module Enumerable

static VALUE
enum_all(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo = MEMO_ENUM_NEW(Qtrue);
    WARN_UNUSED_BLOCK(argc);
    ENUM_BLOCK_CALL(all);
    return memo->v1;
}

Returns whether every element meets a given criterion.

If self has no element, returns true and argument or block are not used.

With no argument and no block, returns whether every element is truthy:

(1..4).all?           # => true
%w[a b c d].all?      # => true
[1, 2, nil].all?      # => false
['a','b', false].all? # => false
[].all?               # => true

With argument pattern and no block, returns whether for each element element, pattern === element:

(1..4).all?(Integer)                 # => true
(1..4).all?(Numeric)                 # => true
(1..4).all?(Float)                   # => false
%w[bar baz bat bam].all?(/ba/)       # => true
%w[bar baz bat bam].all?(/bar/)      # => false
%w[bar baz bat bam].all?('ba')       # => false
{foo: 0, bar: 1, baz: 2}.all?(Array) # => true
{foo: 0, bar: 1, baz: 2}.all?(Hash)  # => false
[].all?(Integer)                     # => true

With a block given, returns whether the block returns a truthy value for every element:

(1..4).all? {|element| element < 5 }                    # => true
(1..4).all? {|element| element < 4 }                    # => false
{foo: 0, bar: 1, baz: 2}.all? {|key, value| value < 3 } # => true
{foo: 0, bar: 1, baz: 2}.all? {|key, value| value < 2 } # => false

Related: any?, none? one?.

static VALUE
enum_any(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo = MEMO_ENUM_NEW(Qfalse);
    WARN_UNUSED_BLOCK(argc);
    ENUM_BLOCK_CALL(any);
    return memo->v1;
}

Returns whether any element meets a given criterion.

If self has no element, returns false and argument or block are not used.

With no argument and no block, returns whether any element is truthy:

(1..4).any?          # => true
%w[a b c d].any?     # => true
[1, false, nil].any? # => true
[].any?              # => false

With argument pattern and no block, returns whether for any element element, pattern === element:

[nil, false, 0].any?(Integer)        # => true
[nil, false, 0].any?(Numeric)        # => true
[nil, false, 0].any?(Float)          # => false
%w[bar baz bat bam].any?(/m/)        # => true
%w[bar baz bat bam].any?(/foo/)      # => false
%w[bar baz bat bam].any?('ba')       # => false
{foo: 0, bar: 1, baz: 2}.any?(Array) # => true
{foo: 0, bar: 1, baz: 2}.any?(Hash)  # => false
[].any?(Integer)                     # => false

With a block given, returns whether the block returns a truthy value for any element:

(1..4).any? {|element| element < 2 }                    # => true
(1..4).any? {|element| element < 1 }                    # => false
{foo: 0, bar: 1, baz: 2}.any? {|key, value| value < 1 } # => true
{foo: 0, bar: 1, baz: 2}.any? {|key, value| value < 0 } # => false

Related: all?, none?, one?.

static VALUE
enum_chain(int argc, VALUE *argv, VALUE obj)
{
    VALUE enums = rb_ary_new_from_values(1, &obj);
    rb_ary_cat(enums, argv, argc);
    return new_enum_chain(enums);
}

Returns an enumerator object generated from this enumerator and given enumerables.

e = (1..3).chain([4, 5])
e.to_a #=> [1, 2, 3, 4, 5]

static VALUE
enum_chunk(VALUE enumerable)
{
    VALUE enumerator;

    RETURN_SIZED_ENUMERATOR(enumerable, 0, 0, enum_size);

    enumerator = rb_obj_alloc(rb_cEnumerator);
    rb_ivar_set(enumerator, id_chunk_enumerable, enumerable);
    rb_ivar_set(enumerator, id_chunk_categorize, rb_block_proc());
    rb_block_call(enumerator, idInitialize, 0, 0, chunk_i, enumerator);
    return enumerator;
}

Each element in the returned enumerator is a 2-element array consisting of:

• A value returned by the block.

• An array (“chunk”) containing the element for which that value was returned, and all following elements for which the block returned the same value:

So that:

• Each block return value that is different from its predecessor begins a new chunk.

• Each block return value that is the same as its predecessor continues the same chunk.

Example:

e = (0..10).chunk {|i| (i / 3).floor } # => #<Enumerator: ...>
# The enumerator elements.
e.next # => [0, [0, 1, 2]]
e.next # => [1, [3, 4, 5]]
e.next # => [2, [6, 7, 8]]
e.next # => [3, [9, 10]]

Method chunk is especially useful for an enumerable that is already sorted. This example counts words for each initial letter in a large array of words:

# Get sorted words from a web page.
url = 'https://raw.githubusercontent.com/eneko/data-repository/master/data/words.txt'
words = URI::open(url).readlines
# Make chunks, one for each letter.
e = words.chunk {|word| word.upcase[0] } # => #<Enumerator: ...>
# Display 'A' through 'F'.
e.each {|c, words| p [c, words.length]; break if c == 'F' }

Output:

["A", 17096]
["B", 11070]
["C", 19901]
["D", 10896]
["E", 8736]
["F", 6860]

You can use the special symbol :_alone to force an element into its own separate chuck:

a = [0, 0, 1, 1]
e = a.chunk{|i| i.even? ? :_alone : true }
e.to_a # => [[:_alone, [0]], [:_alone, [0]], [true, [1, 1]]]

For example, you can put each line that contains a URL into its own chunk:

pattern = /http/
open(filename) { |f|
  f.chunk { |line| line =~ pattern ? :_alone : true }.each { |key, lines|
    pp lines
  }
}

You can use the special symbol :_separator or nil to force an element to be ignored (not included in any chunk):

a = [0, 0, -1, 1, 1]
e = a.chunk{|i| i < 0 ? :_separator : true }
e.to_a # => [[true, [0, 0]], [true, [1, 1]]]

Note that the separator does end the chunk:

a = [0, 0, -1, 1, -1, 1]
e = a.chunk{|i| i < 0 ? :_separator : true }
e.to_a # => [[true, [0, 0]], [true, [1]], [true, [1]]]

For example, the sequence of hyphens in svn log can be eliminated as follows:

sep = "-"*72 + "\n"
IO.popen("svn log README") { |f|
  f.chunk { |line|
    line != sep || nil
  }.each { |_, lines|
    pp lines
  }
}
#=> ["r20018 | knu | 2008-10-29 13:20:42 +0900 (Wed, 29 Oct 2008) | 2 lines\n",
#    "\n",
#    "* README, README.ja: Update the portability section.\n",
#    "\n"]
#   ["r16725 | knu | 2008-05-31 23:34:23 +0900 (Sat, 31 May 2008) | 2 lines\n",
#    "\n",
#    "* README, README.ja: Add a note about default C flags.\n",
#    "\n"]
#   ...

Paragraphs separated by empty lines can be parsed as follows:

File.foreach("README").chunk { |line|
  /\A\s*\z/ !~ line || nil
}.each { |_, lines|
  pp lines
}

static VALUE
enum_chunk_while(VALUE enumerable)
{
    VALUE enumerator;
    VALUE pred;

    pred = rb_block_proc();

    enumerator = rb_obj_alloc(rb_cEnumerator);
    rb_ivar_set(enumerator, id_slicewhen_enum, enumerable);
    rb_ivar_set(enumerator, id_slicewhen_pred, pred);
    rb_ivar_set(enumerator, id_slicewhen_inverted, Qtrue);

    rb_block_call(enumerator, idInitialize, 0, 0, slicewhen_i, enumerator);
    return enumerator;
}

Creates an enumerator for each chunked elements. The beginnings of chunks are defined by the block.

This method splits each chunk using adjacent elements, elt_before and elt_after, in the receiver enumerator. This method split chunks between elt_before and elt_after where the block returns false.

The block is called the length of the receiver enumerator minus one.

The result enumerator yields the chunked elements as an array. So each method can be called as follows:

enum.chunk_while { |elt_before, elt_after| bool }.each { |ary| ... }

Other methods of the Enumerator class and Enumerable module, such as to_a, map, etc., are also usable.

