array.d

// Written in the D programming language.
/**
Functions and types that manipulate built-in arrays and associative arrays.

This module provides all kinds of functions to create, manipulate or convert arrays:

$(SCRIPT inhibitQuickIndex = 1;)
$(DIVC quickindex,
$(BOOKTABLE ,
$(TR $(TH Function Name) $(TH Description)
)
    $(TR $(TD $(LREF array))
        $(TD Returns a copy of the input in a newly allocated dynamic array.
    ))
    $(TR $(TD $(LREF appender))
        $(TD Returns a new $(LREF Appender) or $(LREF RefAppender) initialized with a given array.
    ))
    $(TR $(TD $(LREF assocArray))
        $(TD Returns a newly allocated associative array from a range of key/value tuples.
    ))
    $(TR $(TD $(LREF byPair))
        $(TD Construct a range iterating over an associative array by key/value tuples.
    ))
    $(TR $(TD $(LREF insertInPlace))
        $(TD Inserts into an existing array at a given position.
    ))
    $(TR $(TD $(LREF join))
        $(TD Concatenates a range of ranges into one array.
    ))
    $(TR $(TD $(LREF minimallyInitializedArray))
        $(TD Returns a new array of type `T`.
    ))
    $(TR $(TD $(LREF replace))
        $(TD Returns a new array with all occurrences of a certain subrange replaced.
    ))
    $(TR $(TD $(LREF replaceFirst))
        $(TD Returns a new array with the first occurrence of a certain subrange replaced.
    ))
    $(TR $(TD $(LREF replaceInPlace))
        $(TD Replaces all occurrences of a certain subrange and puts the result into a given array.
    ))
    $(TR $(TD $(LREF replaceInto))
        $(TD Replaces all occurrences of a certain subrange and puts the result into an output range.
    ))
    $(TR $(TD $(LREF replaceLast))
        $(TD Returns a new array with the last occurrence of a certain subrange replaced.
    ))
    $(TR $(TD $(LREF replaceSlice))
        $(TD Returns a new array with a given slice replaced.
    ))
    $(TR $(TD $(LREF replicate))
        $(TD Creates a new array out of several copies of an input array or range.
    ))
    $(TR $(TD $(LREF sameHead))
        $(TD Checks if the initial segments of two arrays refer to the same
        place in memory.
    ))
    $(TR $(TD $(LREF sameTail))
        $(TD Checks if the final segments of two arrays refer to the same place
        in memory.
    ))
    $(TR $(TD $(LREF split))
        $(TD Eagerly split a range or string into an array.
    ))
    $(TR $(TD $(LREF staticArray))
        $(TD Creates a new static array from given data.
    ))
    $(TR $(TD $(LREF uninitializedArray))
        $(TD Returns a new array of type `T` without initializing its elements.
    ))
))

Copyright: Copyright Andrei Alexandrescu 2008- and Jonathan M Davis 2011-.

License:   $(HTTP boost.org/LICENSE_1_0.txt, Boost License 1.0).

Authors:   $(HTTP erdani.org, Andrei Alexandrescu) and
           $(HTTP jmdavisprog.com, Jonathan M Davis)

Source: $(PHOBOSSRC std/array.d)
*/
module std.array;

import std.functional;
import std.meta;
import std.traits;

import std.range.primitives;
public import std.range.primitives : save, empty, popFront, popBack, front, back;

/**
 * Allocates an array and initializes it with copies of the elements
 * of range `r`.
 *
 * Narrow strings are handled as follows:
 * - If autodecoding is turned on (default), then they are handled as a separate overload.
 * - If autodecoding is turned off, then this is equivalent to duplicating the array.
 *
 * Params:
 *      r = range (or aggregate with `opApply` function) whose elements are copied into the allocated array
 * Returns:
 *      allocated and initialized array
 */
ForeachType!Range[] array(Range)(Range r)
if (isIterable!Range && !isAutodecodableString!Range && !isInfinite!Range)
{
    if (__ctfe)
    {
        // Compile-time version to avoid memcpy calls.
        // Also used to infer attributes of array().
        typeof(return) result;
        foreach (e; r)
            result ~= e;
        return result;
    }

    alias E = ForeachType!Range;
    static if (hasLength!Range)
    {
        const length = r.length;
        if (length == 0)
            return null;

        import core.internal.lifetime : emplaceRef;

        auto result = (() @trusted => uninitializedArray!(Unqual!E[])(length))();

        // Every element of the uninitialized array must be initialized
        size_t cnt; //Number of elements that have been initialized
        try
        {
            foreach (e; r)
            {
                emplaceRef!E(result[cnt], e);
                ++cnt;
            }
        } catch (Exception e)
        {
            //https://issues.dlang.org/show_bug.cgi?id=22185
            //Make any uninitialized elements safely destructible.
            foreach (ref elem; result[cnt..$])
            {
                import core.internal.lifetime : emplaceInitializer;
                emplaceInitializer(elem);
            }
            throw e;
        }
        /*
            https://issues.dlang.org/show_bug.cgi?id=22673

            We preallocated an array, we should ensure that enough range elements
            were gathered such that every slot in the array is filled. If not, the GC
            will collect the allocated array, leading to the `length - cnt` left over elements
            being collected too - despite their contents having no guarantee of destructibility.
         */
        assert(length == cnt,
               "Range .length property was not equal to the length yielded by the range before becoming empty");
        return (() @trusted => cast(E[]) result)();
    }
    else
    {
        auto a = appender!(E[])();
        foreach (e; r)
        {
            a.put(e);
        }
        return a.data;
    }
}

/// ditto
ForeachType!(typeof((*Range).init))[] array(Range)(Range r)
if (is(Range == U*, U) && isIterable!U && !isAutodecodableString!Range && !isInfinite!Range)
{
    return array(*r);
}

///
@safe pure nothrow unittest
{
    auto a = array([1, 2, 3, 4, 5][]);
    assert(a == [ 1, 2, 3, 4, 5 ]);
}

@safe pure nothrow unittest
{
    import std.algorithm.comparison : equal;
    struct Foo
    {
        int a;
    }
    auto a = array([Foo(1), Foo(2), Foo(3), Foo(4), Foo(5)][]);
    assert(equal(a, [Foo(1), Foo(2), Foo(3), Foo(4), Foo(5)]));
}

@safe pure nothrow unittest
{
    struct MyRange
    {
        enum front = 123;
        enum empty = true;
        void popFront() {}
    }

    auto arr = (new MyRange).array;
    assert(arr.empty);
}

@safe pure nothrow unittest
{
    immutable int[] a = [1, 2, 3, 4];
    auto b = (&a).array;
    assert(b == a);
}

@safe pure nothrow unittest
{
    import std.algorithm.comparison : equal;
    struct Foo
    {
        int a;
        noreturn opAssign(Foo)
        {
            assert(0);
        }
        auto opEquals(Foo foo)
        {
            return a == foo.a;
        }
    }
    auto a = array([Foo(1), Foo(2), Foo(3), Foo(4), Foo(5)][]);
    assert(equal(a, [Foo(1), Foo(2), Foo(3), Foo(4), Foo(5)]));
}

// https://issues.dlang.org/show_bug.cgi?id=12315
@safe pure nothrow unittest
{
    static struct Bug12315 { immutable int i; }
    enum bug12315 = [Bug12315(123456789)].array();
    static assert(bug12315[0].i == 123456789);
}

@safe pure nothrow unittest
{
    import std.range;
    static struct S{int* p;}
    auto a = array(immutable(S).init.repeat(5));
    assert(a.length == 5);
}

// https://issues.dlang.org/show_bug.cgi?id=18995
@system unittest
{
    import core.memory : __delete;
    int nAlive = 0;
    struct S
    {
        bool alive;
        this(int) { alive = true; ++nAlive; }
        this(this) { nAlive += alive; }
        ~this() { nAlive -= alive; alive = false; }
    }

    import std.algorithm.iteration : map;
    import std.range : iota;

    auto arr = iota(3).map!(a => S(a)).array;
    assert(nAlive == 3);

    // No good way to ensure the GC frees this, just call the lifetime function
    // directly.
    __delete(arr);

    assert(nAlive == 0);
}

@safe pure nothrow @nogc unittest
{
    //Turn down infinity:
    static assert(!is(typeof(
        repeat(1).array()
    )));
}

// https://issues.dlang.org/show_bug.cgi?id=20937
@safe pure nothrow unittest
{
    struct S {int* x;}
    struct R
    {
        immutable(S) front;
        bool empty;
        @safe pure nothrow void popFront(){empty = true;}
    }
    R().array;
}

/**
Convert a narrow autodecoding string to an array type that fully supports
random access.  This is handled as a special case and always returns an array
of `dchar`

NOTE: This function is never used when autodecoding is turned off.

Params:
    str = `isNarrowString` to be converted to an array of `dchar`
Returns:
    a `dchar[]`, `const(dchar)[]`, or `immutable(dchar)[]` depending on the constness of
    the input.
*/
CopyTypeQualifiers!(ElementType!String,dchar)[] array(String)(scope String str)
if (isAutodecodableString!String)
{
    import std.utf : toUTF32;
    auto temp = str.toUTF32;
    /* Unsafe cast. Allowed because toUTF32 makes a new array
       and copies all the elements.
     */
    return () @trusted { return cast(CopyTypeQualifiers!(ElementType!String, dchar)[]) temp; } ();
}

///
@safe pure nothrow unittest
{
    import std.range.primitives : isRandomAccessRange;
    import std.traits : isAutodecodableString;

    // note that if autodecoding is turned off, `array` will not transcode these.
    static if (isAutodecodableString!string)
        assert("Hello D".array == "Hello D"d);
    else
        assert("Hello D".array == "Hello D");

    static if (isAutodecodableString!wstring)
        assert("Hello D"w.array == "Hello D"d);
    else
        assert("Hello D"w.array == "Hello D"w);

    static assert(isRandomAccessRange!dstring == true);
}

@safe unittest
{
    import std.conv : to;

    static struct TestArray { int x; string toString() @safe { return to!string(x); } }

    static struct OpAssign
    {
        uint num;
        this(uint num) { this.num = num; }

        // Templating opAssign to make sure the bugs with opAssign being
        // templated are fixed.
        void opAssign(T)(T rhs) { this.num = rhs.num; }
    }

    static struct OpApply
    {
        int opApply(scope int delegate(ref int) @safe dg)
        {
            int res;
            foreach (i; 0 .. 10)
            {
                res = dg(i);
                if (res) break;
            }

            return res;
        }
    }

    auto a = array([1, 2, 3, 4, 5][]);
    assert(a == [ 1, 2, 3, 4, 5 ]);

    auto b = array([TestArray(1), TestArray(2)][]);
    assert(b == [TestArray(1), TestArray(2)]);

    class C
    {
        int x;
        this(int y) { x = y; }
        override string toString() const @safe { return to!string(x); }
    }
    auto c = array([new C(1), new C(2)][]);
    assert(c[0].x == 1);
    assert(c[1].x == 2);

    auto d = array([1.0, 2.2, 3][]);
    assert(is(typeof(d) == double[]));
    assert(d == [1.0, 2.2, 3]);

    auto e = [OpAssign(1), OpAssign(2)];
    auto f = array(e);
    assert(e == f);

    assert(array(OpApply.init) == [0,1,2,3,4,5,6,7,8,9]);
    static if (isAutodecodableString!string)
    {
        assert(array("ABC") == "ABC"d);
        assert(array("ABC".dup) == "ABC"d.dup);
    }
}

// https://issues.dlang.org/show_bug.cgi?id=8233
@safe pure nothrow unittest
{
    assert(array("hello world"d) == "hello world"d);
    immutable a = [1, 2, 3, 4, 5];
    assert(array(a) == a);
    const b = a;
    assert(array(b) == a);

    //To verify that the opAssign branch doesn't get screwed up by using Unqual.
    //EDIT: array no longer calls opAssign.
    struct S
    {
        ref S opAssign(S)(const ref S rhs)
        {
            assert(0);
        }

        int i;
    }

    static foreach (T; AliasSeq!(S, const S, immutable S))
    {{
        auto arr = [T(1), T(2), T(3), T(4)];
        assert(array(arr) == arr);
    }}
}

// https://issues.dlang.org/show_bug.cgi?id=9824
@safe pure nothrow unittest
{
    static struct S
    {
        @disable void opAssign(S);
        int i;
    }
    auto arr = [S(0), S(1), S(2)];
    arr.array();
}

// https://issues.dlang.org/show_bug.cgi?id=10220
@safe pure nothrow unittest
{
    import std.algorithm.comparison : equal;
    import std.exception;
    import std.range : repeat;

    static struct S
    {
        int val;

        @disable this();
        this(int v) { val = v; }
    }
    assertCTFEable!(
    {
        auto r = S(1).repeat(2).array();
        assert(equal(r, [S(1), S(1)]));
    });
}
//https://issues.dlang.org/show_bug.cgi?id=22673
@system unittest
{
    struct LyingRange
    {
        enum size_t length = 100;
        enum theRealLength = 50;
        size_t idx = 0;
        bool empty()
        {
            return idx <= theRealLength;
        }
        void popFront()
        {
            ++idx;
        }
        size_t front()
        {
            return idx;
        }
    }
    static assert(hasLength!LyingRange);
    LyingRange rng;
    import std.exception : assertThrown;
    assertThrown!Error(array(rng));
}
//https://issues.dlang.org/show_bug.cgi?id=22185
@system unittest
{
    import std.stdio;
    static struct ThrowingCopy
    {
        int x = 420;
        this(ref return scope ThrowingCopy rhs)
        {
            rhs.x = 420;
            //
            throw new Exception("This throws");
        }
        ~this()
        {
            /*
                Any time this destructor runs, it should be running on "valid"
                data. This is is mimicked by having a .init other than 0 (the value the memory
                practically will be from the GC).
            */
            if (x != 420)
            {
                //This will only trigger during GC finalization so avoid writefln for now.
                printf("Destructor failure in ThrowingCopy(%d) @ %p", x, &this);
                assert(x == 420, "unittest destructor failed");
            }
        }
    }
    static struct LyingThrowingRange
    {
        enum size_t length = 100;
        enum size_t evilRealLength = 50;
        size_t idx;
        ThrowingCopy front()
        {
            return ThrowingCopy(12);
        }
        bool empty()
        {
            return idx == evilRealLength;
        }
        void popFront()
        {
            ++idx;
        }
    }
    static assert(hasLength!LyingThrowingRange);
    import std.exception : assertThrown;
    {
        assertThrown(array(LyingThrowingRange()));
    }
    import core.memory : GC;
    /*
        Force a collection early. Doesn't always actually finalize the bad objects
        but trying to collect soon after the allocation is thrown away means any potential failures
        will happen earlier.
    */
    GC.collect();
}

/**
Returns a newly allocated associative array from a range of key/value tuples
or from a range of keys and a range of values.

Params:
    r = An $(REF_ALTTEXT input range, isInputRange, std,range,primitives)
    of tuples of keys and values.
    keys = An $(REF_ALTTEXT input range, isInputRange, std,range,primitives) of keys
    values = An $(REF_ALTTEXT input range, isInputRange, std,range,primitives) of values

Returns:

    A newly allocated associative array out of elements of the input
    range, which must be a range of tuples (Key, Value) or
    a range of keys and a range of values. If given two ranges of unequal
    lengths after the elements of the shorter are exhausted the remaining
    elements of the longer will not be considered.
    Returns a null associative array reference when given an empty range.
    Duplicates: Associative arrays have unique keys. If r contains duplicate keys,
    then the result will contain the value of the last pair for that key in r.

See_Also: $(REF Tuple, std,typecons), $(REF zip, std,range)
 */

auto assocArray(Range)(Range r)
if (isInputRange!Range)
{
    import std.typecons : isTuple;

    alias E = ElementType!Range;
    static assert(isTuple!E, "assocArray: argument must be a range of tuples,"
        ~" but was a range of "~E.stringof);
    static assert(E.length == 2, "assocArray: tuple dimension must be 2");
    alias KeyType = E.Types[0];
    alias ValueType = E.Types[1];
    static assert(isMutable!ValueType, "assocArray: value type must be mutable");

    ValueType[KeyType] aa;
    foreach (ref t; r)
        aa[t[0]] = t[1];
    return aa;
}

/// ditto
auto assocArray(Keys, Values)(Keys keys, Values values)
if (isInputRange!Values && isInputRange!Keys)
{
    static if (isDynamicArray!Keys && isDynamicArray!Values
        && !isNarrowString!Keys && !isNarrowString!Values)
    {
        void* aa;
        {
            // aaLiteral is nothrow when the destructors don't throw
            static if (is(typeof(() nothrow
            {
                import std.range : ElementType;
                import std.traits : hasElaborateDestructor;
                alias KeyElement = ElementType!Keys;
                static if (hasElaborateDestructor!KeyElement)
                    KeyElement.init.__xdtor();

                alias ValueElement = ElementType!Values;
                static if (hasElaborateDestructor!ValueElement)
                    ValueElement.init.__xdtor();
            })))
            {
                scope(failure) assert(false, "aaLiteral must not throw");
            }
            if (values.length > keys.length)
                values = values[0 .. keys.length];
            else if (keys.length > values.length)
                keys = keys[0 .. values.length];
            aa = aaLiteral(keys, values);
        }
        alias Key = typeof(keys[0]);
        alias Value = typeof(values[0]);
        return (() @trusted => cast(Value[Key]) aa)();
    }
    else
    {
        // zip is not always able to infer nothrow
        alias Key = ElementType!Keys;
        alias Value = ElementType!Values;
        static assert(isMutable!Value, "assocArray: value type must be mutable");
        Value[Key] aa;
        foreach (key; keys)
        {
            if (values.empty) break;

            // aa[key] is incorrectly not @safe if the destructor throws
            // https://issues.dlang.org/show_bug.cgi?id=18592
            static if (is(typeof(() @safe
            {
                import std.range : ElementType;
                import std.traits : hasElaborateDestructor;
                alias KeyElement = ElementType!Keys;
                static if (hasElaborateDestructor!KeyElement)
                    KeyElement.init.__xdtor();

                alias ValueElement = ElementType!Values;
                static if (hasElaborateDestructor!ValueElement)
                    ValueElement.init.__xdtor();
            })))
            {
                () @trusted {
                    aa[key] = values.front;
                }();
            }
            else
            {
                aa[key] = values.front;
            }
            values.popFront();
        }
        return aa;
    }
}

