

auto reduce(S, R)(S seed, R r)
if (isIterable!R);

Seed version. The seed should be a single value if fun is a single function. If fun is multiple functions, then seed should be a std.typecons.Tuple, with one field per function in f.

For convenience, if the seed is const, or has qualified fields, then reduce will operate on an unqualified copy. If this happens then the returned type will not perfectly match S.

Use fold instead of reduce to use the seed version in a UFCS chain.

Parameters:

fun	one or more functions
S seed	the initial value of the accumulator
R r	an iterable value as defined by isIterable

Returns:

the final result of the accumulator applied to the iterable

template fold(fun...) if (fun.length >= 1)

Implements the homonym function (also known as accumulate, compress, inject, or foldl) present in various programming languages of functional flavor. The call fold!(fun)(range, seed) first assigns seed to an internal variable result, also called the accumulator. Then, for each element x in range, result = fun(result, x) gets evaluated. Finally, result is returned. The one-argument version fold!(fun)(range) works similarly, but it uses the first element of the range as the seed (the range must be non-empty).

Returns:

the accumulated result

See Also:

Fold (higher-order function)

sum is similar to fold!((a, b) => a + b) that offers precise summing of floating point numbers.

This is functionally equivalent to reduce with the argument order reversed, and without the need to use tuple for multiple seeds.

Examples:

immutable arr = [1, 2, 3, 4, 5];

// Sum all elements
assert(arr.fold!((a, b) => a + b) == 15);

// Sum all elements with explicit seed
assert(arr.fold!((a, b) => a + b)(6) == 21);

import std.algorithm.comparison: min, max;
import std.typecons: tuple;

// Compute minimum and maximum at the same time
assert(arr.fold!(min, max) == tuple(1, 5));

// Compute minimum and maximum at the same time with seeds
assert(arr.fold!(min, max)(0, 7) == tuple(0, 7));

// Can be used in a UFCS chain
assert(arr.map!(a => a + 1).fold!((a, b) => a + b) == 20);

auto splitter(alias pred = "a == b", Range, Separator)(Range r, Separator s)
if (is(typeof(binaryFun!pred(r.front, s)) : bool) && (hasSlicing!Range && hasLength!Range || isNarrowString!Range));

Lazily splits a range using an element as a separator. This can be used with any narrow string type or sliceable range type, but is most popular with string types.

Two adjacent separators are considered to surround an empty element in the split range. Use filter!(a => !a.empty) on the result to compress empty elements.

The predicate is passed to std.functional.binaryFun, and can either accept a string, or any callable that can be executed via pred(element, s).

If the empty range is given, the result is a range with one empty element. If a range with one separator is given, the result is a range with two empty elements.

If splitting a string on whitespace and token compression is desired, consider using splitter without specifying a separator (see fourth overload below).

Parameters:

pred	The predicate for comparing each element with the separator, defaulting to "a == b".
Range r	The input range to be split. Must support slicing and .length.
Separator s	The element to be treated as the separator between range segments to be split.

Constraints: The predicate pred needs to accept an element of r and the separator s.

Returns:

An input range of the subranges of elements between separators. If r is a forward range or bidirectional range, the returned range will be likewise.

See Also:

std.regex.splitter for a version that splits using a regular expression defined separator.

Examples:

import std.algorithm.comparison : equal;

assert(equal(splitter("hello  world", ' '), [ "hello", "", "world" ]));
int[] a = [ 1, 2, 0, 0, 3, 0, 4, 5, 0 ];
int[][] w = [ [1, 2], [], [3], [4, 5], [] ];
assert(equal(splitter(a, 0), w));
a = [ 0 ];
assert(equal(splitter(a, 0), [ (int[]).init, (int[]).init ]));
a = [ 0, 1 ];
assert(equal(splitter(a, 0), [ [], [1] ]));
w = [ [0], [1], [2] ];
assert(equal(splitter!"a.front == b"(w, 1), [ [[0]], [[2]] ]));

auto splitter(alias pred = "a == b", Range, Separator)(Range r, Separator s)
if (is(typeof(binaryFun!pred(r.front, s.front)) : bool) && (hasSlicing!Range || isNarrowString!Range) && isForwardRange!Separator && (hasLength!Separator || isNarrowString!Separator));

Similar to the previous overload of splitter, except this one uses another range as a separator. This can be used with any narrow string type or sliceable range type, but is most popular with string types. The predicate is passed to std.functional.binaryFun, and can either accept a string, or any callable that can be executed via pred(r.front, s.front).

Two adjacent separators are considered to surround an empty element in the split range. Use filter!(a => !a.empty) on the result to compress empty elements.

Parameters:

pred	The predicate for comparing each element with the separator, defaulting to "a == b".
Range r	The input range to be split.
Separator s	The forward range to be treated as the separator between segments of r to be split.

Constraints: The predicate pred needs to accept an element of r and an element of s.

Returns:

An input range of the subranges of elements between separators. If r is a forward range or bidirectional range, the returned range will be likewise.

See Also:

std.regex.splitter for a version that splits using a regular expression defined separator.

Examples:

import std.algorithm.comparison : equal;

assert(equal(splitter("hello  world", "  "), [ "hello", "world" ]));
int[] a = [ 1, 2, 0, 0, 3, 0, 4, 5, 0 ];
int[][] w = [ [1, 2], [3, 0, 4, 5, 0] ];
assert(equal(splitter(a, [0, 0]), w));
a = [ 0, 0 ];
assert(equal(splitter(a, [0, 0]), [ (int[]).init, (int[]).init ]));
a = [ 0, 0, 1 ];
assert(equal(splitter(a, [0, 0]), [ [], [1] ]));

auto splitter(alias isTerminator, Range)(Range input)
if (isForwardRange!Range && is(typeof(unaryFun!isTerminator(input.front))));

Similar to the previous overload of splitter, except this one does not use a separator. Instead, the predicate is an unary function on the input range's element type. The isTerminator predicate is passed to std.functional.unaryFun and can either accept a string, or any callable that can be executed via pred(element, s).