For example, one-by-one increasing subsequence can be chunked as follows:

a = [1,2,4,9,10,11,12,15,16,19,20,21]
b = a.chunk_while {|i, j| i+1 == j }
p b.to_a #=> [[1, 2], [4], [9, 10, 11, 12], [15, 16], [19, 20, 21]]
c = b.map {|a| a.length < 3 ? a : "#{a.first}-#{a.last}" }
p c #=> [[1, 2], [4], "9-12", [15, 16], "19-21"]
d = c.join(",")
p d #=> "1,2,4,9-12,15,16,19-21"

Increasing (non-decreasing) subsequence can be chunked as follows:

a = [0, 9, 2, 2, 3, 2, 7, 5, 9, 5]
p a.chunk_while {|i, j| i <= j }.to_a
#=> [[0, 9], [2, 2, 3], [2, 7], [5, 9], [5]]

Adjacent evens and odds can be chunked as follows: (Enumerable#chunk is another way to do it.)

a = [7, 5, 9, 2, 0, 7, 9, 4, 2, 0]
p a.chunk_while {|i, j| i.even? == j.even? }.to_a
#=> [[7, 5, 9], [2, 0], [7, 9], [4, 2, 0]]

Enumerable#slice_when does the same, except splitting when the block returns true instead of false.

static VALUE
enum_compact(VALUE obj)
{
    VALUE ary;

    ary = rb_ary_new();
    rb_block_call(obj, id_each, 0, 0, compact_i, ary);

    return ary;
}

Returns an array of all non-nil elements:

a = [nil, 0, nil, 'a', false, nil, false, nil, 'a', nil, 0, nil]
a.compact # => [0, "a", false, false, "a", 0]

static VALUE
enum_count(int argc, VALUE *argv, VALUE obj)
{
    VALUE item = Qnil;
    struct MEMO *memo;
    rb_block_call_func *func;

    if (argc == 0) {
        if (rb_block_given_p()) {
            func = count_iter_i;
        }
        else {
            func = count_all_i;
        }
    }
    else {
        rb_scan_args(argc, argv, "1", &item);
        if (rb_block_given_p()) {
            rb_warn("given block not used");
        }
        func = count_i;
    }

    memo = MEMO_NEW(item, 0, 0);
    rb_block_call(obj, id_each, 0, 0, func, (VALUE)memo);
    return imemo_count_value(memo);
}

Returns the count of elements, based on an argument or block criterion, if given.

With no argument and no block given, returns the number of elements:

[0, 1, 2].count                # => 3
{foo: 0, bar: 1, baz: 2}.count # => 3

With argument object given, returns the number of elements that are == to object:

[0, 1, 2, 1].count(1)           # => 2

With a block given, calls the block with each element and returns the number of elements for which the block returns a truthy value:

[0, 1, 2, 3].count {|element| element < 2}              # => 2
{foo: 0, bar: 1, baz: 2}.count {|key, value| value < 2} # => 2

static VALUE
enum_cycle(int argc, VALUE *argv, VALUE obj)
{
    VALUE ary;
    VALUE nv = Qnil;
    long n, i, len;

    rb_check_arity(argc, 0, 1);

    RETURN_SIZED_ENUMERATOR(obj, argc, argv, enum_cycle_size);
    if (!argc || NIL_P(nv = argv[0])) {
        n = -1;
    }
    else {
        n = NUM2LONG(nv);
        if (n <= 0) return Qnil;
    }
    ary = rb_ary_new();
    RBASIC_CLEAR_CLASS(ary);
    rb_block_call(obj, id_each, 0, 0, cycle_i, ary);
    len = RARRAY_LEN(ary);
    if (len == 0) return Qnil;
    while (n < 0 || 0 < --n) {
        for (i=0; i<len; i++) {
            enum_yield_array(RARRAY_AREF(ary, i));
        }
    }
    return Qnil;
}

When called with positive integer argument n and a block, calls the block with each element, then does so again, until it has done so n times; returns nil:

a = []
(1..4).cycle(3) {|element| a.push(element) } # => nil
a # => [1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4]
a = []
('a'..'d').cycle(2) {|element| a.push(element) }
a # => ["a", "b", "c", "d", "a", "b", "c", "d"]
a = []
{foo: 0, bar: 1, baz: 2}.cycle(2) {|element| a.push(element) }
a # => [[:foo, 0], [:bar, 1], [:baz, 2], [:foo, 0], [:bar, 1], [:baz, 2]]

If count is zero or negative, does not call the block.

When called with a block and n is nil, cycles forever.

When no block is given, returns an Enumerator.

static VALUE
enum_drop(VALUE obj, VALUE n)
{
    VALUE result;
    struct MEMO *memo;
    long len = NUM2LONG(n);

    if (len < 0) {
        rb_raise(rb_eArgError, "attempt to drop negative size");
    }

    result = rb_ary_new();
    memo = MEMO_NEW(result, 0, len);
    rb_block_call(obj, id_each, 0, 0, drop_i, (VALUE)memo);
    return result;
}

For positive integer n, returns an array containing all but the first n elements:

r = (1..4)
r.drop(3)  # => [4]
r.drop(2)  # => [3, 4]
r.drop(1)  # => [2, 3, 4]
r.drop(0)  # => [1, 2, 3, 4]
r.drop(50) # => []

h = {foo: 0, bar: 1, baz: 2, bat: 3}
h.drop(2) # => [[:baz, 2], [:bat, 3]]

static VALUE
enum_drop_while(VALUE obj)
{
    VALUE result;
    struct MEMO *memo;

    RETURN_ENUMERATOR(obj, 0, 0);
    result = rb_ary_new();
    memo = MEMO_NEW(result, 0, FALSE);
    rb_block_call(obj, id_each, 0, 0, drop_while_i, (VALUE)memo);
    return result;
}

Calls the block with successive elements as long as the block returns a truthy value; returns an array of all elements after that point:

(1..4).drop_while{|i| i < 3 } # => [3, 4]
h = {foo: 0, bar: 1, baz: 2}
a = h.drop_while{|element| key, value = *element; value < 2 }
a # => [[:baz, 2]]

With no block given, returns an Enumerator.

static VALUE
enum_each_cons(VALUE obj, VALUE n)
{
    long size = NUM2LONG(n);
    struct MEMO *memo;
    int arity;

    if (size <= 0) rb_raise(rb_eArgError, "invalid size");
    RETURN_SIZED_ENUMERATOR(obj, 1, &n, enum_each_cons_size);
    arity = rb_block_arity();
    if (enum_size_over_p(obj, size)) return obj;
    memo = MEMO_NEW(rb_ary_new2(size), dont_recycle_block_arg(arity), size);
    rb_block_call(obj, id_each, 0, 0, each_cons_i, (VALUE)memo);

    return obj;
}

Calls the block with each successive overlapped n-tuple of elements; returns self:

a = []
(1..5).each_cons(3) {|element| a.push(element) }
a # => [[1, 2, 3], [2, 3, 4], [3, 4, 5]]

a = []
h = {foo: 0,  bar: 1, baz: 2, bam: 3}
h.each_cons(2) {|element| a.push(element) }
a # => [[[:foo, 0], [:bar, 1]], [[:bar, 1], [:baz, 2]], [[:baz, 2], [:bam, 3]]]

With no block given, returns an Enumerator.

static VALUE
enum_each_entry(int argc, VALUE *argv, VALUE obj)
{
    RETURN_SIZED_ENUMERATOR(obj, argc, argv, enum_size);
    rb_block_call(obj, id_each, argc, argv, each_val_i, 0);
    return obj;
}

Calls the given block with each element, converting multiple values from yield to an array; returns self:

a = []
(1..4).each_entry {|element| a.push(element) } # => 1..4
a # => [1, 2, 3, 4]

a = []
h = {foo: 0, bar: 1, baz:2}
h.each_entry {|element| a.push(element) }
# => {:foo=>0, :bar=>1, :baz=>2}
a # => [[:foo, 0], [:bar, 1], [:baz, 2]]

class Foo
  include Enumerable
  def each
    yield 1
    yield 1, 2
    yield
  end
end
Foo.new.each_entry {|yielded| p yielded }

Output:

1
[1, 2]
nil

With no block given, returns an Enumerator.