///
@safe pure /*nothrow*/ unittest
{
    import std.range : repeat, zip;
    import std.typecons : tuple;
    import std.range.primitives : autodecodeStrings;
    auto a = assocArray(zip([0, 1, 2], ["a", "b", "c"])); // aka zipMap
    static assert(is(typeof(a) == string[int]));
    assert(a == [0:"a", 1:"b", 2:"c"]);

    auto b = assocArray([ tuple("foo", "bar"), tuple("baz", "quux") ]);
    static assert(is(typeof(b) == string[string]));
    assert(b == ["foo":"bar", "baz":"quux"]);

    static if (autodecodeStrings)
        alias achar = dchar;
    else
        alias achar = immutable(char);
    auto c = assocArray("ABCD", true.repeat);
    static assert(is(typeof(c) == bool[achar]));
    bool[achar] expected = ['D':true, 'A':true, 'B':true, 'C':true];
    assert(c == expected);
}

// Cannot be version (StdUnittest) - recursive instantiation error
// https://issues.dlang.org/show_bug.cgi?id=11053
@safe pure nothrow unittest
{
    import std.typecons;
    static assert(!__traits(compiles, [ 1, 2, 3 ].assocArray()));
    static assert(!__traits(compiles, [ tuple("foo", "bar", "baz") ].assocArray()));
    static assert(!__traits(compiles, [ tuple("foo") ].assocArray()));
    assert([ tuple("foo", "bar") ].assocArray() == ["foo": "bar"]);
}

// https://issues.dlang.org/show_bug.cgi?id=13909
@safe pure nothrow unittest
{
    import std.typecons;
    auto a = [tuple!(const string, string)("foo", "bar")];
    auto b = [tuple!(string, const string)("foo", "bar")];
    assert(a == b);
    assert(assocArray(a) == [cast(const(string)) "foo": "bar"]);
    static assert(!__traits(compiles, assocArray(b)));
}

// https://issues.dlang.org/show_bug.cgi?id=5502
@safe pure nothrow unittest
{
    auto a = assocArray([0, 1, 2], ["a", "b", "c"]);
    static assert(is(typeof(a) == string[int]));
    assert(a == [0:"a", 1:"b", 2:"c"]);

    auto b = assocArray([0, 1, 2], [3L, 4, 5]);
    static assert(is(typeof(b) == long[int]));
    assert(b == [0: 3L, 1: 4, 2: 5]);
}

// https://issues.dlang.org/show_bug.cgi?id=5502
@safe pure unittest
{
    import std.algorithm.iteration : filter, map;
    import std.range : enumerate;
    import std.range.primitives : autodecodeStrings;

    auto r = "abcde".enumerate.filter!(a => a.index == 2);
    auto a = assocArray(r.map!(a => a.value), r.map!(a => a.index));
    static if (autodecodeStrings)
        alias achar = dchar;
    else
        alias achar = immutable(char);
    static assert(is(typeof(a) == size_t[achar]));
    assert(a == [achar('c'): size_t(2)]);
}

@safe nothrow pure unittest
{
    import std.range : iota;
    auto b = assocArray(3.iota, 3.iota(6));
    static assert(is(typeof(b) == int[int]));
    assert(b == [0: 3, 1: 4, 2: 5]);

    b = assocArray([0, 1, 2], [3, 4, 5]);
    assert(b == [0: 3, 1: 4, 2: 5]);
}

@safe unittest
{
    struct ThrowingElement
    {
        int i;
        static bool b;
        ~this(){
            if (b)
                throw new Exception("");
        }
    }
    static assert(!__traits(compiles, () nothrow { assocArray([ThrowingElement()], [0]);}));
    assert(assocArray([ThrowingElement()], [0]) == [ThrowingElement(): 0]);

    static assert(!__traits(compiles, () nothrow { assocArray([0], [ThrowingElement()]);}));
    assert(assocArray([0], [ThrowingElement()]) == [0: ThrowingElement()]);

    import std.range : iota;
    static assert(!__traits(compiles, () nothrow { assocArray(1.iota, [ThrowingElement()]);}));
    assert(assocArray(1.iota, [ThrowingElement()]) == [0: ThrowingElement()]);
}

@system unittest
{
    import std.range : iota;
    struct UnsafeElement
    {
        int i;
        static bool b;
        ~this(){
            int[] arr;
            void* p = arr.ptr + 1; // unsafe
        }
    }
    static assert(!__traits(compiles, () @safe { assocArray(1.iota, [UnsafeElement()]);}));
    assert(assocArray(1.iota, [UnsafeElement()]) == [0: UnsafeElement()]);
}

/**
Construct a range iterating over an associative array by key/value tuples.

Params:
    aa = The associative array to iterate over.

Returns: A $(REF_ALTTEXT forward range, isForwardRange, std,range,primitives)
of Tuple's of key and value pairs from the given associative array. The members
of each pair can be accessed by name (`.key` and `.value`). or by integer
index (0 and 1 respectively).
*/
auto byPair(AA)(AA aa)
if (isAssociativeArray!AA)
{
    import std.algorithm.iteration : map;
    import std.typecons : tuple;

    return aa.byKeyValue
        .map!(pair => tuple!("key", "value")(pair.key, pair.value));
}

///
@safe pure nothrow unittest
{
    import std.algorithm.sorting : sort;
    import std.typecons : tuple, Tuple;

    auto aa = ["a": 1, "b": 2, "c": 3];
    Tuple!(string, int)[] pairs;

    // Iteration over key/value pairs.
    foreach (pair; aa.byPair)
    {
        if (pair.key == "b")
            pairs ~= tuple("B", pair.value);
        else
            pairs ~= pair;
    }

    // Iteration order is implementation-dependent, so we should sort it to get
    // a fixed order.
    pairs.sort();
    assert(pairs == [
        tuple("B", 2),
        tuple("a", 1),
        tuple("c", 3)
    ]);
}

@safe pure nothrow unittest
{
    import std.typecons : tuple, Tuple;
    import std.meta : AliasSeq;

    auto aa = ["a":2];
    auto pairs = aa.byPair();

    alias PT = typeof(pairs.front);
    static assert(is(PT : Tuple!(string,int)));
    static assert(PT.fieldNames == AliasSeq!("key", "value"));
    static assert(isForwardRange!(typeof(pairs)));

    assert(!pairs.empty);
    assert(pairs.front == tuple("a", 2));

    auto savedPairs = pairs.save;

    pairs.popFront();
    assert(pairs.empty);
    assert(!savedPairs.empty);
    assert(savedPairs.front == tuple("a", 2));
}

// https://issues.dlang.org/show_bug.cgi?id=17711
@safe pure nothrow unittest
{
    const(int[string]) aa = [ "abc": 123 ];

    // Ensure that byKeyValue is usable with a const AA.
    auto kv = aa.byKeyValue;
    assert(!kv.empty);
    assert(kv.front.key == "abc" && kv.front.value == 123);
    kv.popFront();
    assert(kv.empty);

    // Ensure byPair is instantiable with const AA.
    auto r = aa.byPair;
    static assert(isInputRange!(typeof(r)));
    assert(!r.empty && r.front[0] == "abc" && r.front[1] == 123);
    r.popFront();
    assert(r.empty);
}

private template blockAttribute(T)
{
    import core.memory;
    static if (hasIndirections!(T) || is(T == void))
    {
        enum blockAttribute = 0;
    }
    else
    {
        enum blockAttribute = GC.BlkAttr.NO_SCAN;
    }
}

@safe unittest
{
    import core.memory : UGC = GC;
    static assert(!(blockAttribute!void & UGC.BlkAttr.NO_SCAN));
}

// Returns the number of dimensions in an array T.
private template nDimensions(T)
{
    static if (isArray!T)
    {
        enum nDimensions = 1 + nDimensions!(typeof(T.init[0]));
    }
    else
    {
        enum nDimensions = 0;
    }
}

@safe unittest
{
    static assert(nDimensions!(uint[]) == 1);
    static assert(nDimensions!(float[][]) == 2);
}

/++
Returns a new array of type `T` allocated on the garbage collected heap
without initializing its elements. This can be a useful optimization if every
element will be immediately initialized. `T` may be a multidimensional
array. In this case sizes may be specified for any number of dimensions from 0
to the number in `T`.

uninitializedArray is `nothrow` and weakly `pure`.

uninitializedArray is `@system` if the uninitialized element type has pointers.

Params:
    T = The type of the resulting array elements
    sizes = The length dimension(s) of the resulting array
Returns:
    An array of `T` with `I.length` dimensions.
+/
auto uninitializedArray(T, I...)(I sizes) nothrow @system
if (isDynamicArray!T && allSatisfy!(isIntegral, I) && hasIndirections!(ElementEncodingType!T))
{
    enum isSize_t(E) = is (E : size_t);
    alias toSize_t(E) = size_t;

    static assert(allSatisfy!(isSize_t, I),
        "Argument types in "~I.stringof~" are not all convertible to size_t: "
        ~Filter!(templateNot!(isSize_t), I).stringof);

    //Eagerlly transform non-size_t into size_t to avoid template bloat
    alias ST = staticMap!(toSize_t, I);

    return arrayAllocImpl!(false, T, ST)(sizes);
}

/// ditto
auto uninitializedArray(T, I...)(I sizes) nothrow @trusted
if (isDynamicArray!T && allSatisfy!(isIntegral, I) && !hasIndirections!(ElementEncodingType!T))
{
    enum isSize_t(E) = is (E : size_t);
    alias toSize_t(E) = size_t;

    static assert(allSatisfy!(isSize_t, I),
        "Argument types in "~I.stringof~" are not all convertible to size_t: "
        ~Filter!(templateNot!(isSize_t), I).stringof);

    //Eagerlly transform non-size_t into size_t to avoid template bloat
    alias ST = staticMap!(toSize_t, I);

    return arrayAllocImpl!(false, T, ST)(sizes);
}
///
@system nothrow pure unittest
{
    double[] arr = uninitializedArray!(double[])(100);
    assert(arr.length == 100);

    double[][] matrix = uninitializedArray!(double[][])(42, 31);
    assert(matrix.length == 42);
    assert(matrix[0].length == 31);

    char*[] ptrs = uninitializedArray!(char*[])(100);
    assert(ptrs.length == 100);
}

/++
Returns a new array of type `T` allocated on the garbage collected heap.

Partial initialization is done for types with indirections, for preservation
of memory safety. Note that elements will only be initialized to 0, but not
necessarily the element type's `.init`.

minimallyInitializedArray is `nothrow` and weakly `pure`.

Params:
    T = The type of the array elements
    sizes = The length dimension(s) of the resulting array
Returns:
    An array of `T` with `I.length` dimensions.
+/
auto minimallyInitializedArray(T, I...)(I sizes) nothrow @trusted
if (isDynamicArray!T && allSatisfy!(isIntegral, I))
{
    enum isSize_t(E) = is (E : size_t);
    alias toSize_t(E) = size_t;

    static assert(allSatisfy!(isSize_t, I),
        "Argument types in "~I.stringof~" are not all convertible to size_t: "
        ~Filter!(templateNot!(isSize_t), I).stringof);
    //Eagerlly transform non-size_t into size_t to avoid template bloat
    alias ST = staticMap!(toSize_t, I);

    return arrayAllocImpl!(true, T, ST)(sizes);
}

///
@safe pure nothrow unittest
{
    import std.algorithm.comparison : equal;
    import std.range : repeat;

    auto arr = minimallyInitializedArray!(int[])(42);
    assert(arr.length == 42);

    // Elements aren't necessarily initialized to 0, so don't do this:
    // assert(arr.equal(0.repeat(42)));
    // If that is needed, initialize the array normally instead:
    auto arr2 = new int[42];
    assert(arr2.equal(0.repeat(42)));
}

@safe pure nothrow unittest
{
    cast(void) minimallyInitializedArray!(int[][][][][])();
    double[] arr = minimallyInitializedArray!(double[])(100);
    assert(arr.length == 100);

    double[][] matrix = minimallyInitializedArray!(double[][])(42);
    assert(matrix.length == 42);
    foreach (elem; matrix)
    {
        assert(elem.ptr is null);
    }
}

// from rt/lifetime.d
private extern(C) void[] _d_newarrayU(const TypeInfo ti, size_t length) pure nothrow;

// from rt/tracegc.d
version (D_ProfileGC)
private extern (C) void[] _d_newarrayUTrace(string file, size_t line,
    string funcname, const scope TypeInfo ti, size_t length) pure nothrow;

private auto arrayAllocImpl(bool minimallyInitialized, T, I...)(I sizes) nothrow
{
    static assert(I.length <= nDimensions!T,
        I.length.stringof~"dimensions specified for a "~nDimensions!T.stringof~" dimensional array.");

    alias E = ElementEncodingType!T;

    E[] ret;

    static if (I.length != 0)
    {
        static assert(is(I[0] == size_t), "I[0] must be of type size_t not "
                ~ I[0].stringof);
        alias size = sizes[0];
    }

    static if (I.length == 1)
    {
        if (__ctfe)
        {
            static if (__traits(compiles, new E[](size)))
                ret = new E[](size);
            else static if (__traits(compiles, ret ~= E.init))
            {
                try
                {
                    //Issue: if E has an impure postblit, then all of arrayAllocImpl
                    //Will be impure, even during non CTFE.
                    foreach (i; 0 .. size)
                        ret ~= E.init;
                }
                catch (Exception e)
                    assert(0, e.msg);
            }
            else
                assert(0, "No postblit nor default init on " ~ E.stringof ~
                    ": At least one is required for CTFE.");
        }
        else
        {
            import core.stdc.string : memset;

            /+
              NOTES:
              _d_newarrayU is part of druntime, and creates an uninitialized
              block, just like GC.malloc. However, it also sets the appropriate
              bits, and sets up the block as an appendable array of type E[],
              which will inform the GC how to destroy the items in the block
              when it gets collected.

              _d_newarrayU returns a void[], but with the length set according
              to E.sizeof.
            +/
            version (D_ProfileGC)
            {
                // FIXME: file, line, function should be propagated from the
                // caller, not here.
                *(cast(void[]*)&ret) = _d_newarrayUTrace(__FILE__, __LINE__,
                    __FUNCTION__, typeid(E[]), size);
            }
            else
                *(cast(void[]*)&ret) = _d_newarrayU(typeid(E[]), size);
            static if (minimallyInitialized && hasIndirections!E)
                // _d_newarrayU would have asserted if the multiplication below
                // had overflowed, so we don't have to check it again.
                memset(ret.ptr, 0, E.sizeof * ret.length);
        }
    }
    else static if (I.length > 1)
    {
        ret = arrayAllocImpl!(false, E[])(size);
        foreach (ref elem; ret)
            elem = arrayAllocImpl!(minimallyInitialized, E)(sizes[1..$]);
    }

    return ret;
}

@safe nothrow pure unittest
{
    auto s1 = uninitializedArray!(int[])();
    auto s2 = minimallyInitializedArray!(int[])();
    assert(s1.length == 0);
    assert(s2.length == 0);
}

// https://issues.dlang.org/show_bug.cgi?id=9803
@safe nothrow pure unittest
{
    auto a = minimallyInitializedArray!(int*[])(1);
    assert(a[0] == null);
    auto b = minimallyInitializedArray!(int[][])(1);
    assert(b[0].empty);
    auto c = minimallyInitializedArray!(int*[][])(1, 1);
    assert(c[0][0] == null);
}

// https://issues.dlang.org/show_bug.cgi?id=10637
@safe pure nothrow unittest
{
    static struct S
    {
        static struct I{int i; alias i this;}
        int* p;
        this() @disable;
        this(int i)
        {
            p = &(new I(i)).i;
        }
        this(this)
        {
            p = &(new I(*p)).i;
        }
        ~this()
        {
            // note, this assert is invalid -- a struct should always be able
            // to run its dtor on the .init value, I'm leaving it here
            // commented out because the original test case had it. I'm not
            // sure what it's trying to prove.
            //
            // What happens now that minimallyInitializedArray adds the
            // destructor run to the GC, is that this assert would fire in the
            // GC, which triggers an invalid memory operation.
            //assert(p != null);
        }
    }
    auto a = minimallyInitializedArray!(S[])(1);
    assert(a[0].p == null);
    enum b = minimallyInitializedArray!(S[])(1);
    assert(b[0].p == null);
}

@safe pure nothrow unittest
{
    static struct S1
    {
        this() @disable;
        this(this) @disable;
    }
    auto a1 = minimallyInitializedArray!(S1[][])(2, 2);
    assert(a1);
    static struct S2
    {
        this() @disable;
        //this(this) @disable;
    }
    auto a2 = minimallyInitializedArray!(S2[][])(2, 2);
    assert(a2);
    enum b2 = minimallyInitializedArray!(S2[][])(2, 2);
    assert(b2 !is null);
    static struct S3
    {
        //this() @disable;
        this(this) @disable;
    }
    auto a3 = minimallyInitializedArray!(S3[][])(2, 2);
    assert(a3);
    enum b3 = minimallyInitializedArray!(S3[][])(2, 2);
    assert(b3 !is null);
}

/++
Returns the overlapping portion, if any, of two arrays. Unlike `equal`,
`overlap` only compares the pointers and lengths in the
ranges, not the values referred by them. If `r1` and `r2` have an
overlapping slice, returns that slice. Otherwise, returns the null
slice.