Two adjacent separators are considered to surround an empty element in the split range. Use filter!(a => !a.empty) on the result to compress empty elements.

Parameters:

isTerminator	The predicate for deciding where to split the range.
Range input	The input range to be split.

Constraints: The predicate isTerminator needs to accept an element of input.

Returns:

An input range of the subranges of elements between separators. If input is a forward range or bidirectional range, the returned range will be likewise.

See Also:

std.regex.splitter for a version that splits using a regular expression defined separator.

Examples:

import std.algorithm.comparison : equal;

assert(equal(splitter!(a => a == ' ')("hello  world"), [ "hello", "", "world" ]));
int[] a = [ 1, 2, 0, 0, 3, 0, 4, 5, 0 ];
int[][] w = [ [1, 2], [], [3], [4, 5], [] ];
assert(equal(splitter!(a => a == 0)(a), w));
a = [ 0 ];
assert(equal(splitter!(a => a == 0)(a), [ (int[]).init, (int[]).init ]));
a = [ 0, 1 ];
assert(equal(splitter!(a => a == 0)(a), [ [], [1] ]));
w = [ [0], [1], [2] ];
assert(equal(splitter!(a => a.front == 1)(w), [ [[0]], [[2]] ]));

auto splitter(C)(C[] s)
if (isSomeChar!C);

Lazily splits the string s into words, using whitespace as the delimiter.

This function is string specific and, contrary to splitter!(std.uni.isWhite), runs of whitespace will be merged together (no empty tokens will be produced).

Parameters:

C[] s

The string to be split.

Returns:

An input range of slices of the original string split by whitespace.

Examples:

import std.algorithm.comparison : equal;
auto a = " a     bcd   ef gh ";
assert(equal(splitter(a), ["a", "bcd", "ef", "gh"][]));

auto sum(R)(R r)
if (isInputRange!R && !isInfinite!R && is(typeof(r.front + r.front)));

auto sum(R, E)(R r, E seed)
if (isInputRange!R && !isInfinite!R && is(typeof(seed = seed + r.front)));

Sums elements of r, which must be a finite input range. Although conceptually sum(r) is equivalent to fold!((a, b) => a + b)(r, 0), sum uses specialized algorithms to maximize accuracy, as follows.

• If std.range.primitives.ElementType!R is a floating-point type and R is a random-access range with length and slicing, then sum uses the pairwise summation algorithm.
• If ElementType!R is a floating-point type and R is a finite input range (but not a random-access range with slicing), then sum uses the Kahan summation algorithm.
• In all other cases, a simple element by element addition is done.

For floating point inputs, calculations are made in real precision for real inputs and in double precision otherwise (Note this is a special case that deviates from fold's behavior, which would have kept float precision for a float range). For all other types, the calculations are done in the same type obtained from from adding two elements of the range, which may be a different type from the elements themselves (for example, in case of integral promotion).

A seed may be passed to sum. Not only will this seed be used as an initial value, but its type will override all the above, and determine the algorithm and precision used for summation.

Note that these specialized summing algorithms execute more primitive operations than vanilla summation. Therefore, if in certain cases maximum speed is required at expense of precision, one can use fold!((a, b) => a + b)(r, 0), which is not specialized for summation.

Parameters:

E seed	the initial value of the summation
R r	a finite input range

Returns:

The sum of all the elements in the range r.

auto uniq(alias pred = "a == b", Range)(Range r)
if (isInputRange!Range && is(typeof(binaryFun!pred(r.front, r.front)) == bool));

Lazily iterates unique consecutive elements of the given range (functionality akin to the uniq system utility). Equivalence of elements is assessed by using the predicate pred, by default "a == b". The predicate is passed to std.functional.binaryFun, and can either accept a string, or any callable that can be executed via pred(element, element). If the given range is bidirectional, uniq also yields a bidirectional range.

Parameters:

pred	Predicate for determining equivalence between range elements.
Range r	An input range of elements to filter.

Returns:

An input range of consecutively unique elements in the original range. If r is also a forward range or bidirectional range, the returned range will be likewise.

Examples:

import std.algorithm.mutation : copy;
import std.algorithm.comparison : equal;

int[] arr = [ 1, 2, 2, 2, 2, 3, 4, 4, 4, 5 ];
assert(equal(uniq(arr), [ 1, 2, 3, 4, 5 ][]));

// Filter duplicates in-place using copy
arr.length -= arr.uniq().copy(arr).length;
assert(arr == [ 1, 2, 3, 4, 5 ]);

// Note that uniqueness is only determined consecutively; duplicated
// elements separated by an intervening different element will not be
// eliminated:
assert(equal(uniq([ 1, 1, 2, 1, 1, 3, 1]), [1, 2, 1, 3, 1]));

Permutations!Range permutations(Range)(Range r)
if (isRandomAccessRange!Range && hasLength!Range);

Lazily computes all permutations of r using Heap's algorithm.

Returns:

A forward range the elements of which are an std.range.indexed view into r.

See Also:

std.algorithm.sorting.nextPermutation.

Examples:

import std.algorithm.comparison : equal;
import std.range : iota;
assert(equal!equal(iota(3).permutations,
    [[0, 1, 2],
     [1, 0, 2],
     [2, 0, 1],
     [0, 2, 1],
     [1, 2, 0],
     [2, 1, 0]]));