static VALUE
enum_each_slice(VALUE obj, VALUE n)
{
    long size = NUM2LONG(n);
    VALUE ary;
    struct MEMO *memo;
    int arity;

    if (size <= 0) rb_raise(rb_eArgError, "invalid slice size");
    RETURN_SIZED_ENUMERATOR(obj, 1, &n, enum_each_slice_size);
    size = limit_by_enum_size(obj, size);
    ary = rb_ary_new2(size);
    arity = rb_block_arity();
    memo = MEMO_NEW(ary, dont_recycle_block_arg(arity), size);
    rb_block_call(obj, id_each, 0, 0, each_slice_i, (VALUE)memo);
    ary = memo->v1;
    if (RARRAY_LEN(ary) > 0) rb_yield(ary);

    return obj;
}

Calls the block with each successive disjoint n-tuple of elements; returns self:

a = []
(1..10).each_slice(3) {|tuple| a.push(tuple) }
a # => [[1, 2, 3], [4, 5, 6], [7, 8, 9], [10]]

a = []
h = {foo: 0, bar: 1, baz: 2, bat: 3, bam: 4}
h.each_slice(2) {|tuple| a.push(tuple) }
a # => [[[:foo, 0], [:bar, 1]], [[:baz, 2], [:bat, 3]], [[:bam, 4]]]

With no block given, returns an Enumerator.

static VALUE
enum_each_with_index(int argc, VALUE *argv, VALUE obj)
{
    RETURN_SIZED_ENUMERATOR(obj, argc, argv, enum_size);

    rb_block_call(obj, id_each, argc, argv, each_with_index_i, INT2FIX(0));
    return obj;
}

Invoke self.each with *args. With a block given, the block receives each element and its index; returns self:

h = {}
(1..4).each_with_index {|element, i| h[element] = i } # => 1..4
h # => {1=>0, 2=>1, 3=>2, 4=>3}

h = {}
%w[a b c d].each_with_index {|element, i| h[element] = i }
# => ["a", "b", "c", "d"]
h # => {"a"=>0, "b"=>1, "c"=>2, "d"=>3}

a = []
h = {foo: 0, bar: 1, baz: 2}
h.each_with_index {|element, i| a.push([i, element]) }
# => {:foo=>0, :bar=>1, :baz=>2}
a # => [[0, [:foo, 0]], [1, [:bar, 1]], [2, [:baz, 2]]]

With no block given, returns an Enumerator.

static VALUE
enum_each_with_object(VALUE obj, VALUE memo)
{
    RETURN_SIZED_ENUMERATOR(obj, 1, &memo, enum_size);

    rb_block_call(obj, id_each, 0, 0, each_with_object_i, memo);

    return memo;
}

Calls the block once for each element, passing both the element and the given object:

(1..4).each_with_object([]) {|i, a| a.push(i**2) }
# => [1, 4, 9, 16]

{foo: 0, bar: 1, baz: 2}.each_with_object({}) {|(k, v), h| h[v] = k }
# => {0=>:foo, 1=>:bar, 2=>:baz}

With no block given, returns an Enumerator.

static VALUE
enum_filter_map(VALUE obj)
{
    VALUE ary;

    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    ary = rb_ary_new();
    rb_block_call(obj, id_each, 0, 0, filter_map_i, ary);

    return ary;
}

Returns an array containing truthy elements returned by the block.

With a block given, calls the block with successive elements; returns an array containing each truthy value returned by the block:

(0..9).filter_map {|i| i * 2 if i.even? }                              # => [0, 4, 8, 12, 16]
{foo: 0, bar: 1, baz: 2}.filter_map {|key, value| key if value.even? } # => [:foo, :baz]

When no block given, returns an Enumerator.

static VALUE
enum_find(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo;
    VALUE if_none;

    if_none = rb_check_arity(argc, 0, 1) ? argv[0] : Qnil;
    RETURN_ENUMERATOR(obj, argc, argv);
    memo = MEMO_NEW(Qundef, 0, 0);
    if (rb_block_pair_yield_optimizable())
        rb_block_call2(obj, id_each, 0, 0, find_i_fast, (VALUE)memo, RB_BLOCK_NO_USE_PACKED_ARGS);
    else
        rb_block_call2(obj, id_each, 0, 0, find_i, (VALUE)memo, RB_BLOCK_NO_USE_PACKED_ARGS);
    if (memo->u3.cnt) {
        return memo->v1;
    }
    if (!NIL_P(if_none)) {
        return rb_funcallv(if_none, id_call, 0, 0);
    }
    return Qnil;
}

Returns the first element for which the block returns a truthy value.

With a block given, calls the block with successive elements of the collection; returns the first element for which the block returns a truthy value:

(0..9).find {|element| element > 2}                # => 3

If no such element is found, calls if_none_proc and returns its return value.

(0..9).find(proc {false}) {|element| element > 12} # => false
{foo: 0, bar: 1, baz: 2}.find {|key, value| key.start_with?('b') }            # => [:bar, 1]
{foo: 0, bar: 1, baz: 2}.find(proc {[]}) {|key, value| key.start_with?('c') } # => []

With no block given, returns an Enumerator.

static VALUE
enum_find_index(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo;  /* [return value, current index, ] */
    VALUE condition_value = Qnil;
    rb_block_call_func *func;

    if (argc == 0) {
        RETURN_ENUMERATOR(obj, 0, 0);
        func = find_index_iter_i;
    }
    else {
        rb_scan_args(argc, argv, "1", &condition_value);
        if (rb_block_given_p()) {
            rb_warn("given block not used");
        }
        func = find_index_i;
    }

    memo = MEMO_NEW(Qnil, condition_value, 0);
    rb_block_call(obj, id_each, 0, 0, func, (VALUE)memo);
    return memo->v1;
}

Returns the index of the first element that meets a specified criterion, or nil if no such element is found.

With argument object given, returns the index of the first element that is == object:

['a', 'b', 'c', 'b'].find_index('b') # => 1

With a block given, calls the block with successive elements; returns the first element for which the block returns a truthy value:

['a', 'b', 'c', 'b'].find_index {|element| element.start_with?('b') } # => 1
{foo: 0, bar: 1, baz: 2}.find_index {|key, value| value > 1 }         # => 2

With no argument and no block given, returns an Enumerator.

static VALUE
enum_first(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo;
    rb_check_arity(argc, 0, 1);
    if (argc > 0) {
        return enum_take(obj, argv[0]);
    }
    else {
        memo = MEMO_NEW(Qnil, 0, 0);
        rb_block_call(obj, id_each, 0, 0, first_i, (VALUE)memo);
        return memo->v1;
    }
}

Returns the first element or elements.

With no argument, returns the first element, or nil if there is none:

(1..4).first                   # => 1
%w[a b c].first                # => "a"
{foo: 1, bar: 1, baz: 2}.first # => [:foo, 1]
[].first                       # => nil

With integer argument n, returns an array containing the first n elements that exist:

(1..4).first(2)                   # => [1, 2]
%w[a b c d].first(3)              # => ["a", "b", "c"]
%w[a b c d].first(50)             # => ["a", "b", "c", "d"]
{foo: 1, bar: 1, baz: 2}.first(2) # => [[:foo, 1], [:bar, 1]]
[].first(2)                       # => []

static VALUE
enum_flat_map(VALUE obj)
{
    VALUE ary;

    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    ary = rb_ary_new();
    rb_block_call(obj, id_each, 0, 0, flat_map_i, ary);

    return ary;
}

Returns an array of flattened objects returned by the block.

With a block given, calls the block with successive elements; returns a flattened array of objects returned by the block:

[0, 1, 2, 3].flat_map {|element| -element }                    # => [0, -1, -2, -3]
[0, 1, 2, 3].flat_map {|element| [element, -element] }         # => [0, 0, 1, -1, 2, -2, 3, -3]
[[0, 1], [2, 3]].flat_map {|e| e + [100] }                     # => [0, 1, 100, 2, 3, 100]
{foo: 0, bar: 1, baz: 2}.flat_map {|key, value| [key, value] } # => [:foo, 0, :bar, 1, :baz, 2]

With no block given, returns an Enumerator.

Alias: collect_concat.

static VALUE
enum_grep(VALUE obj, VALUE pat)
{
    return enum_grep0(obj, pat, Qtrue);
}

Returns an array of objects based elements of self that match the given pattern.

With no block given, returns an array containing each element for which pattern === element is true:

a = ['foo', 'bar', 'car', 'moo']
a.grep(/ar/)                   # => ["bar", "car"]
(1..10).grep(3..8)             # => [3, 4, 5, 6, 7, 8]
['a', 'b', 0, 1].grep(Integer) # => [0, 1]

With a block given, calls the block with each matching element and returns an array containing each object returned by the block:

a = ['foo', 'bar', 'car', 'moo']
a.grep(/ar/) {|element| element.upcase } # => ["BAR", "CAR"]

Related: grep_v.

static VALUE
enum_grep_v(VALUE obj, VALUE pat)
{
    return enum_grep0(obj, pat, Qfalse);
}

Returns an array of objects based on elements of self that don’t match the given pattern.