Params:
    a = The first array to compare
    b = The second array to compare
Returns:
    The overlapping portion of the two arrays.
+/
CommonType!(T[], U[]) overlap(T, U)(T[] a, U[] b) @trusted
if (is(typeof(a.ptr < b.ptr) == bool))
{
    import std.algorithm.comparison : min;

    auto end = min(a.ptr + a.length, b.ptr + b.length);
    // CTFE requires pairing pointer comparisons, which forces a
    // slightly inefficient implementation.
    if (a.ptr <= b.ptr && b.ptr < a.ptr + a.length)
    {
        return b.ptr[0 .. end - b.ptr];
    }

    if (b.ptr <= a.ptr && a.ptr < b.ptr + b.length)
    {
        return a.ptr[0 .. end - a.ptr];
    }

    return null;
}

///
@safe pure nothrow unittest
{
    int[] a = [ 10, 11, 12, 13, 14 ];
    int[] b = a[1 .. 3];
    assert(overlap(a, b) == [ 11, 12 ]);
    b = b.dup;
    // overlap disappears even though the content is the same
    assert(overlap(a, b).empty);

    static test()() @nogc
    {
        auto a = "It's three o'clock"d;
        auto b = a[5 .. 10];
        return b.overlap(a);
    }

    //works at compile-time
    static assert(test == "three"d);
}

@safe pure nothrow unittest
{
    static void test(L, R)(L l, R r)
    {
        assert(overlap(l, r) == [ 100, 12 ]);

        assert(overlap(l, l[0 .. 2]) is l[0 .. 2]);
        assert(overlap(l, l[3 .. 5]) is l[3 .. 5]);
        assert(overlap(l[0 .. 2], l) is l[0 .. 2]);
        assert(overlap(l[3 .. 5], l) is l[3 .. 5]);
    }

    int[] a = [ 10, 11, 12, 13, 14 ];
    int[] b = a[1 .. 3];
    a[1] = 100;

    immutable int[] c = a.idup;
    immutable int[] d = c[1 .. 3];

    test(a, b);
    assert(overlap(a, b.dup).empty);
    test(c, d);
    assert(overlap(c, d.dup.idup).empty);
}

 // https://issues.dlang.org/show_bug.cgi?id=9836
@safe pure nothrow unittest
{
    // range primitives for array should work with alias this types
    struct Wrapper
    {
        int[] data;
        alias data this;

        @property Wrapper save() { return this; }
    }
    auto w = Wrapper([1,2,3,4]);
    std.array.popFront(w); // should work

    static assert(isInputRange!Wrapper);
    static assert(isForwardRange!Wrapper);
    static assert(isBidirectionalRange!Wrapper);
    static assert(isRandomAccessRange!Wrapper);
}

private void copyBackwards(T)(T[] src, T[] dest)
{
    import core.stdc.string : memmove;
    import std.format : format;

    assert(src.length == dest.length, format!
            "src.length %s must equal dest.length %s"(src.length, dest.length));

    if (!__ctfe || hasElaborateCopyConstructor!T)
    {
        /* insertInPlace relies on dest being uninitialized, so no postblits allowed,
         * as this is a MOVE that overwrites the destination, not a COPY.
         * BUG: insertInPlace will not work with ctfe and postblits
         */
        memmove(dest.ptr, src.ptr, src.length * T.sizeof);
    }
    else
    {
        immutable len = src.length;
        for (size_t i = len; i-- > 0;)
        {
            dest[i] = src[i];
        }
    }
}

/++
    Inserts `stuff` (which must be an input range or any number of
    implicitly convertible items) in `array` at position `pos`.

    Params:
        array = The array that `stuff` will be inserted into.
        pos   = The position in `array` to insert the `stuff`.
        stuff = An $(REF_ALTTEXT input range, isInputRange, std,range,primitives),
        or any number of implicitly convertible items to insert into `array`.
 +/
void insertInPlace(T, U...)(ref T[] array, size_t pos, U stuff)
if (!isSomeString!(T[])
    && allSatisfy!(isInputRangeOrConvertible!T, U) && U.length > 0)
{
    static if (allSatisfy!(isInputRangeWithLengthOrConvertible!T, U))
    {
        import core.internal.lifetime : emplaceRef;

        immutable oldLen = array.length;

        size_t to_insert = 0;
        foreach (i, E; U)
        {
            static if (is(E : T)) //a single convertible value, not a range
                to_insert += 1;
            else
                to_insert += stuff[i].length;
        }
        if (to_insert)
        {
            array.length += to_insert;

            // Takes arguments array, pos, stuff
            // Spread apart array[] at pos by moving elements
            (() @trusted { copyBackwards(array[pos .. oldLen], array[pos+to_insert..$]); })();

            // Initialize array[pos .. pos+to_insert] with stuff[]
            auto j = 0;
            foreach (i, E; U)
            {
                static if (is(E : T))
                {
                    emplaceRef!T(array[pos + j++], stuff[i]);
                }
                else
                {
                    foreach (v; stuff[i])
                    {
                        emplaceRef!T(array[pos + j++], v);
                    }
                }
            }
        }
    }
    else
    {
        // stuff has some InputRanges in it that don't have length
        // assume that stuff to be inserted is typically shorter
        // then the array that can be arbitrary big
        // TODO: needs a better implementation as there is no need to build an _array_
        // a singly-linked list of memory blocks (rope, etc.) will do
        auto app = appender!(T[])();
        foreach (i, E; U)
            app.put(stuff[i]);
        insertInPlace(array, pos, app.data);
    }
}

/// Ditto
void insertInPlace(T, U...)(ref T[] array, size_t pos, U stuff)
if (isSomeString!(T[]) && allSatisfy!(isCharOrStringOrDcharRange, U))
{
    static if (is(Unqual!T == T)
        && allSatisfy!(isInputRangeWithLengthOrConvertible!dchar, U))
    {
        import std.utf : codeLength, byDchar;
        // mutable, can do in place
        //helper function: re-encode dchar to Ts and store at *ptr
        static T* putDChar(T* ptr, dchar ch)
        {
            static if (is(T == dchar))
            {
                *ptr++ = ch;
                return ptr;
            }
            else
            {
                import std.utf : encode;
                T[dchar.sizeof/T.sizeof] buf;
                immutable len = encode(buf, ch);
                final switch (len)
                {
                    static if (T.sizeof == char.sizeof)
                    {
                case 4:
                        ptr[3] = buf[3];
                        goto case;
                case 3:
                        ptr[2] = buf[2];
                        goto case;
                    }
                case 2:
                    ptr[1] = buf[1];
                    goto case;
                case 1:
                    ptr[0] = buf[0];
                }
                ptr += len;
                return ptr;
            }
        }
        size_t to_insert = 0;
        //count up the number of *codeunits* to insert
        foreach (i, E; U)
            to_insert += codeLength!T(stuff[i]);
        array.length += to_insert;

        @trusted static void moveToRight(T[] arr, size_t gap)
        {
            static assert(!hasElaborateCopyConstructor!T,
                    "T must not have an elaborate copy constructor");
            import core.stdc.string : memmove;
            if (__ctfe)
            {
                for (size_t i = arr.length - gap; i; --i)
                    arr[gap + i - 1] = arr[i - 1];
            }
            else
                memmove(arr.ptr + gap, arr.ptr, (arr.length - gap) * T.sizeof);
        }
        moveToRight(array[pos .. $], to_insert);
        auto ptr = array.ptr + pos;
        foreach (i, E; U)
        {
            static if (is(E : dchar))
            {
                ptr = putDChar(ptr, stuff[i]);
            }
            else
            {
                foreach (ch; stuff[i].byDchar)
                    ptr = putDChar(ptr, ch);
            }
        }
        assert(ptr == array.ptr + pos + to_insert, "(ptr == array.ptr + pos + to_insert) is false");
    }
    else
    {
        // immutable/const, just construct a new array
        auto app = appender!(T[])();
        app.put(array[0 .. pos]);
        foreach (i, E; U)
            app.put(stuff[i]);
        app.put(array[pos..$]);
        array = app.data;
    }
}

///
@safe pure unittest
{
    int[] a = [ 1, 2, 3, 4 ];
    a.insertInPlace(2, [ 1, 2 ]);
    assert(a == [ 1, 2, 1, 2, 3, 4 ]);
    a.insertInPlace(3, 10u, 11);
    assert(a == [ 1, 2, 1, 10, 11, 2, 3, 4]);

    union U
    {
        float a = 3.0;
        int b;
    }

    U u1 = { b : 3 };
    U u2 = { b : 4 };
    U u3 = { b : 5 };
    U[] unionArr = [u2, u3];
    unionArr.insertInPlace(2, [u1]);
    assert(unionArr == [u2, u3, u1]);
    unionArr.insertInPlace(0, [u3, u2]);
    assert(unionArr == [u3, u2, u2, u3, u1]);

    static class C
    {
        int a;
        float b;

        this(int a, float b) { this.a = a; this.b = b; }
    }

    C c1 = new C(42, 1.0);
    C c2 = new C(0, 0.0);
    C c3 = new C(int.max, float.init);

    C[] classArr = [c1, c2, c3];
    insertInPlace(classArr, 3, [c2, c3]);
    C[5] classArr1 = classArr;
    assert(classArr1 == [c1, c2, c3, c2, c3]);
    insertInPlace(classArr, 0, c3, c1);
    C[7] classArr2 = classArr;
    assert(classArr2 == [c3, c1, c1, c2, c3, c2, c3]);
}

//constraint helpers
private template isInputRangeWithLengthOrConvertible(E)
{
    template isInputRangeWithLengthOrConvertible(R)
    {
        //hasLength not defined for char[], wchar[] and dchar[]
        enum isInputRangeWithLengthOrConvertible =
            (isInputRange!R && is(typeof(R.init.length))
                && is(ElementType!R : E))  || is(R : E);
    }
}

//ditto
private template isCharOrStringOrDcharRange(T)
{
    enum isCharOrStringOrDcharRange = isSomeString!T || isSomeChar!T ||
        (isInputRange!T && is(ElementType!T : dchar));
}

//ditto
private template isInputRangeOrConvertible(E)
{
    template isInputRangeOrConvertible(R)
    {
        enum isInputRangeOrConvertible =
            (isInputRange!R && is(ElementType!R : E))  || is(R : E);
    }
}

@system unittest
{
    // @system due to insertInPlace
    import core.exception;
    import std.algorithm.comparison : equal;
    import std.algorithm.iteration : filter;
    import std.conv : to;
    import std.exception;


    bool test(T, U, V)(T orig, size_t pos, U toInsert, V result)
    {
        {
            static if (is(T == typeof(T.init.dup)))
                auto a = orig.dup;
            else
                auto a = orig.idup;

            a.insertInPlace(pos, toInsert);
            if (!equal(a, result))
                return false;
        }

        static if (isInputRange!U)
        {
            orig.insertInPlace(pos, filter!"true"(toInsert));
            return equal(orig, result);
        }
        else
            return true;
    }


    assert(test([1, 2, 3, 4], 0, [6, 7], [6, 7, 1, 2, 3, 4]));
    assert(test([1, 2, 3, 4], 2, [8, 9], [1, 2, 8, 9, 3, 4]));
    assert(test([1, 2, 3, 4], 4, [10, 11], [1, 2, 3, 4, 10, 11]));

    assert(test([1, 2, 3, 4], 0, 22, [22, 1, 2, 3, 4]));
    assert(test([1, 2, 3, 4], 2, 23, [1, 2, 23, 3, 4]));
    assert(test([1, 2, 3, 4], 4, 24, [1, 2, 3, 4, 24]));

    void testStr(T, U)(string file = __FILE__, size_t line = __LINE__)
    {

        auto l = to!T("hello");
        auto r = to!U(" વિશ્વ");

        enforce(test(l, 0, r, " વિશ્વhello"),
                new AssertError("testStr failure 1", file, line));
        enforce(test(l, 3, r, "hel વિશ્વlo"),
                new AssertError("testStr failure 2", file, line));
        enforce(test(l, l.length, r, "hello વિશ્વ"),
                new AssertError("testStr failure 3", file, line));
    }

    static foreach (T; AliasSeq!(char, wchar, dchar,
        immutable(char), immutable(wchar), immutable(dchar)))
    {
        static foreach (U; AliasSeq!(char, wchar, dchar,
            immutable(char), immutable(wchar), immutable(dchar)))
        {
            testStr!(T[], U[])();
        }

    }

    // variadic version
    bool testVar(T, U...)(T orig, size_t pos, U args)
    {
        static if (is(T == typeof(T.init.dup)))
            auto a = orig.dup;
        else
            auto a = orig.idup;
        auto result = args[$-1];

        a.insertInPlace(pos, args[0..$-1]);
        if (!equal(a, result))
            return false;
        return true;
    }
    assert(testVar([1, 2, 3, 4], 0, 6, 7u, [6, 7, 1, 2, 3, 4]));
    assert(testVar([1L, 2, 3, 4], 2, 8, 9L, [1, 2, 8, 9, 3, 4]));
    assert(testVar([1L, 2, 3, 4], 4, 10L, 11, [1, 2, 3, 4, 10, 11]));
    assert(testVar([1L, 2, 3, 4], 4, [10, 11], 40L, 42L,
                    [1, 2, 3, 4, 10, 11, 40, 42]));
    assert(testVar([1L, 2, 3, 4], 4, 10, 11, [40L, 42],
                    [1, 2, 3, 4, 10, 11, 40, 42]));
    assert(testVar("t".idup, 1, 'e', 's', 't', "test"));
    assert(testVar("!!"w.idup, 1, "\u00e9ll\u00f4", 'x', "TTT"w, 'y',
                    "!\u00e9ll\u00f4xTTTy!"));
    assert(testVar("flipflop"d.idup, 4, '_',
                    "xyz"w, '\U00010143', '_', "abc"d, "__",
                    "flip_xyz\U00010143_abc__flop"));
}

@system unittest
{
    import std.algorithm.comparison : equal;
    // insertInPlace interop with postblit
    static struct Int
    {
        int* payload;
        this(int k)
        {
            payload = new int;
            *payload = k;
        }
        this(this)
        {
            int* np = new int;
            *np = *payload;
            payload = np;
        }
        ~this()
        {
            if (payload)
                *payload = 0; //'destroy' it
        }
        @property int getPayload(){ return *payload; }
        alias getPayload this;
    }

    Int[] arr = [Int(1), Int(4), Int(5)];
    assert(arr[0] == 1);
    insertInPlace(arr, 1, Int(2), Int(3));
    assert(equal(arr, [1, 2, 3, 4, 5]));  //check it works with postblit
}

@safe unittest
{
    import std.exception;
    assertCTFEable!(
    {
        int[] a = [1, 2];
        a.insertInPlace(2, 3);
        a.insertInPlace(0, -1, 0);
        return a == [-1, 0, 1, 2, 3];
    });
}

// https://issues.dlang.org/show_bug.cgi?id=6874
@system unittest
{
    import core.memory;
    // allocate some space
    byte[] a;
    a.length = 1;

    // fill it
    a.length = a.capacity;

    // write beyond
    byte[] b = a[$ .. $];
    b.insertInPlace(0, a);

    // make sure that reallocation has happened
    assert(GC.addrOf(&b[0]) == GC.addrOf(&b[$-1]));
}


/++
    Returns whether the `front`s of `lhs` and `rhs` both refer to the
    same place in memory, making one of the arrays a slice of the other which
    starts at index `0`.

    Params:
        lhs = the first array to compare
        rhs = the second array to compare
    Returns:
        `true` if $(D lhs.ptr == rhs.ptr), `false` otherwise.
  +/
@safe
pure nothrow @nogc bool sameHead(T)(in T[] lhs, in T[] rhs)
{
    return lhs.ptr == rhs.ptr;
}

///
@safe pure nothrow unittest
{
    auto a = [1, 2, 3, 4, 5];
    auto b = a[0 .. 2];

    assert(a.sameHead(b));
}


/++
    Returns whether the `back`s of `lhs` and `rhs` both refer to the
    same place in memory, making one of the arrays a slice of the other which
    end at index `$`.

    Params:
        lhs = the first array to compare
        rhs = the second array to compare
    Returns:
        `true` if both arrays are the same length and $(D lhs.ptr == rhs.ptr),
        `false` otherwise.
  +/
@trusted
pure nothrow @nogc bool sameTail(T)(in T[] lhs, in T[] rhs)
{
    return lhs.ptr + lhs.length == rhs.ptr + rhs.length;
}

///
@safe pure nothrow unittest
{
    auto a = [1, 2, 3, 4, 5];
    auto b = a[3..$];

    assert(a.sameTail(b));
}

@safe pure nothrow unittest
{
    static foreach (T; AliasSeq!(int[], const(int)[], immutable(int)[], const int[], immutable int[]))
    {{
        T a = [1, 2, 3, 4, 5];
        T b = a;
        T c = a[1 .. $];
        T d = a[0 .. 1];
        T e = null;

        assert(sameHead(a, a));
        assert(sameHead(a, b));
        assert(!sameHead(a, c));
        assert(sameHead(a, d));
        assert(!sameHead(a, e));

        assert(sameTail(a, a));
        assert(sameTail(a, b));
        assert(sameTail(a, c));
        assert(!sameTail(a, d));
        assert(!sameTail(a, e));

        //verifies R-value compatibilty
        assert(a.sameHead(a[0 .. 0]));
        assert(a.sameTail(a[$ .. $]));
    }}
}

/**
Params:
    s = an $(REF_ALTTEXT input range, isInputRange, std,range,primitives)
    or a dynamic array
    n = number of times to repeat `s`

Returns:
    An array that consists of `s` repeated `n` times. This function allocates, fills, and
    returns a new array.

See_Also:
    For a lazy version, refer to $(REF repeat, std,range).
 */
ElementEncodingType!S[] replicate(S)(S s, size_t n)
if (isDynamicArray!S)
{
    alias RetType = ElementEncodingType!S[];

    // Optimization for return join(std.range.repeat(s, n));
    if (n == 0)
        return RetType.init;
    if (n == 1)
        return cast(RetType) s;
    auto r = new Unqual!(typeof(s[0]))[n * s.length];
    if (s.length == 1)
        r[] = s[0];
    else
    {
        immutable len = s.length, nlen = n * len;
        for (size_t i = 0; i < nlen; i += len)
        {
            r[i .. i + len] = s[];
        }
    }
    return r;
}

/// ditto
ElementType!S[] replicate(S)(S s, size_t n)
if (isInputRange!S && !isDynamicArray!S)
{
    import std.range : repeat;
    return join(std.range.repeat(s, n));
}


///
@safe unittest
{
    auto a = "abc";
    auto s = replicate(a, 3);

    assert(s == "abcabcabc");

    auto b = [1, 2, 3];
    auto c = replicate(b, 3);

    assert(c == [1, 2, 3, 1, 2, 3, 1, 2, 3]);

    auto d = replicate(b, 0);

    assert(d == []);
}

@safe unittest
{
    import std.conv : to;

    static foreach (S; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[]))
    {{
        immutable S t = "abc";

        assert(replicate(to!S("1234"), 0) is null);
        assert(replicate(to!S("1234"), 0) is null);
        assert(replicate(to!S("1234"), 1) == "1234");
        assert(replicate(to!S("1234"), 2) == "12341234");
        assert(replicate(to!S("1"), 4) == "1111");
        assert(replicate(t, 3) == "abcabcabc");
        assert(replicate(cast(S) null, 4) is null);
    }}
}

/++
Eagerly splits `range` into an array, using `sep` as the delimiter.

When no delimiter is provided, strings are split into an array of words,
using whitespace as delimiter.
Runs of whitespace are merged together (no empty words are produced).

The `range` must be a $(REF_ALTTEXT forward range, isForwardRange, std,range,primitives).
The separator can be a value of the same type as the elements in `range`
or it can be another forward `range`.

Params:
    s = the string to split by word if no separator is given
    range = the range to split
    sep = a value of the same type as the elements of `range` or another
    isTerminator = a predicate that splits the range when it returns `true`.

Returns:
    An array containing the divided parts of `range` (or the words of `s`).

See_Also:
$(REF splitter, std,algorithm,iteration) for a lazy version without allocating memory.