With no block given, returns an array containing each element for which pattern === element is false:

a = ['foo', 'bar', 'car', 'moo']
a.grep_v(/ar/)                   # => ["foo", "moo"]
(1..10).grep_v(3..8)             # => [1, 2, 9, 10]
['a', 'b', 0, 1].grep_v(Integer) # => ["a", "b"]

With a block given, calls the block with each non-matching element and returns an array containing each object returned by the block:

a = ['foo', 'bar', 'car', 'moo']
a.grep_v(/ar/) {|element| element.upcase } # => ["FOO", "MOO"]

Related: grep.

static VALUE
enum_group_by(VALUE obj)
{
    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    return enum_hashify(obj, 0, 0, group_by_i);
}

With a block given returns a hash:

• Each key is a return value from the block.

• Each value is an array of those elements for which the block returned that key.

Examples:

g = (1..6).group_by {|i| i%3 }
g # => {1=>[1, 4], 2=>[2, 5], 0=>[3, 6]}
h = {foo: 0, bar: 1, baz: 0, bat: 1}
g = h.group_by {|key, value| value }
g # => {0=>[[:foo, 0], [:baz, 0]], 1=>[[:bar, 1], [:bat, 1]]}

With no block given, returns an Enumerator.

static VALUE
enum_inject(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo;
    VALUE init, op;
    rb_block_call_func *iter = inject_i;
    ID id;
    int num_args;

    if (rb_block_given_p()) {
        num_args = rb_scan_args(argc, argv, "02", &init, &op);
    }
    else {
        num_args = rb_scan_args(argc, argv, "11", &init, &op);
    }

    switch (num_args) {
      case 0:
        init = Qundef;
        break;
      case 1:
        if (rb_block_given_p()) {
            break;
        }
        id = rb_check_id(&init);
        op = id ? ID2SYM(id) : init;
        init = Qundef;
        iter = inject_op_i;
        break;
      case 2:
        if (rb_block_given_p()) {
            rb_warning("given block not used");
        }
        id = rb_check_id(&op);
        if (id) op = ID2SYM(id);
        iter = inject_op_i;
        break;
    }

    if (iter == inject_op_i &&
        SYMBOL_P(op) &&
        RB_TYPE_P(obj, T_ARRAY) &&
        rb_method_basic_definition_p(CLASS_OF(obj), id_each)) {
        return ary_inject_op(obj, init, op);
    }

    memo = MEMO_NEW(init, Qnil, op);
    rb_block_call(obj, id_each, 0, 0, iter, (VALUE)memo);
    if (UNDEF_P(memo->v1)) return Qnil;
    return memo->v1;
}

Returns the result of applying a reducer to an initial value and the first element of the Enumerable. It then takes the result and applies the function to it and the second element of the collection, and so on. The return value is the result returned by the final call to the function.

You can think of

[ a, b, c, d ].inject(i) { |r, v| fn(r, v) }

as being

fn(fn(fn(fn(i, a), b), c), d)

In a way the inject function injects the function between the elements of the enumerable.

inject is aliased as reduce. You use it when you want to reduce a collection to a single value.

The Calling Sequences

Let’s start with the most verbose:

enum.inject(initial_value) do |result, next_value|
  # do something with +result+ and +next_value+
  # the value returned by the block becomes the
  # value passed in to the next iteration
  # as +result+
end

For example:

product = [ 2, 3, 4 ].inject(1) do |result, next_value|
  result * next_value
end
product #=> 24

When this runs, the block is first called with 1 (the initial value) and 2 (the first element of the array). The block returns 1*2, so on the next iteration the block is called with 2 (the previous result) and 3. The block returns 6, and is called one last time with 6 and 4. The result of the block, 24 becomes the value returned by inject. This code returns the product of the elements in the enumerable.

First Shortcut: Default Initial value

In the case of the previous example, the initial value, 1, wasn’t really necessary: the calculation of the product of a list of numbers is self-contained.

In these circumstances, you can omit the initial_value parameter. inject will then initially call the block with the first element of the collection as the result parameter and the second element as the next_value.

[ 2, 3, 4 ].inject do |result, next_value|
  result * next_value
end

This shortcut is convenient, but can only be used when the block produces a result which can be passed back to it as a first parameter.

Here’s an example where that’s not the case: it returns a hash where the keys are words and the values are the number of occurrences of that word in the enumerable.

freqs = File.read("README.md")
  .scan(/\w{2,}/)
  .reduce(Hash.new(0)) do |counts, word|
    counts[word] += 1
    counts
  end
freqs #=> {"Actions"=>4,
           "Status"=>5,
           "MinGW"=>3,
           "https"=>27,
           "github"=>10,
           "com"=>15, ...

Note that the last line of the block is just the word counts. This ensures the return value of the block is the result that’s being calculated.

Second Shortcut: a Reducer function

A reducer function is a function that takes a partial result and the next value, returning the next partial result. The block that is given to inject is a reducer.

You can also write a reducer as a function and pass the name of that function (as a symbol) to inject. However, for this to work, the function

1. Must be defined on the type of the result value

2. Must accept a single parameter, the next value in the collection, and

3. Must return an updated result which will also implement the function.

Here’s an example that adds elements to a string. The two calls invoke the functions String#concat and String#+ on the result so far, passing it the next value.

s = [ "cat", " ", "dog" ].inject("", :concat)
s #=> "cat dog"
s = [ "cat", " ", "dog" ].inject("The result is:", :+)
s #=> "The result is: cat dog"

Here’s a more complex example when the result object maintains state of a different type to the enumerable elements.

class Turtle

  def initialize
    @x = @y = 0
  end

  def move(dir)
    case dir
    when "n" then @y += 1
    when "s" then @y -= 1
    when "e" then @x += 1
    when "w" then @x -= 1
    end
    self
  end
end

position = "nnneesw".chars.reduce(Turtle.new, :move)
position  #=>> #<Turtle:0x00000001052f4698 @y=2, @x=1>

Third Shortcut: Reducer With no Initial Value

If your reducer returns a value that it can accept as a parameter, then you don’t have to pass in an initial value. Here :* is the name of the times function:

product = [ 2, 3, 4 ].inject(:*)
product # => 24

String concatenation again:

s = [ "cat", " ", "dog" ].inject(:+)
s #=> "cat dog"

And an example that converts a hash to an array of two-element subarrays.

nested = {foo: 0, bar: 1}.inject([], :push)
nested # => [[:foo, 0], [:bar, 1]]

static VALUE
enumerable_lazy(VALUE obj)
{
    VALUE result = lazy_to_enum_i(obj, sym_each, 0, 0, lazyenum_size, rb_keyword_given_p());
    /* Qfalse indicates that the Enumerator::Lazy has no method name */
    rb_ivar_set(result, id_method, Qfalse);
    return result;
}

Returns an Enumerator::Lazy, which redefines most Enumerable methods to postpone enumeration and enumerate values only on an as-needed basis.

Example¶ ↑

The following program finds pythagorean triples:

def pythagorean_triples
  (1..Float::INFINITY).lazy.flat_map {|z|
    (1..z).flat_map {|x|
      (x..z).select {|y|
        x**2 + y**2 == z**2
      }.map {|y|
        [x, y, z]
      }
    }
  }
end
# show first ten pythagorean triples
p pythagorean_triples.take(10).force # take is lazy, so force is needed
p pythagorean_triples.first(10)      # first is eager
# show pythagorean triples less than 100
p pythagorean_triples.take_while { |*, z| z < 100 }.force

static VALUE
enum_max(int argc, VALUE *argv, VALUE obj)
{
    VALUE memo;
    struct max_t *m = NEW_MEMO_FOR(struct max_t, memo);
    VALUE result;
    VALUE num;

    if (rb_check_arity(argc, 0, 1) && !NIL_P(num = argv[0]))
       return rb_nmin_run(obj, num, 0, 1, 0);

    m->max = Qundef;
    if (rb_block_given_p()) {
        rb_block_call(obj, id_each, 0, 0, max_ii, (VALUE)memo);
    }
    else {
        rb_block_call(obj, id_each, 0, 0, max_i, (VALUE)memo);
    }
    result = m->max;
    if (UNDEF_P(result)) return Qnil;
    return result;
}

Returns the element with the maximum element according to a given criterion. The ordering of equal elements is indeterminate and may be unstable.