$(REF splitter, std,regex) for a version that splits using a regular
expression defined separator.
+/
S[] split(S)(S s) @safe pure
if (isSomeString!S)
{
    size_t istart;
    bool inword = false;
    auto result = appender!(S[]);

    foreach (i, dchar c ; s)
    {
        import std.uni : isWhite;
        if (isWhite(c))
        {
            if (inword)
            {
                put(result, s[istart .. i]);
                inword = false;
            }
        }
        else
        {
            if (!inword)
            {
                istart = i;
                inword = true;
            }
        }
    }
    if (inword)
        put(result, s[istart .. $]);
    return result.data;
}

///
@safe unittest
{
    import std.uni : isWhite;
    assert("Learning,D,is,fun".split(",") == ["Learning", "D", "is", "fun"]);
    assert("Learning D is fun".split!isWhite == ["Learning", "D", "is", "fun"]);
    assert("Learning D is fun".split(" D ") == ["Learning", "is fun"]);
}

///
@safe unittest
{
    string str = "Hello World!";
    assert(str.split == ["Hello", "World!"]);

    string str2 = "Hello\t\tWorld\t!";
    assert(str2.split == ["Hello", "World", "!"]);
}

@safe unittest
{
    import std.conv : to;
    import std.format : format;
    import std.typecons;

    static auto makeEntry(S)(string l, string[] r)
    {return tuple(l.to!S(), r.to!(S[])());}

    static foreach (S; AliasSeq!(string, wstring, dstring,))
    {{
        auto entries =
        [
            makeEntry!S("", []),
            makeEntry!S(" ", []),
            makeEntry!S("hello", ["hello"]),
            makeEntry!S(" hello ", ["hello"]),
            makeEntry!S("  h  e  l  l  o ", ["h", "e", "l", "l", "o"]),
            makeEntry!S("peter\t\npaul\rjerry", ["peter", "paul", "jerry"]),
            makeEntry!S(" \t\npeter paul\tjerry \n", ["peter", "paul", "jerry"]),
            makeEntry!S("\u2000日\u202F本\u205F語\u3000", ["日", "本", "語"]),
            makeEntry!S("　　哈・郎博尔德｝　　　　___一个", ["哈・郎博尔德｝", "___一个"])
        ];
        foreach (entry; entries)
            assert(entry[0].split() == entry[1], format("got: %s, expected: %s.", entry[0].split(), entry[1]));
    }}

    //Just to test that an immutable is split-able
    immutable string s = " \t\npeter paul\tjerry \n";
    assert(split(s) == ["peter", "paul", "jerry"]);
}

@safe unittest //purity, ctfe ...
{
    import std.exception;
    void dg() @safe pure {
        assert(split("hello world"c) == ["hello"c, "world"c]);
        assert(split("hello world"w) == ["hello"w, "world"w]);
        assert(split("hello world"d) == ["hello"d, "world"d]);
    }
    dg();
    assertCTFEable!dg;
}

///
@safe unittest
{
    assert(split("hello world") == ["hello","world"]);
    assert(split("192.168.0.1", ".") == ["192", "168", "0", "1"]);

    auto a = split([1, 2, 3, 4, 5, 1, 2, 3, 4, 5], [2, 3]);
    assert(a == [[1], [4, 5, 1], [4, 5]]);
}

///ditto
auto split(Range, Separator)(Range range, Separator sep)
if (isForwardRange!Range && (
    is(typeof(ElementType!Range.init == Separator.init)) ||
    is(typeof(ElementType!Range.init == ElementType!Separator.init)) && isForwardRange!Separator
    ))
{
    import std.algorithm.iteration : splitter;
    return range.splitter(sep).array;
}
///ditto
auto split(alias isTerminator, Range)(Range range)
if (isForwardRange!Range && is(typeof(unaryFun!isTerminator(range.front))))
{
    import std.algorithm.iteration : splitter;
    return range.splitter!isTerminator.array;
}

@safe unittest
{
    import std.algorithm.comparison : cmp;
    import std.conv;

    static foreach (S; AliasSeq!(string, wstring, dstring,
                    immutable(string), immutable(wstring), immutable(dstring),
                    char[], wchar[], dchar[],
                    const(char)[], const(wchar)[], const(dchar)[],
                    const(char[]), immutable(char[])))
    {{
        S s = to!S(",peter,paul,jerry,");

        auto words = split(s, ",");
        assert(words.length == 5, text(words.length));
        assert(cmp(words[0], "") == 0);
        assert(cmp(words[1], "peter") == 0);
        assert(cmp(words[2], "paul") == 0);
        assert(cmp(words[3], "jerry") == 0);
        assert(cmp(words[4], "") == 0);

        auto s1 = s[0 .. s.length - 1];   // lop off trailing ','
        words = split(s1, ",");
        assert(words.length == 4);
        assert(cmp(words[3], "jerry") == 0);

        auto s2 = s1[1 .. s1.length];   // lop off leading ','
        words = split(s2, ",");
        assert(words.length == 3);
        assert(cmp(words[0], "peter") == 0);

        auto s3 = to!S(",,peter,,paul,,jerry,,");

        words = split(s3, ",,");
        assert(words.length == 5);
        assert(cmp(words[0], "") == 0);
        assert(cmp(words[1], "peter") == 0);
        assert(cmp(words[2], "paul") == 0);
        assert(cmp(words[3], "jerry") == 0);
        assert(cmp(words[4], "") == 0);

        auto s4 = s3[0 .. s3.length - 2];    // lop off trailing ',,'
        words = split(s4, ",,");
        assert(words.length == 4);
        assert(cmp(words[3], "jerry") == 0);

        auto s5 = s4[2 .. s4.length];    // lop off leading ',,'
        words = split(s5, ",,");
        assert(words.length == 3);
        assert(cmp(words[0], "peter") == 0);
    }}
}

/+
   Conservative heuristic to determine if a range can be iterated cheaply.
   Used by `join` in decision to do an extra iteration of the range to
   compute the resultant length. If iteration is not cheap then precomputing
   length could be more expensive than using `Appender`.

   For now, we only assume arrays are cheap to iterate.
 +/
private enum bool hasCheapIteration(R) = isArray!R;

/++
   Eagerly concatenates all of the ranges in `ror` together (with the GC)
   into one array using `sep` as the separator if present.

   Params:
        ror = An $(REF_ALTTEXT input range, isInputRange, std,range,primitives)
        of input ranges
        sep = An input range, or a single element, to join the ranges on

   Returns:
        An array of elements

   See_Also:
        For a lazy version, see $(REF joiner, std,algorithm,iteration)
  +/
ElementEncodingType!(ElementType!RoR)[] join(RoR, R)(RoR ror, R sep)
if (isInputRange!RoR &&
    isInputRange!(Unqual!(ElementType!RoR)) &&
    isInputRange!R &&
    (is(immutable ElementType!(ElementType!RoR) == immutable ElementType!R) ||
     (isSomeChar!(ElementType!(ElementType!RoR)) && isSomeChar!(ElementType!R))
    ))
{
    alias RetType = typeof(return);
    alias RetTypeElement = Unqual!(ElementEncodingType!RetType);
    alias RoRElem = ElementType!RoR;

    if (ror.empty)
        return RetType.init;

    // Constraint only requires input range for sep.
    // This converts sep to an array (forward range) if it isn't one,
    // and makes sure it has the same string encoding for string types.
    static if (isSomeString!RetType &&
               !is(immutable ElementEncodingType!RetType == immutable ElementEncodingType!R))
    {
        import std.conv : to;
        auto sepArr = to!RetType(sep);
    }
    else static if (!isArray!R)
        auto sepArr = array(sep);
    else
        alias sepArr = sep;

    static if (hasCheapIteration!RoR && (hasLength!RoRElem || isNarrowString!RoRElem))
    {
        import core.internal.lifetime : emplaceRef;
        size_t length;          // length of result array
        size_t rorLength;       // length of range ror
        foreach (r; ror.save)
        {
            length += r.length;
            ++rorLength;
        }
        if (!rorLength)
            return null;
        length += (rorLength - 1) * sepArr.length;

        auto result = (() @trusted => uninitializedArray!(RetTypeElement[])(length))();
        size_t len;
        foreach (e; ror.front)
            emplaceRef(result[len++], e);
        ror.popFront();
        foreach (r; ror)
        {
            foreach (e; sepArr)
                emplaceRef(result[len++], e);
            foreach (e; r)
                emplaceRef(result[len++], e);
        }
        assert(len == result.length);
        return (() @trusted => cast(RetType) result)();
    }
    else
    {
        auto result = appender!RetType();
        put(result, ror.front);
        ror.popFront();
        for (; !ror.empty; ror.popFront())
        {
            put(result, sepArr);
            put(result, ror.front);
        }
        return result.data;
    }
}

// https://issues.dlang.org/show_bug.cgi?id=14230
@safe unittest
{
   string[] ary = ["","aa","bb","cc"]; // leaded by _empty_ element
   assert(ary.join(" @") == " @aa @bb @cc"); // OK in 2.067b1 and olders
}

// https://issues.dlang.org/show_bug.cgi?id=21337
@system unittest
{
    import std.algorithm.iteration : map;

    static class Once
    {
        bool empty;

        void popFront()
        {
            empty = true;
        }

        int front()
        {
            return 0;
        }
    }

    assert([1, 2].map!"[a]".join(new Once) == [1, 0, 2]);
}

/// Ditto
ElementEncodingType!(ElementType!RoR)[] join(RoR, E)(RoR ror, scope E sep)
if (isInputRange!RoR &&
    isInputRange!(Unqual!(ElementType!RoR)) &&
    ((is(E : ElementType!(ElementType!RoR))) ||
     (!autodecodeStrings && isSomeChar!(ElementType!(ElementType!RoR)) &&
      isSomeChar!E)))
{
    alias RetType = typeof(return);
    alias RetTypeElement = Unqual!(ElementEncodingType!RetType);
    alias RoRElem = ElementType!RoR;

    if (ror.empty)
        return RetType.init;

    static if (hasCheapIteration!RoR && (hasLength!RoRElem || isNarrowString!RoRElem))
    {
        static if (isSomeChar!E && isSomeChar!RetTypeElement && E.sizeof > RetTypeElement.sizeof)
        {
            import std.utf : encode;
            RetTypeElement[4 / RetTypeElement.sizeof] encodeSpace;
            immutable size_t sepArrLength = encode(encodeSpace, sep);
            return join(ror, encodeSpace[0 .. sepArrLength]);
        }
        else
        {
            import core.internal.lifetime : emplaceRef;
            import std.format : format;
            size_t length;
            size_t rorLength;
            foreach (r; ror.save)
            {
                length += r.length;
                ++rorLength;
            }
            if (!rorLength)
                return null;
            length += rorLength - 1;
            auto result = uninitializedArray!(RetTypeElement[])(length);


            size_t len;
            foreach (e; ror.front)
                emplaceRef(result[len++], e);
            ror.popFront();
            foreach (r; ror)
            {
                emplaceRef(result[len++], sep);
                foreach (e; r)
                    emplaceRef(result[len++], e);
            }
            assert(len == result.length, format!
                    "len %s must equal result.lenght %s"(len, result.length));
            return (() @trusted => cast(RetType) result)();
        }
    }
    else
    {
        auto result = appender!RetType();
        put(result, ror.front);
        ror.popFront();
        for (; !ror.empty; ror.popFront())
        {
            put(result, sep);
            put(result, ror.front);
        }
        return result.data;
    }
}

// https://issues.dlang.org/show_bug.cgi?id=14230
@safe unittest
{
   string[] ary = ["","aa","bb","cc"];
   assert(ary.join('@') == "@aa@bb@cc");
}

/// Ditto
ElementEncodingType!(ElementType!RoR)[] join(RoR)(RoR ror)
if (isInputRange!RoR &&
    isInputRange!(Unqual!(ElementType!RoR)))
{
    alias RetType = typeof(return);
    alias ConstRetTypeElement = ElementEncodingType!RetType;
    static if (isAssignable!(Unqual!ConstRetTypeElement, ConstRetTypeElement))
    {
        alias RetTypeElement = Unqual!ConstRetTypeElement;
    }
    else
    {
        alias RetTypeElement = ConstRetTypeElement;
    }
    alias RoRElem = ElementType!RoR;

    if (ror.empty)
        return RetType.init;

    static if (hasCheapIteration!RoR && (hasLength!RoRElem || isNarrowString!RoRElem))
    {
        import core.internal.lifetime : emplaceRef;
        size_t length;
        foreach (r; ror.save)
            length += r.length;

        auto result = (() @trusted => uninitializedArray!(RetTypeElement[])(length))();
        size_t len;
        foreach (r; ror)
            foreach (e; r)
                emplaceRef!RetTypeElement(result[len++], e);
        assert(len == result.length,
                "emplaced an unexpected number of elements");
        return (() @trusted => cast(RetType) result)();
    }
    else
    {
        auto result = appender!RetType();
        for (; !ror.empty; ror.popFront())
            put(result, ror.front);
        return result.data;
    }
}

///
@safe pure nothrow unittest
{
    assert(join(["hello", "silly", "world"], " ") == "hello silly world");
    assert(join(["hello", "silly", "world"]) == "hellosillyworld");

    assert(join([[1, 2, 3], [4, 5]], [72, 73]) == [1, 2, 3, 72, 73, 4, 5]);
    assert(join([[1, 2, 3], [4, 5]]) == [1, 2, 3, 4, 5]);

    const string[] arr = ["apple", "banana"];
    assert(arr.join(",") == "apple,banana");
    assert(arr.join() == "applebanana");
}

@safe pure unittest
{
    import std.conv : to;
    import std.range.primitives : autodecodeStrings;

    static foreach (T; AliasSeq!(string,wstring,dstring))
    {{
        auto arr2 = "Здравствуй Мир Unicode".to!(T);
        auto arr = ["Здравствуй", "Мир", "Unicode"].to!(T[]);
        assert(join(arr) == "ЗдравствуйМирUnicode");
        static foreach (S; AliasSeq!(char,wchar,dchar))
        {{
            auto jarr = arr.join(to!S(' '));
            static assert(is(typeof(jarr) == T));
            assert(jarr == arr2);
        }}
        static foreach (S; AliasSeq!(string,wstring,dstring))
        {{
            auto jarr = arr.join(to!S(" "));
            static assert(is(typeof(jarr) == T));
            assert(jarr == arr2);
        }}
    }}

    static foreach (T; AliasSeq!(string,wstring,dstring))
    {{
        auto arr2 = "Здравствуй\u047CМир\u047CUnicode".to!(T);
        auto arr = ["Здравствуй", "Мир", "Unicode"].to!(T[]);
        static foreach (S; AliasSeq!(wchar,dchar))
        {{
            auto jarr = arr.join(to!S('\u047C'));
            static assert(is(typeof(jarr) == T));
            assert(jarr == arr2);
        }}
    }}

    const string[] arr = ["apple", "banana"];
    assert(arr.join(',') == "apple,banana");
}

@safe unittest
{
    class A { }

    const A[][] array;
    auto result = array.join; // can't remove constness, so don't try

    static assert(is(typeof(result) == const(A)[]));
}

@safe unittest
{
    import std.algorithm;
    import std.conv : to;
    import std.range;

    static foreach (R; AliasSeq!(string, wstring, dstring))
    {{
        R word1 = "日本語";
        R word2 = "paul";
        R word3 = "jerry";
        R[] words = [word1, word2, word3];

        auto filteredWord1    = filter!"true"(word1);
        auto filteredLenWord1 = takeExactly(filteredWord1, word1.walkLength());
        auto filteredWord2    = filter!"true"(word2);
        auto filteredLenWord2 = takeExactly(filteredWord2, word2.walkLength());
        auto filteredWord3    = filter!"true"(word3);
        auto filteredLenWord3 = takeExactly(filteredWord3, word3.walkLength());
        auto filteredWordsArr = [filteredWord1, filteredWord2, filteredWord3];
        auto filteredLenWordsArr = [filteredLenWord1, filteredLenWord2, filteredLenWord3];
        auto filteredWords    = filter!"true"(filteredWordsArr);

        static foreach (S; AliasSeq!(string, wstring, dstring))
        {{
            assert(join(filteredWords, to!S(", ")) == "日本語, paul, jerry");
            assert(join(filteredWords, to!(ElementType!S)(',')) == "日本語,paul,jerry");
            assert(join(filteredWordsArr, to!(ElementType!(S))(',')) == "日本語,paul,jerry");
            assert(join(filteredWordsArr, to!S(", ")) == "日本語, paul, jerry");
            assert(join(filteredWordsArr, to!(ElementType!(S))(',')) == "日本語,paul,jerry");
            assert(join(filteredLenWordsArr, to!S(", ")) == "日本語, paul, jerry");
            assert(join(filter!"true"(words), to!S(", ")) == "日本語, paul, jerry");
            assert(join(words, to!S(", ")) == "日本語, paul, jerry");

            assert(join(filteredWords, to!S("")) == "日本語pauljerry");
            assert(join(filteredWordsArr, to!S("")) == "日本語pauljerry");
            assert(join(filteredLenWordsArr, to!S("")) == "日本語pauljerry");
            assert(join(filter!"true"(words), to!S("")) == "日本語pauljerry");
            assert(join(words, to!S("")) == "日本語pauljerry");

            assert(join(filter!"true"([word1]), to!S(", ")) == "日本語");
            assert(join([filteredWord1], to!S(", ")) == "日本語");
            assert(join([filteredLenWord1], to!S(", ")) == "日本語");
            assert(join(filter!"true"([filteredWord1]), to!S(", ")) == "日本語");
            assert(join([word1], to!S(", ")) == "日本語");

            assert(join(filteredWords, to!S(word1)) == "日本語日本語paul日本語jerry");
            assert(join(filteredWordsArr, to!S(word1)) == "日本語日本語paul日本語jerry");
            assert(join(filteredLenWordsArr, to!S(word1)) == "日本語日本語paul日本語jerry");
            assert(join(filter!"true"(words), to!S(word1)) == "日本語日本語paul日本語jerry");
            assert(join(words, to!S(word1)) == "日本語日本語paul日本語jerry");

            auto filterComma = filter!"true"(to!S(", "));
            assert(join(filteredWords, filterComma) == "日本語, paul, jerry");
            assert(join(filteredWordsArr, filterComma) == "日本語, paul, jerry");
            assert(join(filteredLenWordsArr, filterComma) == "日本語, paul, jerry");
            assert(join(filter!"true"(words), filterComma) == "日本語, paul, jerry");
            assert(join(words, filterComma) == "日本語, paul, jerry");
        }}

        assert(join(filteredWords) == "日本語pauljerry");
        assert(join(filteredWordsArr) == "日本語pauljerry");
        assert(join(filteredLenWordsArr) == "日本語pauljerry");
        assert(join(filter!"true"(words)) == "日本語pauljerry");
        assert(join(words) == "日本語pauljerry");