With no argument and no block, returns the maximum element, using the elements’ own method #<=> for comparison:

(1..4).max                   # => 4
(-4..-1).max                 # => -1
%w[d c b a].max              # => "d"
{foo: 0, bar: 1, baz: 2}.max # => [:foo, 0]
[].max                       # => nil

With positive integer argument n given, and no block, returns an array containing the first n maximum elements that exist:

(1..4).max(2)                   # => [4, 3]
(-4..-1).max(2)                # => [-1, -2]
%w[d c b a].max(2)              # => ["d", "c"]
{foo: 0, bar: 1, baz: 2}.max(2) # => [[:foo, 0], [:baz, 2]]
[].max(2)                       # => []

With a block given, the block determines the maximum elements. The block is called with two elements a and b, and must return:

• A negative integer if a < b.

• Zero if a == b.

• A positive integer if a > b.

With a block given and no argument, returns the maximum element as determined by the block:

%w[xxx x xxxx xx].max {|a, b| a.size <=> b.size } # => "xxxx"
h = {foo: 0, bar: 1, baz: 2}
h.max {|pair1, pair2| pair1[1] <=> pair2[1] }     # => [:baz, 2]
[].max {|a, b| a <=> b }                          # => nil

With a block given and positive integer argument n given, returns an array containing the first n maximum elements that exist, as determined by the block.

%w[xxx x xxxx xx].max(2) {|a, b| a.size <=> b.size } # => ["xxxx", "xxx"]
h = {foo: 0, bar: 1, baz: 2}
h.max(2) {|pair1, pair2| pair1[1] <=> pair2[1] }
# => [[:baz, 2], [:bar, 1]]
[].max(2) {|a, b| a <=> b }                          # => []

Related: min, minmax, max_by.

static VALUE
enum_max_by(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo;
    VALUE num;

    rb_check_arity(argc, 0, 1);

    RETURN_SIZED_ENUMERATOR(obj, argc, argv, enum_size);

    if (argc && !NIL_P(num = argv[0]))
        return rb_nmin_run(obj, num, 1, 1, 0);

    memo = MEMO_NEW(Qundef, Qnil, 0);
    rb_block_call(obj, id_each, 0, 0, max_by_i, (VALUE)memo);
    return memo->v2;
}

Returns the elements for which the block returns the maximum values.

With a block given and no argument, returns the element for which the block returns the maximum value:

(1..4).max_by {|element| -element }                    # => 1
%w[a b c d].max_by {|element| -element.ord }           # => "a"
{foo: 0, bar: 1, baz: 2}.max_by {|key, value| -value } # => [:foo, 0]
[].max_by {|element| -element }                        # => nil

With a block given and positive integer argument n given, returns an array containing the n elements for which the block returns maximum values:

(1..4).max_by(2) {|element| -element }
# => [1, 2]
%w[a b c d].max_by(2) {|element| -element.ord }
# => ["a", "b"]
{foo: 0, bar: 1, baz: 2}.max_by(2) {|key, value| -value }
# => [[:foo, 0], [:bar, 1]]
[].max_by(2) {|element| -element }
# => []

Returns an Enumerator if no block is given.

Related: max, minmax, min_by.

static VALUE
enum_min(int argc, VALUE *argv, VALUE obj)
{
    VALUE memo;
    struct min_t *m = NEW_MEMO_FOR(struct min_t, memo);
    VALUE result;
    VALUE num;

    if (rb_check_arity(argc, 0, 1) && !NIL_P(num = argv[0]))
       return rb_nmin_run(obj, num, 0, 0, 0);

    m->min = Qundef;
    if (rb_block_given_p()) {
        rb_block_call(obj, id_each, 0, 0, min_ii, memo);
    }
    else {
        rb_block_call(obj, id_each, 0, 0, min_i, memo);
    }
    result = m->min;
    if (UNDEF_P(result)) return Qnil;
    return result;
}

Returns the element with the minimum element according to a given criterion. The ordering of equal elements is indeterminate and may be unstable.

With no argument and no block, returns the minimum element, using the elements’ own method #<=> for comparison:

(1..4).min                   # => 1
(-4..-1).min                 # => -4
%w[d c b a].min              # => "a"
{foo: 0, bar: 1, baz: 2}.min # => [:bar, 1]
[].min                       # => nil

With positive integer argument n given, and no block, returns an array containing the first n minimum elements that exist:

(1..4).min(2)                   # => [1, 2]
(-4..-1).min(2)                 # => [-4, -3]
%w[d c b a].min(2)              # => ["a", "b"]
{foo: 0, bar: 1, baz: 2}.min(2) # => [[:bar, 1], [:baz, 2]]
[].min(2)                       # => []

With a block given, the block determines the minimum elements. The block is called with two elements a and b, and must return:

• A negative integer if a < b.

• Zero if a == b.

• A positive integer if a > b.

With a block given and no argument, returns the minimum element as determined by the block:

%w[xxx x xxxx xx].min {|a, b| a.size <=> b.size } # => "x"
h = {foo: 0, bar: 1, baz: 2}
h.min {|pair1, pair2| pair1[1] <=> pair2[1] } # => [:foo, 0]
[].min {|a, b| a <=> b }                          # => nil

With a block given and positive integer argument n given, returns an array containing the first n minimum elements that exist, as determined by the block.

%w[xxx x xxxx xx].min(2) {|a, b| a.size <=> b.size } # => ["x", "xx"]
h = {foo: 0, bar: 1, baz: 2}
h.min(2) {|pair1, pair2| pair1[1] <=> pair2[1] }
# => [[:foo, 0], [:bar, 1]]
[].min(2) {|a, b| a <=> b }                          # => []

Related: min_by, minmax, max.

static VALUE
enum_min_by(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo;
    VALUE num;

    rb_check_arity(argc, 0, 1);

    RETURN_SIZED_ENUMERATOR(obj, argc, argv, enum_size);

    if (argc && !NIL_P(num = argv[0]))
        return rb_nmin_run(obj, num, 1, 0, 0);

    memo = MEMO_NEW(Qundef, Qnil, 0);
    rb_block_call(obj, id_each, 0, 0, min_by_i, (VALUE)memo);
    return memo->v2;
}

Returns the elements for which the block returns the minimum values.

With a block given and no argument, returns the element for which the block returns the minimum value:

(1..4).min_by {|element| -element }                    # => 4
%w[a b c d].min_by {|element| -element.ord }           # => "d"
{foo: 0, bar: 1, baz: 2}.min_by {|key, value| -value } # => [:baz, 2]
[].min_by {|element| -element }                        # => nil

With a block given and positive integer argument n given, returns an array containing the n elements for which the block returns minimum values:

(1..4).min_by(2) {|element| -element }
# => [4, 3]
%w[a b c d].min_by(2) {|element| -element.ord }
# => ["d", "c"]
{foo: 0, bar: 1, baz: 2}.min_by(2) {|key, value| -value }
# => [[:baz, 2], [:bar, 1]]
[].min_by(2) {|element| -element }
# => []

Returns an Enumerator if no block is given.

Related: min, minmax, max_by.

static VALUE
enum_minmax(VALUE obj)
{
    VALUE memo;
    struct minmax_t *m = NEW_MEMO_FOR(struct minmax_t, memo);

    m->min = Qundef;
    m->last = Qundef;
    if (rb_block_given_p()) {
        rb_block_call(obj, id_each, 0, 0, minmax_ii, memo);
        if (!UNDEF_P(m->last))
            minmax_ii_update(m->last, m->last, m);
    }
    else {
        rb_block_call(obj, id_each, 0, 0, minmax_i, memo);
        if (!UNDEF_P(m->last))
            minmax_i_update(m->last, m->last, m);
    }
    if (!UNDEF_P(m->min)) {
        return rb_assoc_new(m->min, m->max);
    }
    return rb_assoc_new(Qnil, Qnil);
}

Returns a 2-element array containing the minimum and maximum elements according to a given criterion. The ordering of equal elements is indeterminate and may be unstable.