        assert(join(filteredWords, filter!"true"(", ")) == "日本語, paul, jerry");
        assert(join(filteredWordsArr, filter!"true"(", ")) == "日本語, paul, jerry");
        assert(join(filteredLenWordsArr, filter!"true"(", ")) == "日本語, paul, jerry");
        assert(join(filter!"true"(words), filter!"true"(", ")) == "日本語, paul, jerry");
        assert(join(words, filter!"true"(", ")) == "日本語, paul, jerry");

        assert(join(filter!"true"(cast(typeof(filteredWordsArr))[]), ", ").empty);
        assert(join(cast(typeof(filteredWordsArr))[], ", ").empty);
        assert(join(cast(typeof(filteredLenWordsArr))[], ", ").empty);
        assert(join(filter!"true"(cast(R[])[]), ", ").empty);
        assert(join(cast(R[])[], ", ").empty);

        assert(join(filter!"true"(cast(typeof(filteredWordsArr))[])).empty);
        assert(join(cast(typeof(filteredWordsArr))[]).empty);
        assert(join(cast(typeof(filteredLenWordsArr))[]).empty);

        assert(join(filter!"true"(cast(R[])[])).empty);
        assert(join(cast(R[])[]).empty);
    }}

    assert(join([[1, 2], [41, 42]], [5, 6]) == [1, 2, 5, 6, 41, 42]);
    assert(join([[1, 2], [41, 42]], cast(int[])[]) == [1, 2, 41, 42]);
    assert(join([[1, 2]], [5, 6]) == [1, 2]);
    assert(join(cast(int[][])[], [5, 6]).empty);

    assert(join([[1, 2], [41, 42]]) == [1, 2, 41, 42]);
    assert(join(cast(int[][])[]).empty);

    alias f = filter!"true";
    assert(join([[1, 2], [41, 42]],          [5, 6]) == [1, 2, 5, 6, 41, 42]);
    assert(join(f([[1, 2], [41, 42]]),       [5, 6]) == [1, 2, 5, 6, 41, 42]);
    assert(join([f([1, 2]), f([41, 42])],    [5, 6]) == [1, 2, 5, 6, 41, 42]);
    assert(join(f([f([1, 2]), f([41, 42])]), [5, 6]) == [1, 2, 5, 6, 41, 42]);
    assert(join([[1, 2], [41, 42]],          f([5, 6])) == [1, 2, 5, 6, 41, 42]);
    assert(join(f([[1, 2], [41, 42]]),       f([5, 6])) == [1, 2, 5, 6, 41, 42]);
    assert(join([f([1, 2]), f([41, 42])],    f([5, 6])) == [1, 2, 5, 6, 41, 42]);
    assert(join(f([f([1, 2]), f([41, 42])]), f([5, 6])) == [1, 2, 5, 6, 41, 42]);
}

// https://issues.dlang.org/show_bug.cgi?id=10683
@safe unittest
{
    import std.range : join;
    import std.typecons : tuple;
    assert([[tuple(1)]].join == [tuple(1)]);
    assert([[tuple("x")]].join == [tuple("x")]);
}

// https://issues.dlang.org/show_bug.cgi?id=13877
@safe unittest
{
    // Test that the range is iterated only once.
    import std.algorithm.iteration : map;
    int c = 0;
    auto j1 = [1, 2, 3].map!(_ => [c++]).join;
    assert(c == 3);
    assert(j1 == [0, 1, 2]);

    c = 0;
    auto j2 = [1, 2, 3].map!(_ => [c++]).join(9);
    assert(c == 3);
    assert(j2 == [0, 9, 1, 9, 2]);

    c = 0;
    auto j3 = [1, 2, 3].map!(_ => [c++]).join([9]);
    assert(c == 3);
    assert(j3 == [0, 9, 1, 9, 2]);
}


/++
    Replace occurrences of `from` with `to` in `subject` in a new array.

    Params:
        subject = the array to scan
        from = the item to replace
        to = the item to replace all instances of `from` with

    Returns:
        A new array without changing the contents of `subject`, or the original
        array if no match is found.

    See_Also:
        $(REF substitute, std,algorithm,iteration) for a lazy replace.
 +/
E[] replace(E, R1, R2)(E[] subject, R1 from, R2 to)
if ((isForwardRange!R1 && isForwardRange!R2 && (hasLength!R2 || isSomeString!R2)) ||
    is(Unqual!E : Unqual!R1))
{
    size_t changed = 0;
    return replace(subject, from, to, changed);
}

///
@safe unittest
{
    assert("Hello Wörld".replace("o Wö", "o Wo") == "Hello World");
    assert("Hello Wörld".replace("l", "h") == "Hehho Wörhd");
}

@safe unittest
{
    assert([1, 2, 3, 4, 2].replace([2], [5]) == [1, 5, 3, 4, 5]);
    assert([3, 3, 3].replace([3], [0]) == [0, 0, 0]);
    assert([3, 3, 4, 3].replace([3, 3], [1, 1, 1]) == [1, 1, 1, 4, 3]);
}

// https://issues.dlang.org/show_bug.cgi?id=18215
@safe unittest
{
    auto arr = ["aaa.dd", "b"];
    arr = arr.replace("aaa.dd", ".");
    assert(arr == [".", "b"]);

    arr = ["_", "_", "aaa.dd", "b", "c", "aaa.dd", "e"];
    arr = arr.replace("aaa.dd", ".");
    assert(arr == ["_", "_", ".", "b", "c", ".", "e"]);
}

// https://issues.dlang.org/show_bug.cgi?id=18215
@safe unittest
{
    assert([[0], [1, 2], [0], [3]].replace([0], [4]) == [[4], [1, 2], [4], [3]]);
    assert([[0], [1, 2], [0], [3], [1, 2]]
            .replace([1, 2], [0]) == [[0], [0], [0], [3], [0]]);
    assert([[0], [1, 2], [0], [3], [1, 2], [0], [1, 2]]
            .replace([[0], [1, 2]], [[4]]) == [[4], [0], [3], [1, 2], [4]]);
}

// https://issues.dlang.org/show_bug.cgi?id=10930
@safe unittest
{
    assert([0, 1, 2].replace(1, 4) == [0, 4, 2]);
    assert("äbö".replace('ä', 'a') == "abö");
}

// empty array
@safe unittest
{
    int[] arr;
    assert(replace(arr, 1, 2) == []);
}

/++
    Replace occurrences of `from` with `to` in `subject` in a new array.
    `changed` counts how many replacements took place.

    Params:
        subject = the array to scan
        from = the item to replace
        to = the item to replace all instances of `from` with
        changed = the number of replacements

    Returns:
        A new array without changing the contents of `subject`, or the original
        array if no match is found.
 +/
E[] replace(E, R1, R2)(E[] subject, R1 from, R2 to, ref size_t changed)
if ((isForwardRange!R1 && isForwardRange!R2 && (hasLength!R2 || isSomeString!R2)) ||
    is(Unqual!E : Unqual!R1))
{
    import std.algorithm.searching : find;
    import std.range : dropOne;

    static if (isInputRange!R1)
    {
        if (from.empty) return subject;
        alias rSave = a => a.save;
    }
    else
    {
        alias rSave = a => a;
    }

    auto balance = find(subject, rSave(from));
    if (balance.empty)
        return subject;

    auto app = appender!(E[])();
    app.put(subject[0 .. subject.length - balance.length]);
    app.put(rSave(to));
    ++changed;
    // replacing an element in an array is different to a range replacement
    static if (is(Unqual!E : Unqual!R1))
        replaceInto(app, balance.dropOne, from, to, changed);
    else
        replaceInto(app, balance[from.length .. $], from, to, changed);

    return app.data;
}

///
@safe unittest
{
    size_t changed = 0;
    assert("Hello Wörld".replace("o Wö", "o Wo", changed) == "Hello World");
    assert(changed == 1);

    changed = 0;
    assert("Hello Wörld".replace("l", "h", changed) == "Hehho Wörhd");
    import std.stdio : writeln;
    writeln(changed);
    assert(changed == 3);
}

/++
    Replace occurrences of `from` with `to` in `subject` and output the result into
    `sink`.

    Params:
        sink = an $(REF_ALTTEXT output range, isOutputRange, std,range,primitives)
        subject = the array to scan
        from = the item to replace
        to = the item to replace all instances of `from` with

    See_Also:
        $(REF substitute, std,algorithm,iteration) for a lazy replace.
 +/
void replaceInto(E, Sink, R1, R2)(Sink sink, E[] subject, R1 from, R2 to)
if (isOutputRange!(Sink, E) &&
    ((isForwardRange!R1 && isForwardRange!R2 && (hasLength!R2 || isSomeString!R2)) ||
    is(Unqual!E : Unqual!R1)))
{
    size_t changed = 0;
    replaceInto(sink, subject, from, to, changed);
}

///
@safe unittest
{
    auto arr = [1, 2, 3, 4, 5];
    auto from = [2, 3];
    auto to = [4, 6];
    auto sink = appender!(int[])();

    replaceInto(sink, arr, from, to);

    assert(sink.data == [1, 4, 6, 4, 5]);
}

// empty array
@safe unittest
{
    auto sink = appender!(int[])();
    int[] arr;
    replaceInto(sink, arr, 1, 2);
    assert(sink.data == []);
}

@safe unittest
{
    import std.algorithm.comparison : cmp;
    import std.conv : to;

    static foreach (S; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[]))
    {
        static foreach (T; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[]))
        {{
            auto s = to!S("This is a foo foo list");
            auto from = to!T("foo");
            auto into = to!S("silly");
            S r;
            int i;

            r = replace(s, from, into);
            i = cmp(r, "This is a silly silly list");
            assert(i == 0);

            r = replace(s, to!S(""), into);
            i = cmp(r, "This is a foo foo list");
            assert(i == 0);

            assert(replace(r, to!S("won't find this"), to!S("whatever")) is r);
        }}
    }

    immutable s = "This is a foo foo list";
    assert(replace(s, "foo", "silly") == "This is a silly silly list");
}

@safe unittest
{
    import std.algorithm.searching : skipOver;
    import std.conv : to;

    struct CheckOutput(C)
    {
        C[] desired;
        this(C[] arr){ desired = arr; }
        void put(C[] part){ assert(skipOver(desired, part)); }
    }
    static foreach (S; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[]))
    {{
        alias Char = ElementEncodingType!S;
        S s = to!S("yet another dummy text, yet another ...");
        S from = to!S("yet another");
        S into = to!S("some");
        replaceInto(CheckOutput!(Char)(to!S("some dummy text, some ..."))
                    , s, from, into);
    }}
}

// https://issues.dlang.org/show_bug.cgi?id=10930
@safe unittest
{
    auto sink = appender!(int[])();
    replaceInto(sink, [0, 1, 2], 1, 5);
    assert(sink.data == [0, 5, 2]);

    auto sink2 = appender!(dchar[])();
    replaceInto(sink2, "äbö", 'ä', 'a');
    assert(sink2.data == "abö");
}

/++
    Replace occurrences of `from` with `to` in `subject` and output the result into
    `sink`. `changed` counts how many replacements took place.

    Params:
        sink = an $(REF_ALTTEXT output range, isOutputRange, std,range,primitives)
        subject = the array to scan
        from = the item to replace
        to = the item to replace all instances of `from` with
        changed = the number of replacements
 +/
void replaceInto(E, Sink, R1, R2)(Sink sink, E[] subject, R1 from, R2 to, ref size_t changed)
if (isOutputRange!(Sink, E) &&
    ((isForwardRange!R1 && isForwardRange!R2 && (hasLength!R2 || isSomeString!R2)) ||
    is(Unqual!E : Unqual!R1)))
{
    import std.algorithm.searching : find;
    import std.range : dropOne;

    static if (isInputRange!R1)
    {
        if (from.empty)
        {
            sink.put(subject);
            return;
        }
        alias rSave = a => a.save;
    }
    else
    {
        alias rSave = a => a;
    }
    for (;;)
    {
        auto balance = find(subject, rSave(from));
        if (balance.empty)
        {
            sink.put(subject);
            break;
        }
        sink.put(subject[0 .. subject.length - balance.length]);
        sink.put(rSave(to));
        ++changed;
        // replacing an element in an array is different to a range replacement
        static if (is(Unqual!E : Unqual!R1))
            subject = balance.dropOne;
        else
            subject = balance[from.length .. $];
    }
}

///
@safe unittest
{
    auto arr = [1, 2, 3, 4, 5];
    auto from = [2, 3];
    auto to = [4, 6];
    auto sink = appender!(int[])();

    size_t changed = 0;
    replaceInto(sink, arr, from, to, changed);

    assert(sink.data == [1, 4, 6, 4, 5]);
    assert(changed == 1);
}

/++
    Replaces elements from `array` with indices ranging from `from`
    (inclusive) to `to` (exclusive) with the range `stuff`.

    Params:
        subject = the array to scan
        from = the starting index
        to = the ending index
        stuff = the items to replace in-between `from` and `to`

    Returns:
        A new array without changing the contents of `subject`.

    See_Also:
        $(REF substitute, std,algorithm,iteration) for a lazy replace.
 +/
T[] replace(T, Range)(T[] subject, size_t from, size_t to, Range stuff)
if (isInputRange!Range &&
    (is(ElementType!Range : T) ||
    isSomeString!(T[]) && is(ElementType!Range : dchar)))
{
    static if (hasLength!Range && is(ElementEncodingType!Range : T))
    {
        import std.algorithm.mutation : copy;
        assert(from <= to, "from must be before or equal to to");
        immutable sliceLen = to - from;
        auto retval = new Unqual!(T)[](subject.length - sliceLen + stuff.length);
        retval[0 .. from] = subject[0 .. from];

        if (!stuff.empty)
            copy(stuff, retval[from .. from + stuff.length]);

        retval[from + stuff.length .. $] = subject[to .. $];
        static if (is(T == const) || is(T == immutable))
        {
            return () @trusted { return cast(T[]) retval; } ();
        }
        else
        {
            return cast(T[]) retval;
        }
    }
    else
    {
        auto app = appender!(T[])();
        app.put(subject[0 .. from]);
        app.put(stuff);
        app.put(subject[to .. $]);
        return app.data;
    }
}

///
@safe unittest
{
    auto a = [ 1, 2, 3, 4 ];
    auto b = a.replace(1, 3, [ 9, 9, 9 ]);
    assert(a == [ 1, 2, 3, 4 ]);
    assert(b == [ 1, 9, 9, 9, 4 ]);
}

@system unittest
{
    import core.exception;
    import std.algorithm.iteration : filter;
    import std.conv : to;
    import std.exception;


    auto a = [ 1, 2, 3, 4 ];
    assert(replace(a, 0, 0, [5, 6, 7]) == [5, 6, 7, 1, 2, 3, 4]);
    assert(replace(a, 0, 2, cast(int[])[]) == [3, 4]);
    assert(replace(a, 0, 4, [5, 6, 7]) == [5, 6, 7]);
    assert(replace(a, 0, 2, [5, 6, 7]) == [5, 6, 7, 3, 4]);
    assert(replace(a, 2, 4, [5, 6, 7]) == [1, 2, 5, 6, 7]);

    assert(replace(a, 0, 0, filter!"true"([5, 6, 7])) == [5, 6, 7, 1, 2, 3, 4]);
    assert(replace(a, 0, 2, filter!"true"(cast(int[])[])) == [3, 4]);
    assert(replace(a, 0, 4, filter!"true"([5, 6, 7])) == [5, 6, 7]);
    assert(replace(a, 0, 2, filter!"true"([5, 6, 7])) == [5, 6, 7, 3, 4]);
    assert(replace(a, 2, 4, filter!"true"([5, 6, 7])) == [1, 2, 5, 6, 7]);
    assert(a == [ 1, 2, 3, 4 ]);

    void testStr(T, U)(string file = __FILE__, size_t line = __LINE__)
    {

        auto l = to!T("hello");
        auto r = to!U(" world");

        enforce(replace(l, 0, 0, r) == " worldhello",
                new AssertError("testStr failure 1", file, line));
        enforce(replace(l, 0, 3, r) == " worldlo",
                new AssertError("testStr failure 2", file, line));
        enforce(replace(l, 3, l.length, r) == "hel world",
                new AssertError("testStr failure 3", file, line));
        enforce(replace(l, 0, l.length, r) == " world",
                new AssertError("testStr failure 4", file, line));
        enforce(replace(l, l.length, l.length, r) == "hello world",
                new AssertError("testStr failure 5", file, line));
    }

    testStr!(string, string)();
    testStr!(string, wstring)();
    testStr!(string, dstring)();
    testStr!(wstring, string)();
    testStr!(wstring, wstring)();
    testStr!(wstring, dstring)();
    testStr!(dstring, string)();
    testStr!(dstring, wstring)();
    testStr!(dstring, dstring)();

    enum s = "0123456789";
    enum w = "⁰¹²³⁴⁵⁶⁷⁸⁹"w;
    enum d = "⁰¹²³⁴⁵⁶⁷⁸⁹"d;

    assert(replace(s, 0, 0, "***") == "***0123456789");
    assert(replace(s, 10, 10, "***") == "0123456789***");
    assert(replace(s, 3, 8, "1012") == "012101289");
    assert(replace(s, 0, 5, "43210") == "4321056789");
    assert(replace(s, 5, 10, "43210") == "0123443210");

    assert(replace(w, 0, 0, "***"w) == "***⁰¹²³⁴⁵⁶⁷⁸⁹"w);
    assert(replace(w, 10, 10, "***"w) == "⁰¹²³⁴⁵⁶⁷⁸⁹***"w);
    assert(replace(w, 3, 8, "¹⁰¹²"w) == "⁰¹²¹⁰¹²⁸⁹"w);
    assert(replace(w, 0, 5, "⁴³²¹⁰"w) == "⁴³²¹⁰⁵⁶⁷⁸⁹"w);
    assert(replace(w, 5, 10, "⁴³²¹⁰"w) == "⁰¹²³⁴⁴³²¹⁰"w);

    assert(replace(d, 0, 0, "***"d) == "***⁰¹²³⁴⁵⁶⁷⁸⁹"d);
    assert(replace(d, 10, 10, "***"d) == "⁰¹²³⁴⁵⁶⁷⁸⁹***"d);
    assert(replace(d, 3, 8, "¹⁰¹²"d) == "⁰¹²¹⁰¹²⁸⁹"d);
    assert(replace(d, 0, 5, "⁴³²¹⁰"d) == "⁴³²¹⁰⁵⁶⁷⁸⁹"d);
    assert(replace(d, 5, 10, "⁴³²¹⁰"d) == "⁰¹²³⁴⁴³²¹⁰"d);
}

// https://issues.dlang.org/show_bug.cgi?id=18166
@safe pure unittest
{
    auto str = replace("aaaaa"d, 1, 4, "***"d);
    assert(str == "a***a");
}

/++
    Replaces elements from `array` with indices ranging from `from`
    (inclusive) to `to` (exclusive) with the range `stuff`. Expands or
    shrinks the array as needed.

    Params:
        array = the array to scan
        from = the starting index
        to = the ending index
        stuff = the items to replace in-between `from` and `to`
 +/
void replaceInPlace(T, Range)(ref T[] array, size_t from, size_t to, Range stuff)
if (is(typeof(replace(array, from, to, stuff))))
{
    static if (isDynamicArray!Range &&
              is(Unqual!(ElementEncodingType!Range) == T) &&
              !isNarrowString!(T[]))
    {
        // optimized for homogeneous arrays that can be overwritten.
        import std.algorithm.mutation : remove;
        import std.typecons : tuple;

        if (overlap(array, stuff).length)
        {
            // use slower/conservative method
            array = array[0 .. from] ~ stuff ~ array[to .. $];
        }
        else if (stuff.length <= to - from)
        {
            // replacement reduces length
            immutable stuffEnd = from + stuff.length;
            array[from .. stuffEnd] = stuff[];
            if (stuffEnd < to)
                array = remove(array, tuple(stuffEnd, to));
        }
        else
        {
            // replacement increases length
            // @@@TODO@@@: optimize this
            immutable replaceLen = to - from;
            array[from .. to] = stuff[0 .. replaceLen];
            insertInPlace(array, to, stuff[replaceLen .. $]);
        }
    }
    else
    {
        // default implementation, just do what replace does.
        array = replace(array, from, to, stuff);
    }
}

///
@safe unittest
{
    int[] a = [1, 4, 5];
    replaceInPlace(a, 1u, 2u, [2, 3, 4]);
    assert(a == [1, 2, 3, 4, 5]);
    replaceInPlace(a, 1u, 2u, cast(int[])[]);
    assert(a == [1, 3, 4, 5]);
    replaceInPlace(a, 1u, 3u, a[2 .. 4]);
    assert(a == [1, 4, 5, 5]);
}

// https://issues.dlang.org/show_bug.cgi?id=12889
@safe unittest
{
    int[1][] arr = [[0], [1], [2], [3], [4], [5], [6]];
    int[1][] stuff = [[0], [1]];
    replaceInPlace(arr, 4, 6, stuff);
    assert(arr == [[0], [1], [2], [3], [0], [1], [6]]);
}

@system unittest
{
    // https://issues.dlang.org/show_bug.cgi?id=14925
    char[] a = "mon texte 1".dup;
    char[] b = "abc".dup;
    replaceInPlace(a, 4, 9, b);
    assert(a == "mon abc 1");

    // ensure we can replace in place with different encodings
    string unicoded = "\U00010437";
    string unicodedLong = "\U00010437aaaaa";
    string base = "abcXXXxyz";
    string result = "abc\U00010437xyz";
    string resultLong = "abc\U00010437aaaaaxyz";
    size_t repstart = 3;
    size_t repend = 3 + 3;

    void testStringReplaceInPlace(T, U)()
    {
        import std.algorithm.comparison : equal;
        import std.conv;
        auto a = unicoded.to!(U[]);
        auto b = unicodedLong.to!(U[]);

        auto test = base.to!(T[]);

        test.replaceInPlace(repstart, repend, a);
        assert(equal(test, result), "Failed for types " ~ T.stringof ~ " and " ~ U.stringof);

        test = base.to!(T[]);

        test.replaceInPlace(repstart, repend, b);
        assert(equal(test, resultLong), "Failed for types " ~ T.stringof ~ " and " ~ U.stringof);
    }

    import std.meta : AliasSeq;
    alias allChars = AliasSeq!(char, immutable(char), const(char),
                         wchar, immutable(wchar), const(wchar),
                         dchar, immutable(dchar), const(dchar));
    foreach (T; allChars)
        foreach (U; allChars)
            testStringReplaceInPlace!(T, U)();

    void testInout(inout(int)[] a)
    {
        // will be transferred to the 'replace' function
        replaceInPlace(a, 1, 2, [1,2,3]);
    }
}

@safe unittest
{
    // the constraint for the first overload used to match this, which wouldn't compile.
    import std.algorithm.comparison : equal;
    long[] a = [1L, 2, 3];
    int[] b = [4, 5, 6];
    a.replaceInPlace(1, 2, b);
    assert(equal(a, [1L, 4, 5, 6, 3]));
}

@system unittest
{
    import core.exception;
    import std.algorithm.comparison : equal;
    import std.algorithm.iteration : filter;
    import std.conv : to;
    import std.exception;


    bool test(T, U, V)(T orig, size_t from, size_t to, U toReplace, V result)
    {
        {
            static if (is(T == typeof(T.init.dup)))
                auto a = orig.dup;
            else
                auto a = orig.idup;

            a.replaceInPlace(from, to, toReplace);
            if (!equal(a, result))
                return false;
        }

        static if (isInputRange!U)
        {
            orig.replaceInPlace(from, to, filter!"true"(toReplace));
            return equal(orig, result);
        }
        else
            return true;
    }

    assert(test([1, 2, 3, 4], 0, 0, [5, 6, 7], [5, 6, 7, 1, 2, 3, 4]));
    assert(test([1, 2, 3, 4], 0, 2, cast(int[])[], [3, 4]));
    assert(test([1, 2, 3, 4], 0, 4, [5, 6, 7], [5, 6, 7]));
    assert(test([1, 2, 3, 4], 0, 2, [5, 6, 7], [5, 6, 7, 3, 4]));
    assert(test([1, 2, 3, 4], 2, 4, [5, 6, 7], [1, 2, 5, 6, 7]));

    assert(test([1, 2, 3, 4], 0, 0, filter!"true"([5, 6, 7]), [5, 6, 7, 1, 2, 3, 4]));
    assert(test([1, 2, 3, 4], 0, 2, filter!"true"(cast(int[])[]), [3, 4]));
    assert(test([1, 2, 3, 4], 0, 4, filter!"true"([5, 6, 7]), [5, 6, 7]));
    assert(test([1, 2, 3, 4], 0, 2, filter!"true"([5, 6, 7]), [5, 6, 7, 3, 4]));
    assert(test([1, 2, 3, 4], 2, 4, filter!"true"([5, 6, 7]), [1, 2, 5, 6, 7]));

    void testStr(T, U)(string file = __FILE__, size_t line = __LINE__)
    {

        auto l = to!T("hello");
        auto r = to!U(" world");

        enforce(test(l, 0, 0, r, " worldhello"),
                new AssertError("testStr failure 1", file, line));
        enforce(test(l, 0, 3, r, " worldlo"),
                new AssertError("testStr failure 2", file, line));
        enforce(test(l, 3, l.length, r, "hel world"),
                new AssertError("testStr failure 3", file, line));
        enforce(test(l, 0, l.length, r, " world"),
                new AssertError("testStr failure 4", file, line));
        enforce(test(l, l.length, l.length, r, "hello world"),
                new AssertError("testStr failure 5", file, line));
    }

    testStr!(string, string)();
    testStr!(string, wstring)();
    testStr!(string, dstring)();
    testStr!(wstring, string)();
    testStr!(wstring, wstring)();
    testStr!(wstring, dstring)();
    testStr!(dstring, string)();
    testStr!(dstring, wstring)();
    testStr!(dstring, dstring)();
}

/++
    Replaces the first occurrence of `from` with `to` in `subject`.

    Params:
        subject = the array to scan
        from = the item to replace
        to = the item to replace `from` with

    Returns:
        A new array without changing the contents of `subject`, or the original
        array if no match is found.
 +/
E[] replaceFirst(E, R1, R2)(E[] subject, R1 from, R2 to)
if (isDynamicArray!(E[]) &&
    isForwardRange!R1 && is(typeof(appender!(E[])().put(from[0 .. 1]))) &&
    isForwardRange!R2 && is(typeof(appender!(E[])().put(to[0 .. 1]))))
{
    if (from.empty) return subject;
    static if (isSomeString!(E[]))
    {
        import std.string : indexOf;
        immutable idx = subject.indexOf(from);
    }
    else
    {
        import std.algorithm.searching : countUntil;
        immutable idx = subject.countUntil(from);
    }
    if (idx == -1)
        return subject;

    auto app = appender!(E[])();
    app.put(subject[0 .. idx]);
    app.put(to);

    static if (isSomeString!(E[]) && isSomeString!R1)
    {
        import std.utf : codeLength;
        immutable fromLength = codeLength!(Unqual!E, R1)(from);
    }
    else
        immutable fromLength = from.length;

    app.put(subject[idx + fromLength .. $]);

    return app.data;
}

///
@safe unittest
{
    auto a = [1, 2, 2, 3, 4, 5];
    auto b = a.replaceFirst([2], [1337]);
    assert(b == [1, 1337, 2, 3, 4, 5]);

    auto s = "This is a foo foo list";
    auto r = s.replaceFirst("foo", "silly");
    assert(r == "This is a silly foo list");
}

@safe unittest
{
    import std.algorithm.comparison : cmp;
    import std.conv : to;

    static foreach (S; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[],
                          const(char[]), immutable(char[])))
    {
        static foreach (T; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[],
                              const(char[]), immutable(char[])))
        {{
            auto s = to!S("This is a foo foo list");
            auto s2 = to!S("Thüs is a ßöö foo list");
            auto from = to!T("foo");
            auto from2 = to!T("ßöö");
            auto into = to!T("silly");
            auto into2 = to!T("sälly");

            S r1 = replaceFirst(s, from, into);
            assert(cmp(r1, "This is a silly foo list") == 0);

            S r11 = replaceFirst(s2, from2, into2);
            assert(cmp(r11, "Thüs is a sälly foo list") == 0,
                to!string(r11) ~ " : " ~ S.stringof ~ " " ~ T.stringof);

            S r2 = replaceFirst(r1, from, into);
            assert(cmp(r2, "This is a silly silly list") == 0);

            S r3 = replaceFirst(s, to!T(""), into);
            assert(cmp(r3, "This is a foo foo list") == 0);

            assert(replaceFirst(r3, to!T("won't find"), to!T("whatever")) is r3);
        }}
    }
}

// https://issues.dlang.org/show_bug.cgi?id=8187
@safe unittest
{
    auto res = ["a", "a"];
    assert(replace(res, "a", "b") == ["b", "b"]);
    assert(replaceFirst(res, "a", "b") == ["b", "a"]);
}

/++
    Replaces the last occurrence of `from` with `to` in `subject`.

    Params:
        subject = the array to scan
        from = the item to replace
        to = the item to replace `from` with

    Returns:
        A new array without changing the contents of `subject`, or the original
        array if no match is found.
 +/
E[] replaceLast(E, R1, R2)(E[] subject, R1 from , R2 to)
if (isDynamicArray!(E[]) &&
    isForwardRange!R1 && is(typeof(appender!(E[])().put(from[0 .. 1]))) &&
    isForwardRange!R2 && is(typeof(appender!(E[])().put(to[0 .. 1]))))
{
    import std.range : retro;
    if (from.empty) return subject;
    static if (isSomeString!(E[]))
    {
        import std.string : lastIndexOf;
        auto idx = subject.lastIndexOf(from);
    }
    else
    {
        import std.algorithm.searching : countUntil;
        auto idx = retro(subject).countUntil(retro(from));
    }

    if (idx == -1)
        return subject;

    static if (isSomeString!(E[]) && isSomeString!R1)
    {
        import std.utf : codeLength;
        auto fromLength = codeLength!(Unqual!E, R1)(from);
    }
    else
        auto fromLength = from.length;

    auto app = appender!(E[])();
    static if (isSomeString!(E[]))
        app.put(subject[0 .. idx]);
    else
        app.put(subject[0 .. $ - idx - fromLength]);

    app.put(to);

    static if (isSomeString!(E[]))
        app.put(subject[idx+fromLength .. $]);
    else
        app.put(subject[$ - idx .. $]);

    return app.data;
}

///
@safe unittest
{
    auto a = [1, 2, 2, 3, 4, 5];
    auto b = a.replaceLast([2], [1337]);
    assert(b == [1, 2, 1337, 3, 4, 5]);

    auto s = "This is a foo foo list";
    auto r = s.replaceLast("foo", "silly");
    assert(r == "This is a foo silly list", r);
}

@safe unittest
{
    import std.algorithm.comparison : cmp;
    import std.conv : to;

    static foreach (S; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[],
                          const(char[]), immutable(char[])))
    {
        static foreach (T; AliasSeq!(string, wstring, dstring, char[], wchar[], dchar[],
                              const(char[]), immutable(char[])))
        {{
            auto s = to!S("This is a foo foo list");
            auto s2 = to!S("Thüs is a ßöö ßöö list");
            auto from = to!T("foo");
            auto from2 = to!T("ßöö");
            auto into = to!T("silly");
            auto into2 = to!T("sälly");

            S r1 = replaceLast(s, from, into);
            assert(cmp(r1, "This is a foo silly list") == 0, to!string(r1));

            S r11 = replaceLast(s2, from2, into2);
            assert(cmp(r11, "Thüs is a ßöö sälly list") == 0,
                to!string(r11) ~ " : " ~ S.stringof ~ " " ~ T.stringof);

            S r2 = replaceLast(r1, from, into);
            assert(cmp(r2, "This is a silly silly list") == 0);

            S r3 = replaceLast(s, to!T(""), into);
            assert(cmp(r3, "This is a foo foo list") == 0);

            assert(replaceLast(r3, to!T("won't find"), to!T("whatever")) is r3);
        }}
    }
}

/++
    Creates a new array such that the items in `slice` are replaced with the
    items in `replacement`. `slice` and `replacement` do not need to be the
    same length. The result will grow or shrink based on the items given.

    Params:
        s = the base of the new array
        slice = the slice of `s` to be replaced
        replacement = the items to replace `slice` with

    Returns:
        A new array that is `s` with `slice` replaced by
        `replacement[]`.

    See_Also:
        $(REF substitute, std,algorithm,iteration) for a lazy replace.
 +/
inout(T)[] replaceSlice(T)(inout(T)[] s, in T[] slice, in T[] replacement)
in
{
    // Verify that slice[] really is a slice of s[]
    assert(overlap(s, slice) is slice, "slice[] is not a subslice of s[]");
}
do
{
    auto result = new T[s.length - slice.length + replacement.length];
    immutable so = &slice[0] - &s[0];
    result[0 .. so] = s[0 .. so];
    result[so .. so + replacement.length] = replacement[];
    result[so + replacement.length .. result.length] =
        s[so + slice.length .. s.length];

    return () @trusted inout {
        return cast(inout(T)[]) result;
    }();
}

///
@safe unittest
{
    auto a = [1, 2, 3, 4, 5];
    auto b = replaceSlice(a, a[1 .. 4], [0, 0, 0]);

    assert(b == [1, 0, 0, 0, 5]);
}

@safe unittest
{
    import std.algorithm.comparison : cmp;

    string s = "hello";
    string slice = s[2 .. 4];

    auto r = replaceSlice(s, slice, "bar");
    int i;
    i = cmp(r, "hebaro");
    assert(i == 0);
}

/**
Implements an output range that appends data to an array. This is
recommended over $(D array ~= data) when appending many elements because it is more
efficient. `Appender` maintains its own array metadata locally, so it can avoid
the $(DDSUBLINK spec/arrays, capacity-reserve, performance hit of looking up slice `capacity`)
for each append.

Params:
    A = the array type to simulate.

See_Also: $(LREF appender)
 */
struct Appender(A)
if (isDynamicArray!A)
{
    import core.memory : GC;

    private alias T = ElementEncodingType!A;

    private struct Data
    {
        size_t capacity;
        Unqual!T[] arr;
        bool tryExtendBlock = false;
    }

    private Data* _data;

    /**
     * Constructs an `Appender` with a given array.  Note that this does not copy the
     * data.  If the array has a larger capacity as determined by `arr.capacity`,
     * it will be used by the appender.  After initializing an appender on an array,
     * appending to the original array will reallocate.
     */
    this(A arr) @trusted
    {
        // initialize to a given array.
        _data = new Data;
        _data.arr = cast(Unqual!T[]) arr; //trusted

        if (__ctfe)
            return;

        // We want to use up as much of the block the array is in as possible.
        // if we consume all the block that we can, then array appending is
        // safe WRT built-in append, and we can use the entire block.
        // We only do this for mutable types that can be extended.
        static if (isMutable!T && is(typeof(arr.length = size_t.max)))
        {
            immutable cap = arr.capacity; //trusted
            // Replace with "GC.setAttr( Not Appendable )" once pure (and fixed)
            if (cap > arr.length)
                arr.length = cap;
        }
        _data.capacity = arr.length;
    }

    /**
     * Reserve at least newCapacity elements for appending.  Note that more elements
     * may be reserved than requested. If `newCapacity <= capacity`, then nothing is
     * done.
     *
     * Params:
     *     newCapacity = the capacity the `Appender` should have
     */
    void reserve(size_t newCapacity)
    {
        if (_data)
        {
            if (newCapacity > _data.capacity)
                ensureAddable(newCapacity - _data.arr.length);
        }
        else
        {
            ensureAddable(newCapacity);
        }
    }

    /**
     * Returns: the capacity of the array (the maximum number of elements the
     * managed array can accommodate before triggering a reallocation). If any
     * appending will reallocate, `0` will be returned.
     */
    @property size_t capacity() const
    {
        return _data ? _data.capacity : 0;
    }

    /**
     * Use opSlice() from now on.
     * Returns: The managed array.
     */
    @property inout(T)[] data() inout @trusted
    {
        return this[];
    }

    /**
     * Returns: The managed array.
     */
    @property inout(T)[] opSlice() inout @trusted
    {
        /* @trusted operation:
         * casting Unqual!T[] to inout(T)[]
         */
        return cast(typeof(return))(_data ? _data.arr : null);
    }

    // ensure we can add nelems elements, resizing as necessary
    private void ensureAddable(size_t nelems)
    {
        if (!_data)
            _data = new Data;
        immutable len = _data.arr.length;
        immutable reqlen = len + nelems;

        if (_data.capacity >= reqlen)
            return;