With no argument and no block, returns the minimum and maximum elements, using the elements’ own method #<=> for comparison:

(1..4).minmax                   # => [1, 4]
(-4..-1).minmax                 # => [-4, -1]
%w[d c b a].minmax              # => ["a", "d"]
{foo: 0, bar: 1, baz: 2}.minmax # => [[:bar, 1], [:foo, 0]]
[].minmax                       # => [nil, nil]

With a block given, returns the minimum and maximum elements as determined by the block:

%w[xxx x xxxx xx].minmax {|a, b| a.size <=> b.size } # => ["x", "xxxx"]
h = {foo: 0, bar: 1, baz: 2}
h.minmax {|pair1, pair2| pair1[1] <=> pair2[1] }
# => [[:foo, 0], [:baz, 2]]
[].minmax {|a, b| a <=> b }                          # => [nil, nil]

Related: min, max, minmax_by.

static VALUE
enum_minmax_by(VALUE obj)
{
    VALUE memo;
    struct minmax_by_t *m = NEW_MEMO_FOR(struct minmax_by_t, memo);

    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    m->min_bv = Qundef;
    m->max_bv = Qundef;
    m->min = Qnil;
    m->max = Qnil;
    m->last_bv = Qundef;
    m->last = Qundef;
    rb_block_call(obj, id_each, 0, 0, minmax_by_i, memo);
    if (!UNDEF_P(m->last_bv))
        minmax_by_i_update(m->last_bv, m->last_bv, m->last, m->last, m);
    m = MEMO_FOR(struct minmax_by_t, memo);
    return rb_assoc_new(m->min, m->max);
}

Returns a 2-element array containing the elements for which the block returns minimum and maximum values:

(1..4).minmax_by {|element| -element }
# => [4, 1]
%w[a b c d].minmax_by {|element| -element.ord }
# => ["d", "a"]
{foo: 0, bar: 1, baz: 2}.minmax_by {|key, value| -value }
# => [[:baz, 2], [:foo, 0]]
[].minmax_by {|element| -element }
# => [nil, nil]

Returns an Enumerator if no block is given.

Related: max_by, minmax, min_by.

static VALUE
enum_none(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo = MEMO_ENUM_NEW(Qtrue);

    WARN_UNUSED_BLOCK(argc);
    ENUM_BLOCK_CALL(none);
    return memo->v1;
}

Returns whether no element meets a given criterion.

With no argument and no block, returns whether no element is truthy:

(1..4).none?           # => false
[nil, false].none?     # => true
{foo: 0}.none?         # => false
{foo: 0, bar: 1}.none? # => false
[].none?               # => true

With argument pattern and no block, returns whether for no element element, pattern === element:

[nil, false, 1.1].none?(Integer)      # => true
%w[bar baz bat bam].none?(/m/)        # => false
%w[bar baz bat bam].none?(/foo/)      # => true
%w[bar baz bat bam].none?('ba')       # => true
{foo: 0, bar: 1, baz: 2}.none?(Hash)  # => true
{foo: 0}.none?(Array)                 # => false
[].none?(Integer)                     # => true

With a block given, returns whether the block returns a truthy value for no element:

(1..4).none? {|element| element < 1 }                     # => true
(1..4).none? {|element| element < 2 }                     # => false
{foo: 0, bar: 1, baz: 2}.none? {|key, value| value < 0 }  # => true
{foo: 0, bar: 1, baz: 2}.none? {|key, value| value < 1 } # => false

Related: one?, all?, any?.

static VALUE
enum_one(int argc, VALUE *argv, VALUE obj)
{
    struct MEMO *memo = MEMO_ENUM_NEW(Qundef);
    VALUE result;

    WARN_UNUSED_BLOCK(argc);
    ENUM_BLOCK_CALL(one);
    result = memo->v1;
    if (UNDEF_P(result)) return Qfalse;
    return result;
}

Returns whether exactly one element meets a given criterion.

With no argument and no block, returns whether exactly one element is truthy:

(1..1).one?           # => true
[1, nil, false].one?  # => true
(1..4).one?           # => false
{foo: 0}.one?         # => true
{foo: 0, bar: 1}.one? # => false
[].one?               # => false

With argument pattern and no block, returns whether for exactly one element element, pattern === element:

[nil, false, 0].one?(Integer)        # => true
[nil, false, 0].one?(Numeric)        # => true
[nil, false, 0].one?(Float)          # => false
%w[bar baz bat bam].one?(/m/)        # => true
%w[bar baz bat bam].one?(/foo/)      # => false
%w[bar baz bat bam].one?('ba')       # => false
{foo: 0, bar: 1, baz: 2}.one?(Array) # => false
{foo: 0}.one?(Array)                 # => true
[].one?(Integer)                     # => false

With a block given, returns whether the block returns a truthy value for exactly one element:

(1..4).one? {|element| element < 2 }                     # => true
(1..4).one? {|element| element < 1 }                     # => false
{foo: 0, bar: 1, baz: 2}.one? {|key, value| value < 1 }  # => true
{foo: 0, bar: 1, baz: 2}.one? {|key, value| value < 2 } # => false

Related: none?, all?, any?.

static VALUE
enum_partition(VALUE obj)
{
    struct MEMO *memo;

    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    memo = MEMO_NEW(rb_ary_new(), rb_ary_new(), 0);
    rb_block_call(obj, id_each, 0, 0, partition_i, (VALUE)memo);

    return rb_assoc_new(memo->v1, memo->v2);
}

With a block given, returns an array of two arrays:

• The first having those elements for which the block returns a truthy value.

• The other having all other elements.

Examples:

p = (1..4).partition {|i| i.even? }
p # => [[2, 4], [1, 3]]
p = ('a'..'d').partition {|c| c < 'c' }
p # => [["a", "b"], ["c", "d"]]
h = {foo: 0, bar: 1, baz: 2, bat: 3}
p = h.partition {|key, value| key.start_with?('b') }
p # => [[[:bar, 1], [:baz, 2], [:bat, 3]], [[:foo, 0]]]
p = h.partition {|key, value| value < 2 }
p # => [[[:foo, 0], [:bar, 1]], [[:baz, 2], [:bat, 3]]]

With no block given, returns an Enumerator.

Related: Enumerable#group_by.

static VALUE
enum_reject(VALUE obj)
{
    VALUE ary;

    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    ary = rb_ary_new();
    rb_block_call(obj, id_each, 0, 0, reject_i, ary);

    return ary;
}

Returns an array of objects rejected by the block.

With a block given, calls the block with successive elements; returns an array of those elements for which the block returns nil or false:

(0..9).reject {|i| i * 2 if i.even? }                             # => [1, 3, 5, 7, 9]
{foo: 0, bar: 1, baz: 2}.reject {|key, value| key if value.odd? } # => {:foo=>0, :baz=>2}

When no block given, returns an Enumerator.

Related: select.

static VALUE
enum_reverse_each(int argc, VALUE *argv, VALUE obj)
{
    VALUE ary;
    long len;

    RETURN_SIZED_ENUMERATOR(obj, argc, argv, enum_size);

    ary = enum_to_a(argc, argv, obj);

    len = RARRAY_LEN(ary);
    while (len--) {
        long nlen;
        rb_yield(RARRAY_AREF(ary, len));
        nlen = RARRAY_LEN(ary);
        if (nlen < len) {
            len = nlen;
        }
    }

    return obj;
}

With a block given, calls the block with each element, but in reverse order; returns self:

a = []
(1..4).reverse_each {|element| a.push(-element) } # => 1..4
a # => [-4, -3, -2, -1]

a = []
%w[a b c d].reverse_each {|element| a.push(element) }
# => ["a", "b", "c", "d"]
a # => ["d", "c", "b", "a"]

a = []
h.reverse_each {|element| a.push(element) }
# => {:foo=>0, :bar=>1, :baz=>2}
a # => [[:baz, 2], [:bar, 1], [:foo, 0]]

With no block given, returns an Enumerator.

static VALUE
enum_slice_after(int argc, VALUE *argv, VALUE enumerable)
{
    VALUE enumerator;
    VALUE pat = Qnil, pred = Qnil;

    if (rb_block_given_p()) {
        if (0 < argc)
            rb_raise(rb_eArgError, "both pattern and block are given");
        pred = rb_block_proc();
    }
    else {
        rb_scan_args(argc, argv, "1", &pat);
    }

    enumerator = rb_obj_alloc(rb_cEnumerator);
    rb_ivar_set(enumerator, id_sliceafter_enum, enumerable);
    rb_ivar_set(enumerator, id_sliceafter_pat, pat);
    rb_ivar_set(enumerator, id_sliceafter_pred, pred);

    rb_block_call(enumerator, idInitialize, 0, 0, sliceafter_i, enumerator);
    return enumerator;
}

Creates an enumerator for each chunked elements. The ends of chunks are defined by pattern and the block.