        // need to increase capacity
        if (__ctfe)
        {
            static if (__traits(compiles, new Unqual!T[1]))
            {
                _data.arr.length = reqlen;
            }
            else
            {
                // avoid restriction of @disable this()
                _data.arr = _data.arr[0 .. _data.capacity];
                foreach (i; _data.capacity .. reqlen)
                    _data.arr ~= Unqual!T.init;
            }
            _data.arr = _data.arr[0 .. len];
            _data.capacity = reqlen;
        }
        else
        {
            // Time to reallocate.
            // We need to almost duplicate what's in druntime, except we
            // have better access to the capacity field.
            auto newlen = appenderNewCapacity!(T.sizeof)(_data.capacity, reqlen);
            // first, try extending the current block
            if (_data.tryExtendBlock)
            {
                immutable u = (() @trusted => GC.extend(_data.arr.ptr, nelems * T.sizeof, (newlen - len) * T.sizeof))();
                if (u)
                {
                    // extend worked, update the capacity
                    _data.capacity = u / T.sizeof;
                    return;
                }
            }


            // didn't work, must reallocate
            import core.checkedint : mulu;
            bool overflow;
            const nbytes = mulu(newlen, T.sizeof, overflow);
            if (overflow) assert(false, "the reallocation would exceed the "
                    ~ "available pointer range");

            auto bi = (() @trusted => GC.qalloc(nbytes, blockAttribute!T))();
            _data.capacity = bi.size / T.sizeof;
            import core.stdc.string : memcpy;
            if (len)
                () @trusted { memcpy(bi.base, _data.arr.ptr, len * T.sizeof); }();
            _data.arr = (() @trusted => (cast(Unqual!T*) bi.base)[0 .. len])();
            _data.tryExtendBlock = true;
            // leave the old data, for safety reasons
        }
    }

    private template canPutItem(U)
    {
        enum bool canPutItem =
            is(Unqual!U : Unqual!T) ||
            isSomeChar!T && isSomeChar!U;
    }
    private template canPutConstRange(Range)
    {
        enum bool canPutConstRange =
            isInputRange!(Unqual!Range) &&
            !isInputRange!Range &&
            is(typeof(Appender.init.put(Range.init.front)));
    }
    private template canPutRange(Range)
    {
        enum bool canPutRange =
            isInputRange!Range &&
            is(typeof(Appender.init.put(Range.init.front)));
    }

    /**
     * Appends `item` to the managed array. Performs encoding for
     * `char` types if `A` is a differently typed `char` array.
     *
     * Params:
     *     item = the single item to append
     */
    void put(U)(U item) if (canPutItem!U)
    {
        static if (isSomeChar!T && isSomeChar!U && T.sizeof < U.sizeof)
        {
            /* may throwable operation:
             * - std.utf.encode
             */
            // must do some transcoding around here
            import std.utf : encode;
            Unqual!T[T.sizeof == 1 ? 4 : 2] encoded;
            auto len = encode(encoded, item);
            put(encoded[0 .. len]);
        }
        else
        {
            import core.lifetime : emplace;

            ensureAddable(1);
            immutable len = _data.arr.length;

            auto bigData = (() @trusted => _data.arr.ptr[0 .. len + 1])();
            auto itemUnqual = (() @trusted => & cast() item)();
            emplace(&bigData[len], *itemUnqual);
            //We do this at the end, in case of exceptions
            _data.arr = bigData;
        }
    }

    // Const fixing hack.
    void put(Range)(Range items) if (canPutConstRange!Range)
    {
        alias p = put!(Unqual!Range);
        p(items);
    }

    /**
     * Appends an entire range to the managed array. Performs encoding for
     * `char` elements if `A` is a differently typed `char` array.
     *
     * Params:
     *     items = the range of items to append
     */
    void put(Range)(Range items) if (canPutRange!Range)
    {
        // note, we disable this branch for appending one type of char to
        // another because we can't trust the length portion.
        static if (!(isSomeChar!T && isSomeChar!(ElementType!Range) &&
                     !is(immutable Range == immutable T[])) &&
                    is(typeof(items.length) == size_t))
        {
            // optimization -- if this type is something other than a string,
            // and we are adding exactly one element, call the version for one
            // element.
            static if (!isSomeChar!T)
            {
                if (items.length == 1)
                {
                    put(items.front);
                    return;
                }
            }

            // make sure we have enough space, then add the items
            auto bigDataFun(size_t extra)
            {
                ensureAddable(extra);
                return (() @trusted => _data.arr.ptr[0 .. _data.arr.length + extra])();
            }
            auto bigData = bigDataFun(items.length);

            immutable len = _data.arr.length;
            immutable newlen = bigData.length;

            alias UT = Unqual!T;

            static if (is(typeof(_data.arr[] = items[])) &&
                !hasElaborateAssign!UT && isAssignable!(UT, ElementEncodingType!Range))
            {
                bigData[len .. newlen] = items[];
            }
            else
            {
                import core.internal.lifetime : emplaceRef;
                foreach (ref it ; bigData[len .. newlen])
                {
                    emplaceRef!T(it, items.front);
                    items.popFront();
                }
            }

            //We do this at the end, in case of exceptions
            _data.arr = bigData;
        }
        else static if (isSomeChar!T && isSomeChar!(ElementType!Range) &&
                        !is(immutable T == immutable ElementType!Range))
        {
            // need to decode and encode
            import std.utf : decodeFront;
            while (!items.empty)
            {
                auto c = items.decodeFront;
                put(c);
            }
        }
        else
        {
            //pragma(msg, Range.stringof);
            // Generic input range
            for (; !items.empty; items.popFront())
            {
                put(items.front);
            }
        }
    }

    /**
     * Appends to the managed array.
     *
     * See_Also: $(LREF Appender.put)
     */
    alias opOpAssign(string op : "~") = put;

    // only allow overwriting data on non-immutable and non-const data
    static if (isMutable!T)
    {
        /**
         * Clears the managed array.  This allows the elements of the array to be reused
         * for appending.
         *
         * Note: clear is disabled for immutable or const element types, due to the
         * possibility that `Appender` might overwrite immutable data.
         */
        void clear() @trusted pure nothrow
        {
            if (_data)
            {
                _data.arr = _data.arr.ptr[0 .. 0];
            }
        }

        /**
         * Shrinks the managed array to the given length.
         *
         * Throws: `Exception` if newlength is greater than the current array length.
         * Note: shrinkTo is disabled for immutable or const element types.
         */
        void shrinkTo(size_t newlength) @trusted pure
        {
            import std.exception : enforce;
            if (_data)
            {
                enforce(newlength <= _data.arr.length, "Attempting to shrink Appender with newlength > length");
                _data.arr = _data.arr.ptr[0 .. newlength];
            }
            else
                enforce(newlength == 0, "Attempting to shrink empty Appender with non-zero newlength");
        }
    }

    /**
     * Gives a string in the form of `Appender!(A)(data)`.
     *
     * Params:
     *     w = A `char` accepting
     *     $(REF_ALTTEXT output range, isOutputRange, std, range, primitives).
     *     fmt = A $(REF FormatSpec, std, format) which controls how the array
     *     is formatted.
     * Returns:
     *     A `string` if `writer` is not set; `void` otherwise.
     */
    string toString()() const
    {
        import std.format.spec : singleSpec;

        auto app = appender!string();
        auto spec = singleSpec("%s");
        immutable len = _data ? _data.arr.length : 0;
        // different reserve lengths because each element in a
        // non-string-like array uses two extra characters for `, `.
        static if (isSomeString!A)
        {
            app.reserve(len + 25);
        }
        else
        {
            // Multiplying by three is a very conservative estimate of
            // length, as it assumes each element is only one char
            app.reserve((len * 3) + 25);
        }
        toString(app, spec);
        return app.data;
    }

    import std.format.spec : FormatSpec;

    /// ditto
    template toString(Writer)
    if (isOutputRange!(Writer, char))
    {
        void toString(ref Writer w, scope const ref FormatSpec!char fmt) const
        {
            import std.format.write : formatValue;
            import std.range.primitives : put;
            put(w, Unqual!(typeof(this)).stringof);
            put(w, '(');
            formatValue(w, data, fmt);
            put(w, ')');
        }
    }
}

///
@safe pure nothrow unittest
{
    auto app = appender!string();
    string b = "abcdefg";
    foreach (char c; b)
        app.put(c);
    assert(app[] == "abcdefg");

    int[] a = [ 1, 2 ];
    auto app2 = appender(a);
    app2.put(3);
    app2.put([ 4, 5, 6 ]);
    assert(app2[] == [ 1, 2, 3, 4, 5, 6 ]);
}

@safe pure unittest
{
    import std.format : format;
    import std.format.spec : singleSpec;

    auto app = appender!(int[])();
    app.put(1);
    app.put(2);
    app.put(3);
    assert("%s".format(app) == "Appender!(int[])(%s)".format([1,2,3]));

    auto app2 = appender!string();
    auto spec = singleSpec("%s");
    app.toString(app2, spec);
    assert(app2[] == "Appender!(int[])([1, 2, 3])");

    auto app3 = appender!string();
    spec = singleSpec("%(%04d, %)");
    app.toString(app3, spec);
    assert(app3[] == "Appender!(int[])(0001, 0002, 0003)");
}

// https://issues.dlang.org/show_bug.cgi?id=17251
@safe pure nothrow unittest
{
    static struct R
    {
        int front() const { return 0; }
        bool empty() const { return true; }
        void popFront() {}
    }

    auto app = appender!(R[]);
    const(R)[1] r;
    app.put(r[0]);
    app.put(r[]);
}

// https://issues.dlang.org/show_bug.cgi?id=13300
@safe pure nothrow unittest
{
    static test(bool isPurePostblit)()
    {
        static if (!isPurePostblit)
            static int i;

        struct Simple
        {
            @disable this(); // Without this, it works.
            static if (!isPurePostblit)
                this(this) { i++; }
            else
                pure this(this) { }

            private:
            this(int tmp) { }
        }

        struct Range
        {
            @property Simple front() { return Simple(0); }
            void popFront() { count++; }
            @property empty() { return count < 3; }
            size_t count;
        }

        Range r;
        auto a = r.array();
    }

    static assert(__traits(compiles, () pure { test!true(); }));
    static assert(!__traits(compiles, () pure { test!false(); }));
}

// https://issues.dlang.org/show_bug.cgi?id=19572
@safe pure nothrow unittest
{
    static struct Struct
    {
        int value;

        int fun() const { return 23; }

        alias fun this;
    }

    Appender!(Struct[]) appender;

    appender.put(const(Struct)(42));

    auto result = appender[][0];

    assert(result.value != 23);
}

@safe pure unittest
{
    import std.conv : to;
    import std.utf : byCodeUnit;
    auto str = "ウェブサイト";
    auto wstr = appender!wstring();
    put(wstr, str.byCodeUnit);
    assert(wstr.data == str.to!wstring);
}

// https://issues.dlang.org/show_bug.cgi?id=21256
@safe pure unittest
{
    Appender!string app1;
    app1.toString();

    Appender!(int[]) app2;
    app2.toString();
}

//Calculates an efficient growth scheme based on the old capacity
//of data, and the minimum requested capacity.
//arg curLen: The current length
//arg reqLen: The length as requested by the user
//ret sugLen: A suggested growth.
private size_t appenderNewCapacity(size_t TSizeOf)(size_t curLen, size_t reqLen)
{
    import core.bitop : bsr;
    import std.algorithm.comparison : max;
    if (curLen == 0)
        return max(reqLen,8);
    ulong mult = 100 + (1000UL) / (bsr(curLen * TSizeOf) + 1);
    // limit to doubling the length, we don't want to grow too much
    if (mult > 200)
        mult = 200;
    auto sugLen = cast(size_t)((curLen * mult + 99) / 100);
    return max(reqLen, sugLen);
}

/**
 * A version of $(LREF Appender) that can update an array in-place.
 * It forwards all calls to an underlying appender implementation.
 * Any calls made to the appender also update the pointer to the
 * original array passed in.
 *
 * Tip: Use the `arrayPtr` overload of $(LREF appender) for construction with type-inference.
 *
 * Params:
 *     A = The array type to simulate
 */
struct RefAppender(A)
if (isDynamicArray!A)
{
    private alias T = ElementEncodingType!A;

    private
    {
        Appender!A impl;
        A* arr;
    }

    /**
     * Constructs a `RefAppender` with a given array reference.  This does not copy the
     * data.  If the array has a larger capacity as determined by `arr.capacity`, it
     * will be used by the appender.
     *
     * Note: Do not use built-in appending (i.e. `~=`) on the original array
     * until you are done with the appender, because subsequent calls to the appender
     * will reallocate the array data without those appends.
     *
     * Params:
     * arr = Pointer to an array. Must not be _null.
     */
    this(A* arr)
    {
        impl = Appender!A(*arr);
        this.arr = arr;
    }

    /** Wraps remaining `Appender` methods such as $(LREF put).
     * Params:
     * fn = Method name to call.
     * args = Arguments to pass to the method.
     */
    void opDispatch(string fn, Args...)(Args args)
    if (__traits(compiles, (Appender!A a) => mixin("a." ~ fn ~ "(args)")))
    {
        // we do it this way because we can't cache a void return
        scope(exit) *this.arr = impl[];
        mixin("return impl." ~ fn ~ "(args);");
    }

    /**
     * Appends `rhs` to the managed array.
     * Params:
     * rhs = Element or range.
     */
    void opOpAssign(string op : "~", U)(U rhs)
    if (__traits(compiles, (Appender!A a){ a.put(rhs); }))
    {
        scope(exit) *this.arr = impl[];
        impl.put(rhs);
    }

    /**
     * Returns the capacity of the array (the maximum number of elements the
     * managed array can accommodate before triggering a reallocation).  If any
     * appending will reallocate, `capacity` returns `0`.
     */
    @property size_t capacity() const
    {
        return impl.capacity;
    }

    /* Use opSlice() instead.
     * Returns: the managed array.
     */
    @property inout(T)[] data() inout
    {
        return impl[];
    }

    /**
     * Returns: the managed array.
     */
    @property inout(ElementEncodingType!A)[] opSlice() inout
    {
        return impl[];
    }
}

///
@safe pure nothrow
unittest
{
    int[] a = [1, 2];
    auto app2 = appender(&a);
    assert(app2[] == [1, 2]);
    assert(a == [1, 2]);
    app2 ~= 3;
    app2 ~= [4, 5, 6];
    assert(app2[] == [1, 2, 3, 4, 5, 6]);
    assert(a == [1, 2, 3, 4, 5, 6]);

    app2.reserve(5);
    assert(app2.capacity >= 5);
}

/++
    Convenience function that returns an $(LREF Appender) instance,
    optionally initialized with `array`.
 +/
Appender!A appender(A)()
if (isDynamicArray!A)
{
    return Appender!A(null);
}
/// ditto
Appender!(E[]) appender(A : E[], E)(auto ref A array)
{
    static assert(!isStaticArray!A || __traits(isRef, array),
        "Cannot create Appender from an rvalue static array");

    return Appender!(E[])(array);
}

@safe pure nothrow unittest
{
    auto app = appender!(char[])();
    string b = "abcdefg";
    foreach (char c; b) app.put(c);
    assert(app[] == "abcdefg");
}

@safe pure nothrow unittest
{
    auto app = appender!(char[])();
    string b = "abcdefg";
    foreach (char c; b) app ~= c;
    assert(app[] == "abcdefg");
}

@safe pure nothrow unittest
{
    int[] a = [ 1, 2 ];
    auto app2 = appender(a);
    assert(app2[] == [ 1, 2 ]);
    app2.put(3);
    app2.put([ 4, 5, 6 ][]);
    assert(app2[] == [ 1, 2, 3, 4, 5, 6 ]);
    app2.put([ 7 ]);
    assert(app2[] == [ 1, 2, 3, 4, 5, 6, 7 ]);
}

@safe pure nothrow unittest
{
    auto app4 = appender([]);
    try // shrinkTo may throw
    {
        app4.shrinkTo(0);
    }
    catch (Exception) assert(0);
}

// https://issues.dlang.org/show_bug.cgi?id=5663
// https://issues.dlang.org/show_bug.cgi?id=9725
@safe pure nothrow unittest
{
    import std.exception : assertNotThrown;

    static foreach (S; AliasSeq!(char[], const(char)[], string))
    {
        {
            Appender!S app5663i;
            assertNotThrown(app5663i.put("\xE3"));
            assert(app5663i[] == "\xE3");

            Appender!S app5663c;
            assertNotThrown(app5663c.put(cast(const(char)[])"\xE3"));
            assert(app5663c[] == "\xE3");

            Appender!S app5663m;
            assertNotThrown(app5663m.put("\xE3".dup));
            assert(app5663m[] == "\xE3");
        }
        // ditto for ~=
        {
            Appender!S app5663i;
            assertNotThrown(app5663i ~= "\xE3");
            assert(app5663i[] == "\xE3");

            Appender!S app5663c;
            assertNotThrown(app5663c ~= cast(const(char)[])"\xE3");
            assert(app5663c[] == "\xE3");

            Appender!S app5663m;
            assertNotThrown(app5663m ~= "\xE3".dup);
            assert(app5663m[] == "\xE3");
        }
    }
}

// https://issues.dlang.org/show_bug.cgi?id=10122
@safe pure nothrow unittest
{
    import std.exception : assertCTFEable;

    static struct S10122
    {
        int val;

        @disable this();
        this(int v) @safe pure nothrow { val = v; }
    }
    assertCTFEable!(
    {
        auto w = appender!(S10122[])();
        w.put(S10122(1));
        assert(w[].length == 1 && w[][0].val == 1);
    });
}

@safe pure nothrow unittest
{
    import std.exception : assertThrown;

    int[] a = [ 1, 2 ];
    auto app2 = appender(a);
    assert(app2[] == [ 1, 2 ]);
    app2 ~= 3;
    app2 ~= [ 4, 5, 6 ][];
    assert(app2[] == [ 1, 2, 3, 4, 5, 6 ]);
    app2 ~= [ 7 ];
    assert(app2[] == [ 1, 2, 3, 4, 5, 6, 7 ]);

    app2.reserve(5);
    assert(app2.capacity >= 5);

    try // shrinkTo may throw
    {
        app2.shrinkTo(3);
    }
    catch (Exception) assert(0);
    assert(app2[] == [ 1, 2, 3 ]);
    assertThrown(app2.shrinkTo(5));

    const app3 = app2;
    assert(app3.capacity >= 3);
    assert(app3[] == [1, 2, 3]);
}

///
@safe pure nothrow
unittest
{
    auto w = appender!string;
    // pre-allocate space for at least 10 elements (this avoids costly reallocations)
    w.reserve(10);
    assert(w.capacity >= 10);

    w.put('a'); // single elements
    w.put("bc"); // multiple elements

    // use the append syntax
    w ~= 'd';
    w ~= "ef";

    assert(w[] == "abcdef");
}

@safe pure nothrow unittest
{
    auto w = appender!string();
    w.reserve(4);
    cast(void) w.capacity;
    cast(void) w[];
    try
    {
        wchar wc = 'a';
        dchar dc = 'a';
        w.put(wc);    // decoding may throw
        w.put(dc);    // decoding may throw
    }
    catch (Exception) assert(0);
}