If pattern === elt returns true or the block returns true for the element, the element is end of a chunk.

The === and block is called from the first element to the last element of enum.

The result enumerator yields the chunked elements as an array. So each method can be called as follows:

enum.slice_after(pattern).each { |ary| ... }
enum.slice_after { |elt| bool }.each { |ary| ... }

Other methods of the Enumerator class and Enumerable module, such as map, etc., are also usable.

For example, continuation lines (lines end with backslash) can be concatenated as follows:

lines = ["foo\n", "bar\\\n", "baz\n", "\n", "qux\n"]
e = lines.slice_after(/(?<!\\)\n\z/)
p e.to_a
#=> [["foo\n"], ["bar\\\n", "baz\n"], ["\n"], ["qux\n"]]
p e.map {|ll| ll[0...-1].map {|l| l.sub(/\\\n\z/, "") }.join + ll.last }
#=>["foo\n", "barbaz\n", "\n", "qux\n"]

static VALUE
enum_slice_before(int argc, VALUE *argv, VALUE enumerable)
{
    VALUE enumerator;

    if (rb_block_given_p()) {
        if (argc != 0)
            rb_error_arity(argc, 0, 0);
        enumerator = rb_obj_alloc(rb_cEnumerator);
        rb_ivar_set(enumerator, id_slicebefore_sep_pred, rb_block_proc());
    }
    else {
        VALUE sep_pat;
        rb_scan_args(argc, argv, "1", &sep_pat);
        enumerator = rb_obj_alloc(rb_cEnumerator);
        rb_ivar_set(enumerator, id_slicebefore_sep_pat, sep_pat);
    }
    rb_ivar_set(enumerator, id_slicebefore_enumerable, enumerable);
    rb_block_call(enumerator, idInitialize, 0, 0, slicebefore_i, enumerator);
    return enumerator;
}

With argument pattern, returns an enumerator that uses the pattern to partition elements into arrays (“slices”). An element begins a new slice if element === pattern (or if it is the first element).

a = %w[foo bar fop for baz fob fog bam foy]
e = a.slice_before(/ba/) # => #<Enumerator: ...>
e.each {|array| p array }

Output:

["foo"]
["bar", "fop", "for"]
["baz", "fob", "fog"]
["bam", "foy"]

With a block, returns an enumerator that uses the block to partition elements into arrays. An element begins a new slice if its block return is a truthy value (or if it is the first element):

e = (1..20).slice_before {|i| i % 4 == 2 } # => #<Enumerator: ...>
e.each {|array| p array }

Output:

[1]
[2, 3, 4, 5]
[6, 7, 8, 9]
[10, 11, 12, 13]
[14, 15, 16, 17]
[18, 19, 20]

Other methods of the Enumerator class and Enumerable module, such as to_a, map, etc., are also usable.

For example, iteration over ChangeLog entries can be implemented as follows:

# iterate over ChangeLog entries.
open("ChangeLog") { |f|
  f.slice_before(/\A\S/).each { |e| pp e }
}

# same as above.  block is used instead of pattern argument.
open("ChangeLog") { |f|
  f.slice_before { |line| /\A\S/ === line }.each { |e| pp e }
}

“svn proplist -R” produces multiline output for each file. They can be chunked as follows:

IO.popen([{"LC_ALL"=>"C"}, "svn", "proplist", "-R"]) { |f|
  f.lines.slice_before(/\AProp/).each { |lines| p lines }
}
#=> ["Properties on '.':\n", "  svn:ignore\n", "  svk:merge\n"]
#   ["Properties on 'goruby.c':\n", "  svn:eol-style\n"]
#   ["Properties on 'complex.c':\n", "  svn:mime-type\n", "  svn:eol-style\n"]
#   ["Properties on 'regparse.c':\n", "  svn:eol-style\n"]
#   ...

If the block needs to maintain state over multiple elements, local variables can be used. For example, three or more consecutive increasing numbers can be squashed as follows (see chunk_while for a better way):

a = [0, 2, 3, 4, 6, 7, 9]
prev = a[0]
p a.slice_before { |e|
  prev, prev2 = e, prev
  prev2 + 1 != e
}.map { |es|
  es.length <= 2 ? es.join(",") : "#{es.first}-#{es.last}"
}.join(",")
#=> "0,2-4,6,7,9"

However local variables should be used carefully if the result enumerator is enumerated twice or more. The local variables should be initialized for each enumeration. Enumerator.new can be used to do it.

# Word wrapping.  This assumes all characters have same width.
def wordwrap(words, maxwidth)
  Enumerator.new {|y|
    # cols is initialized in Enumerator.new.
    cols = 0
    words.slice_before { |w|
      cols += 1 if cols != 0
      cols += w.length
      if maxwidth < cols
        cols = w.length
        true
      else
        false
      end
    }.each {|ws| y.yield ws }
  }
end
text = (1..20).to_a.join(" ")
enum = wordwrap(text.split(/\s+/), 10)
puts "-"*10
enum.each { |ws| puts ws.join(" ") } # first enumeration.
puts "-"*10
enum.each { |ws| puts ws.join(" ") } # second enumeration generates same result as the first.
puts "-"*10
#=> ----------
#   1 2 3 4 5
#   6 7 8 9 10
#   11 12 13
#   14 15 16
#   17 18 19
#   20
#   ----------
#   1 2 3 4 5
#   6 7 8 9 10
#   11 12 13
#   14 15 16
#   17 18 19
#   20
#   ----------

mbox contains series of mails which start with Unix From line. So each mail can be extracted by slice before Unix From line.

# parse mbox
open("mbox") { |f|
  f.slice_before { |line|
    line.start_with? "From "
  }.each { |mail|
    unix_from = mail.shift
    i = mail.index("\n")
    header = mail[0...i]
    body = mail[(i+1)..-1]
    body.pop if body.last == "\n"
    fields = header.slice_before { |line| !" \t".include?(line[0]) }.to_a
    p unix_from
    pp fields
    pp body
  }
}

# split mails in mbox (slice before Unix From line after an empty line)
open("mbox") { |f|
  emp = true
  f.slice_before { |line|
    prevemp = emp
    emp = line == "\n"
    prevemp && line.start_with?("From ")
  }.each { |mail|
    mail.pop if mail.last == "\n"
    pp mail
  }
}

static VALUE
enum_slice_when(VALUE enumerable)
{
    VALUE enumerator;
    VALUE pred;

    pred = rb_block_proc();

    enumerator = rb_obj_alloc(rb_cEnumerator);
    rb_ivar_set(enumerator, id_slicewhen_enum, enumerable);
    rb_ivar_set(enumerator, id_slicewhen_pred, pred);
    rb_ivar_set(enumerator, id_slicewhen_inverted, Qfalse);

    rb_block_call(enumerator, idInitialize, 0, 0, slicewhen_i, enumerator);
    return enumerator;
}

Creates an enumerator for each chunked elements. The beginnings of chunks are defined by the block.

This method splits each chunk using adjacent elements, elt_before and elt_after, in the receiver enumerator. This method split chunks between elt_before and elt_after where the block returns true.

The block is called the length of the receiver enumerator minus one.

The result enumerator yields the chunked elements as an array. So each method can be called as follows:

enum.slice_when { |elt_before, elt_after| bool }.each { |ary| ... }

Other methods of the Enumerator class and Enumerable module, such as to_a, map, etc., are also usable.