@safe pure nothrow unittest
{
    auto w = appender!(int[])();
    w.reserve(4);
    cast(void) w.capacity;
    cast(void) w[];
    w.put(10);
    w.put([10]);
    w.clear();
    try
    {
        w.shrinkTo(0);
    }
    catch (Exception) assert(0);

    struct N
    {
        int payload;
        alias payload this;
    }
    w.put(N(1));
    w.put([N(2)]);

    struct S(T)
    {
        @property bool empty() { return true; }
        @property T front() { return T.init; }
        void popFront() {}
    }
    S!int r;
    w.put(r);
}

// https://issues.dlang.org/show_bug.cgi?id=10690
@safe pure nothrow unittest
{
    import std.algorithm.iteration : filter;
    import std.typecons : tuple;
    [tuple(1)].filter!(t => true).array; // No error
    [tuple("A")].filter!(t => true).array; // error
}

@safe pure nothrow unittest
{
    import std.range;
    //Coverage for put(Range)
    struct S1
    {
    }
    struct S2
    {
        void opAssign(S2){}
    }
    auto a1 = Appender!(S1[])();
    auto a2 = Appender!(S2[])();
    auto au1 = Appender!(const(S1)[])();
    a1.put(S1().repeat().take(10));
    a2.put(S2().repeat().take(10));
    auto sc1 = const(S1)();
    au1.put(sc1.repeat().take(10));
}

@system pure unittest
{
    import std.range;
    struct S2
    {
        void opAssign(S2){}
    }
    auto au2 = Appender!(const(S2)[])();
    auto sc2 = const(S2)();
    au2.put(sc2.repeat().take(10));
}

@system pure nothrow unittest
{
    struct S
    {
        int* p;
    }

    auto a0 = Appender!(S[])();
    auto a1 = Appender!(const(S)[])();
    auto a2 = Appender!(immutable(S)[])();
    auto s0 = S(null);
    auto s1 = const(S)(null);
    auto s2 = immutable(S)(null);
    a1.put(s0);
    a1.put(s1);
    a1.put(s2);
    a1.put([s0]);
    a1.put([s1]);
    a1.put([s2]);
    a0.put(s0);
    static assert(!is(typeof(a0.put(a1))));
    static assert(!is(typeof(a0.put(a2))));
    a0.put([s0]);
    static assert(!is(typeof(a0.put([a1]))));
    static assert(!is(typeof(a0.put([a2]))));
    static assert(!is(typeof(a2.put(a0))));
    static assert(!is(typeof(a2.put(a1))));
    a2.put(s2);
    static assert(!is(typeof(a2.put([a0]))));
    static assert(!is(typeof(a2.put([a1]))));
    a2.put([s2]);
}

// https://issues.dlang.org/show_bug.cgi?id=9528
@safe pure nothrow unittest
{
    const(E)[] fastCopy(E)(E[] src) {
            auto app = appender!(const(E)[])();
            foreach (i, e; src)
                    app.put(e);
            return app[];
    }

    static class C {}
    static struct S { const(C) c; }
    S[] s = [ S(new C) ];

    auto t = fastCopy(s); // Does not compile
    assert(t.length == 1);
}

// https://issues.dlang.org/show_bug.cgi?id=10753
@safe pure unittest
{
    import std.algorithm.iteration : map;
    struct Foo {
       immutable dchar d;
    }
    struct Bar {
       immutable int x;
    }
    "12".map!Foo.array;
    [1, 2].map!Bar.array;
}

@safe pure nothrow unittest
{
    import std.algorithm.comparison : equal;

    //New appender signature tests
    alias mutARR = int[];
    alias conARR = const(int)[];
    alias immARR = immutable(int)[];

    mutARR mut;
    conARR con;
    immARR imm;

    auto app1 = Appender!mutARR(mut);                //Always worked. Should work. Should not create a warning.
    app1.put(7);
    assert(equal(app1[], [7]));
    static assert(!is(typeof(Appender!mutARR(con)))); //Never worked.  Should not work.
    static assert(!is(typeof(Appender!mutARR(imm)))); //Never worked.  Should not work.

    auto app2 = Appender!conARR(mut); //Always worked. Should work. Should not create a warning.
    app2.put(7);
    assert(equal(app2[], [7]));
    auto app3 = Appender!conARR(con); //Didn't work.   Now works.   Should not create a warning.
    app3.put(7);
    assert(equal(app3[], [7]));
    auto app4 = Appender!conARR(imm); //Didn't work.   Now works.   Should not create a warning.
    app4.put(7);
    assert(equal(app4[], [7]));

    //{auto app = Appender!immARR(mut);}                //Worked. Will cease to work. Creates warning.
    //static assert(!is(typeof(Appender!immARR(mut)))); //Worked. Will cease to work. Uncomment me after full deprecation.
    static assert(!is(typeof(Appender!immARR(con))));   //Never worked. Should not work.
    auto app5 = Appender!immARR(imm);                  //Didn't work.  Now works. Should not create a warning.
    app5.put(7);
    assert(equal(app5[], [7]));

    //Deprecated. Please uncomment and make sure this doesn't work:
    //char[] cc;
    //static assert(!is(typeof(Appender!string(cc))));

    //This should always work:
    auto app6 = appender!string(null);
    assert(app6[] == null);
    auto app7 = appender!(const(char)[])(null);
    assert(app7[] == null);
    auto app8 = appender!(char[])(null);
    assert(app8[] == null);
}

@safe pure nothrow unittest //Test large allocations (for GC.extend)
{
    import std.algorithm.comparison : equal;
    import std.range;
    Appender!(char[]) app;
    app.reserve(1); //cover reserve on non-initialized
    foreach (_; 0 .. 100_000)
        app.put('a');
    assert(equal(app[], 'a'.repeat(100_000)));
}

@safe pure nothrow unittest
{
    auto reference = new ubyte[](2048 + 1); //a number big enough to have a full page (EG: the GC extends)
    auto arr = reference.dup;
    auto app = appender(arr[0 .. 0]);
    app.reserve(1); //This should not trigger a call to extend
    app.put(ubyte(1)); //Don't clobber arr
    assert(reference[] == arr[]);
}

@safe pure nothrow unittest // clear method is supported only for mutable element types
{
    Appender!string app;
    app.put("foo");
    static assert(!__traits(compiles, app.clear()));
    assert(app[] == "foo");
}

@safe pure nothrow unittest
{
    static struct D//dynamic
    {
        int[] i;
        alias i this;
    }
    static struct S//static
    {
        int[5] i;
        alias i this;
    }
    static assert(!is(Appender!(char[5])));
    static assert(!is(Appender!D));
    static assert(!is(Appender!S));

    enum int[5] a = [];
    int[5] b;
    D d;
    S s;
    int[5] foo(){return a;}

    static assert(!is(typeof(appender(a))));
    static assert( is(typeof(appender(b))));
    static assert( is(typeof(appender(d))));
    static assert( is(typeof(appender(s))));
    static assert(!is(typeof(appender(foo()))));
}

@system unittest
{
    // https://issues.dlang.org/show_bug.cgi?id=13077
    static class A {}

    // reduced case
    auto w = appender!(shared(A)[])();
    w.put(new shared A());

    // original case
    import std.range;
    InputRange!(shared A) foo()
    {
        return [new shared A].inputRangeObject;
    }
    auto res = foo.array;
    assert(res.length == 1);
}

/++
    Convenience function that returns a $(LREF RefAppender) instance initialized
    with `arrayPtr`. Don't use null for the array pointer, use the other
    version of `appender` instead.
 +/
RefAppender!(E[]) appender(P : E[]*, E)(P arrayPtr)
{
    return RefAppender!(E[])(arrayPtr);
}

///
@safe pure nothrow
unittest
{
    int[] a = [1, 2];
    auto app2 = appender(&a);
    assert(app2[] == [1, 2]);
    assert(a == [1, 2]);
    app2 ~= 3;
    app2 ~= [4, 5, 6];
    assert(app2[] == [1, 2, 3, 4, 5, 6]);
    assert(a == [1, 2, 3, 4, 5, 6]);

    app2.reserve(5);
    assert(app2.capacity >= 5);
}

@safe pure nothrow unittest
{
    auto arr = new char[0];
    auto app = appender(&arr);
    string b = "abcdefg";
    foreach (char c; b) app.put(c);
    assert(app[] == "abcdefg");
    assert(arr == "abcdefg");
}

@safe pure nothrow unittest
{
    auto arr = new char[0];
    auto app = appender(&arr);
    string b = "abcdefg";
    foreach (char c; b) app ~= c;
    assert(app[] == "abcdefg");
    assert(arr == "abcdefg");
}

@safe pure nothrow unittest
{
    int[] a = [ 1, 2 ];
    auto app2 = appender(&a);
    assert(app2[] == [ 1, 2 ]);
    assert(a == [ 1, 2 ]);
    app2.put(3);
    app2.put([ 4, 5, 6 ][]);
    assert(app2[] == [ 1, 2, 3, 4, 5, 6 ]);
    assert(a == [ 1, 2, 3, 4, 5, 6 ]);
}

@safe pure nothrow unittest
{
    import std.exception : assertThrown;

    int[] a = [ 1, 2 ];
    auto app2 = appender(&a);
    assert(app2[] == [ 1, 2 ]);
    assert(a == [ 1, 2 ]);
    app2 ~= 3;
    app2 ~= [ 4, 5, 6 ][];
    assert(app2[] == [ 1, 2, 3, 4, 5, 6 ]);
    assert(a == [ 1, 2, 3, 4, 5, 6 ]);

    app2.reserve(5);
    assert(app2.capacity >= 5);

    try // shrinkTo may throw
    {
        app2.shrinkTo(3);
    }
    catch (Exception) assert(0);
    assert(app2[] == [ 1, 2, 3 ]);
    assertThrown(app2.shrinkTo(5));

    const app3 = app2;
    assert(app3.capacity >= 3);
    assert(app3[] == [1, 2, 3]);
}

// https://issues.dlang.org/show_bug.cgi?id=14605
@safe pure nothrow unittest
{
    static assert(isOutputRange!(Appender!(int[]), int));
    static assert(isOutputRange!(RefAppender!(int[]), int));
}

@safe pure nothrow unittest
{
    Appender!(int[]) app;
    short[] range = [1, 2, 3];
    app.put(range);
    assert(app[] == [1, 2, 3]);
}

@safe pure nothrow unittest
{
    string s = "hello".idup;
    char[] a = "hello".dup;
    auto appS = appender(s);
    auto appA = appender(a);
    put(appS, 'w');
    put(appA, 'w');
    s ~= 'a'; //Clobbers here?
    a ~= 'a'; //Clobbers here?
    assert(appS[] == "hellow");
    assert(appA[] == "hellow");
}

/++
Constructs a static array from a dynamic array whose length is known at compile-time.
The element type can be inferred or specified explicitly:

* $(D [1, 2].staticArray) returns `int[2]`
* $(D [1, 2].staticArray!float) returns `float[2]`

Note: `staticArray` returns by value, so expressions involving large arrays may be inefficient.

Params:
    a = The input array.

Returns: A static array constructed from `a`.
+/
pragma(inline, true) T[n] staticArray(T, size_t n)(auto ref T[n] a)
{
    return a;
}

/// static array from array literal
nothrow pure @safe @nogc unittest
{
    auto a = [0, 1].staticArray;
    static assert(is(typeof(a) == int[2]));
    assert(a == [0, 1]);
}

/// ditto
pragma(inline, true) U[n] staticArray(U, T, size_t n)(auto ref T[n] a)
if (!is(T == U) && is(T : U))
{
    return a[].staticArray!(U[n]);
}

/// static array from array with implicit casting of elements
nothrow pure @safe @nogc unittest
{
    auto b = [0, 1].staticArray!long;
    static assert(is(typeof(b) == long[2]));
    assert(b == [0, 1]);
}

nothrow pure @safe @nogc unittest
{
    int val = 3;
    static immutable gold = [1, 2, 3];
    [1, 2, val].staticArray.checkStaticArray!int([1, 2, 3]);

    @nogc void checkNogc()
    {
        [1, 2, val].staticArray.checkStaticArray!int(gold);
    }

    checkNogc();

    [1, 2, val].staticArray!double.checkStaticArray!double(gold);
    [1, 2, 3].staticArray!int.checkStaticArray!int(gold);

    [1, 2, 3].staticArray!(const(int)).checkStaticArray!(const(int))(gold);
    [1, 2, 3].staticArray!(const(double)).checkStaticArray!(const(double))(gold);
    {
        const(int)[3] a2 = [1, 2, 3].staticArray;
    }

    [cast(byte) 1, cast(byte) 129].staticArray.checkStaticArray!byte([1, -127]);
}

/**
Constructs a static array from a range.
When `a.length` is not known at compile time, the number of elements must be
given as a template argument (e.g. `myrange.staticArray!2`).
Size and type can be combined, if the source range elements are implicitly
convertible to the requested element type (eg: `2.iota.staticArray!(long[2])`).

When the range `a` is known at compile time, it can be given as a
template argument to avoid having to specify the number of elements
(e.g.: `staticArray!(2.iota)` or `staticArray!(double, 2.iota)`).

Params:
    a = The input range. If there are less elements than the specified length of the static array,
    the rest of it is default-initialized. If there are more than specified, the first elements
    up to the specified length are used.
    rangeLength = Output for the number of elements used from `a`. Optional.
*/
auto staticArray(size_t n, T)(scope T a)
if (isInputRange!T)
{
    alias U = ElementType!T;
    return staticArray!(U[n], U, n)(a);
}

/// ditto
auto staticArray(size_t n, T)(scope T a, out size_t rangeLength)
if (isInputRange!T)
{
    alias U = ElementType!T;
    return staticArray!(U[n], U, n)(a, rangeLength);
}

/// ditto
auto staticArray(Un : U[n], U, size_t n, T)(scope T a)
if (isInputRange!T && is(ElementType!T : U))
{
    size_t extraStackSpace;
    return staticArray!(Un, U, n)(a, extraStackSpace);
}

/// ditto
auto staticArray(Un : U[n], U, size_t n, T)(scope T a, out size_t rangeLength)
if (isInputRange!T && is(ElementType!T : U))
{
    import std.algorithm.mutation : uninitializedFill;
    import std.range : take;
    import core.internal.lifetime : emplaceRef;

    if (__ctfe)
    {
        size_t i;
        // Compile-time version to avoid unchecked memory access.
        Unqual!U[n] ret;
        for (auto iter = a.take(n); !iter.empty; iter.popFront())
        {
            ret[i] = iter.front;
            i++;
        }

        rangeLength = i;
        return (() @trusted => cast(U[n]) ret)();
    }

    auto ret = (() @trusted
    {
        Unqual!U[n] theArray = void;
        return theArray;
    }());

    size_t i;
    if (true)
    {
        // ret was void-initialized so let's initialize the unfilled part manually.
        // also prevents destructors to be called on uninitialized memory if
        // an exception is thrown
        scope (exit) ret[i .. $].uninitializedFill(U.init);

        for (auto iter = a.take(n); !iter.empty; iter.popFront())
        {
            emplaceRef!U(ret[i++], iter.front);
        }
    }

    rangeLength = i;
    return (() @trusted => cast(U[n]) ret)();
}

/// static array from range + size
nothrow pure @safe @nogc unittest
{
    import std.range : iota;

    auto input = 3.iota;
    auto a = input.staticArray!2;
    static assert(is(typeof(a) == int[2]));
    assert(a == [0, 1]);
    auto b = input.staticArray!(long[4]);
    static assert(is(typeof(b) == long[4]));
    assert(b == [0, 1, 2, 0]);
}

// Tests that code compiles when there is an elaborate destructor and exceptions
// are thrown. Unfortunately can't test that memory is initialized
// before having a destructor called on it.
@safe nothrow unittest
{
    // exists only to allow doing something in the destructor. Not tested
    // at the end because value appears to depend on implementation of the.
    // function.
    static int preventersDestroyed = 0;

    static struct CopyPreventer
    {
        bool on = false;
        this(this)
        {
            if (on) throw new Exception("Thou shalt not copy past me!");
        }

        ~this()
        {
            preventersDestroyed++;
        }
    }
    auto normalArray =
    [
        CopyPreventer(false),
        CopyPreventer(false),
        CopyPreventer(true),
        CopyPreventer(false),
        CopyPreventer(true),
    ];

    try
    {
        auto staticArray = normalArray.staticArray!5;
        assert(false);
    }
    catch (Exception e){}
}


nothrow pure @safe @nogc unittest
{
    auto a = [1, 2].staticArray;
    assert(is(typeof(a) == int[2]) && a == [1, 2]);

    import std.range : iota;

    2.iota.staticArray!2.checkStaticArray!int([0, 1]);
    2.iota.staticArray!(double[2]).checkStaticArray!double([0, 1]);
    2.iota.staticArray!(long[2]).checkStaticArray!long([0, 1]);
}

nothrow pure @safe @nogc unittest
{
    import std.range : iota;
    size_t copiedAmount;
    2.iota.staticArray!1(copiedAmount);
    assert(copiedAmount == 1);
    2.iota.staticArray!3(copiedAmount);
    assert(copiedAmount == 2);
}

/// ditto
auto staticArray(alias a)()
if (isInputRange!(typeof(a)))
{
    return .staticArray!(size_t(a.length))(a);
}

/// ditto
auto staticArray(U, alias a)()
if (isInputRange!(typeof(a)))
{
    return .staticArray!(U[size_t(a.length)])(a);
}

/// static array from CT range
nothrow pure @safe @nogc unittest
{
    import std.range : iota;

    enum a = staticArray!(2.iota);
    static assert(is(typeof(a) == int[2]));
    assert(a == [0, 1]);

    enum b = staticArray!(long, 2.iota);
    static assert(is(typeof(b) == long[2]));
    assert(b == [0, 1]);
}

nothrow pure @safe @nogc unittest
{
    import std.range : iota;

    enum a = staticArray!(2.iota);
    staticArray!(2.iota).checkStaticArray!int([0, 1]);
    staticArray!(double, 2.iota).checkStaticArray!double([0, 1]);
    staticArray!(long, 2.iota).checkStaticArray!long([0, 1]);
}

version (StdUnittest) private void checkStaticArray(T, T1, T2)(T1 a, T2 b) nothrow @safe pure @nogc
{
    static assert(is(T1 == T[T1.length]));
    assert(a == b, "a must be equal to b");
}