For example, one-by-one increasing subsequence can be chunked as follows:

a = [1,2,4,9,10,11,12,15,16,19,20,21]
b = a.slice_when {|i, j| i+1 != j }
p b.to_a #=> [[1, 2], [4], [9, 10, 11, 12], [15, 16], [19, 20, 21]]
c = b.map {|a| a.length < 3 ? a : "#{a.first}-#{a.last}" }
p c #=> [[1, 2], [4], "9-12", [15, 16], "19-21"]
d = c.join(",")
p d #=> "1,2,4,9-12,15,16,19-21"

Near elements (threshold: 6) in sorted array can be chunked as follows:

a = [3, 11, 14, 25, 28, 29, 29, 41, 55, 57]
p a.slice_when {|i, j| 6 < j - i }.to_a
#=> [[3], [11, 14], [25, 28, 29, 29], [41], [55, 57]]

Increasing (non-decreasing) subsequence can be chunked as follows:

a = [0, 9, 2, 2, 3, 2, 7, 5, 9, 5]
p a.slice_when {|i, j| i > j }.to_a
#=> [[0, 9], [2, 2, 3], [2, 7], [5, 9], [5]]

Adjacent evens and odds can be chunked as follows: (Enumerable#chunk is another way to do it.)

a = [7, 5, 9, 2, 0, 7, 9, 4, 2, 0]
p a.slice_when {|i, j| i.even? != j.even? }.to_a
#=> [[7, 5, 9], [2, 0], [7, 9], [4, 2, 0]]

Paragraphs (non-empty lines with trailing empty lines) can be chunked as follows: (See Enumerable#chunk to ignore empty lines.)

lines = ["foo\n", "bar\n", "\n", "baz\n", "qux\n"]
p lines.slice_when {|l1, l2| /\A\s*\z/ =~ l1 && /\S/ =~ l2 }.to_a
#=> [["foo\n", "bar\n", "\n"], ["baz\n", "qux\n"]]

Enumerable#chunk_while does the same, except splitting when the block returns false instead of true.

static VALUE
enum_sort(VALUE obj)
{
    return rb_ary_sort_bang(enum_to_a(0, 0, obj));
}

Returns an array containing the sorted elements of self. The ordering of equal elements is indeterminate and may be unstable.

With no block given, the sort compares using the elements’ own method #<=>:

%w[b c a d].sort              # => ["a", "b", "c", "d"]
{foo: 0, bar: 1, baz: 2}.sort # => [[:bar, 1], [:baz, 2], [:foo, 0]]

With a block given, comparisons in the block determine the ordering. The block is called with two elements a and b, and must return:

• A negative integer if a < b.

• Zero if a == b.

• A positive integer if a > b.

Examples:

a = %w[b c a d]
a.sort {|a, b| b <=> a } # => ["d", "c", "b", "a"]
h = {foo: 0, bar: 1, baz: 2}
h.sort {|a, b| b <=> a } # => [[:foo, 0], [:baz, 2], [:bar, 1]]

See also sort_by. It implements a Schwartzian transform which is useful when key computation or comparison is expensive.

static VALUE
enum_sort_by(VALUE obj)
{
    VALUE ary, buf;
    struct MEMO *memo;
    long i;
    struct sort_by_data *data;

    RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);

    if (RB_TYPE_P(obj, T_ARRAY) && RARRAY_LEN(obj) <= LONG_MAX/2) {
        ary = rb_ary_new2(RARRAY_LEN(obj)*2);
    }
    else {
        ary = rb_ary_new();
    }
    RBASIC_CLEAR_CLASS(ary);
    buf = rb_ary_hidden_new(SORT_BY_BUFSIZE*2);
    rb_ary_store(buf, SORT_BY_BUFSIZE*2-1, Qnil);
    memo = MEMO_NEW(0, 0, 0);
    data = (struct sort_by_data *)&memo->v1;
    RB_OBJ_WRITE(memo, &data->ary, ary);
    RB_OBJ_WRITE(memo, &data->buf, buf);
    data->n = 0;
    data->primitive_uniformed = SORT_BY_UNIFORMED((CMP_OPTIMIZABLE(FLOAT) && CMP_OPTIMIZABLE(INTEGER)),
                                                  CMP_OPTIMIZABLE(FLOAT),
                                                  CMP_OPTIMIZABLE(INTEGER));
    rb_block_call(obj, id_each, 0, 0, sort_by_i, (VALUE)memo);
    ary = data->ary;
    buf = data->buf;
    if (data->n) {
        rb_ary_resize(buf, data->n*2);
        rb_ary_concat(ary, buf);
    }
    if (RARRAY_LEN(ary) > 2) {
        if (data->primitive_uniformed) {
            RARRAY_PTR_USE(ary, ptr,
                           rb_uniform_intro_sort_2((struct rb_uniform_sort_data*)ptr,
                                                   (struct rb_uniform_sort_data*)(ptr + RARRAY_LEN(ary))));
        }
        else {
            RARRAY_PTR_USE(ary, ptr,
                           ruby_qsort(ptr, RARRAY_LEN(ary)/2, 2*sizeof(VALUE),
                                      sort_by_cmp, (void *)ary));
        }
    }
    if (RBASIC(ary)->klass) {
        rb_raise(rb_eRuntimeError, "sort_by reentered");
    }
    for (i=1; i<RARRAY_LEN(ary); i+=2) {
        RARRAY_ASET(ary, i/2, RARRAY_AREF(ary, i));
    }
    rb_ary_resize(ary, RARRAY_LEN(ary)/2);
    RBASIC_SET_CLASS_RAW(ary, rb_cArray);

    return ary;
}

With a block given, returns an array of elements of self, sorted according to the value returned by the block for each element. The ordering of equal elements is indeterminate and may be unstable.

Examples:

a = %w[xx xxx x xxxx]
a.sort_by {|s| s.size }        # => ["x", "xx", "xxx", "xxxx"]
a.sort_by {|s| -s.size }       # => ["xxxx", "xxx", "xx", "x"]
h = {foo: 2, bar: 1, baz: 0}
h.sort_by{|key, value| value } # => [[:baz, 0], [:bar, 1], [:foo, 2]]
h.sort_by{|key, value| key }   # => [[:bar, 1], [:baz, 0], [:foo, 2]]

With no block given, returns an Enumerator.

The current implementation of sort_by generates an array of tuples containing the original collection element and the mapped value. This makes sort_by fairly expensive when the keysets are simple.

require 'benchmark'

a = (1..100000).map { rand(100000) }

Benchmark.bm(10) do |b|
  b.report("Sort")    { a.sort }
  b.report("Sort by") { a.sort_by { |a| a } }
end

produces:

user     system      total        real
Sort        0.180000   0.000000   0.180000 (  0.175469)
Sort by     1.980000   0.040000   2.020000 (  2.013586)

However, consider the case where comparing the keys is a non-trivial operation. The following code sorts some files on modification time using the basic sort method.

files = Dir["*"]
sorted = files.sort { |a, b| File.new(a).mtime <=> File.new(b).mtime }
sorted   #=> ["mon", "tues", "wed", "thurs"]

This sort is inefficient: it generates two new File objects during every comparison. A slightly better technique is to use the Kernel#test method to generate the modification times directly.

files = Dir["*"]
sorted = files.sort { |a, b|
  test(?M, a) <=> test(?M, b)
}
sorted   #=> ["mon", "tues", "wed", "thurs"]

This still generates many unnecessary Time objects. A more efficient technique is to cache the sort keys (modification times in this case) before the sort. Perl users often call this approach a Schwartzian transform, after Randal Schwartz. We construct a temporary array, where each element is an array containing our sort key along with the filename. We sort this array, and then extract the filename from the result.

sorted = Dir["*"].collect { |f|
   [test(?M, f), f]
}.sort.collect { |f| f[1] }
sorted   #=> ["mon", "tues", "wed", "thurs"]

This is exactly what sort_by does internally.

sorted = Dir["*"].sort_by { |f| test(?M, f) }
sorted   #=> ["mon", "tues", "wed", "thurs"]

To produce the reverse of a specific order, the following can be used:

ary.sort_by { ... }.reverse!

static VALUE
enum_sum(int argc, VALUE* argv, VALUE obj)
{
    struct enum_sum_memo memo;
    VALUE beg, end;
    int excl;

    memo.v = (rb_check_arity(argc, 0, 1) == 0) ? LONG2FIX(0) : argv[0];
    memo.block_given = rb_block_given_p();
    memo.n = 0;
    memo.r = Qundef;

    if ((memo.float_value = RB_FLOAT_TYPE_P(memo.v))) {
        memo.f = RFLOAT_VALUE(memo.v);
        memo.c = 0.0;
    }
    else {
        memo.f = 0.0;
        memo.c = 0.0;
    }

    if (RTEST(rb_range_values(obj, &beg, &end, &excl))) {
        if (!memo.block_given && !memo.float_value &&
                (FIXNUM_P(beg) || RB_BIGNUM_TYPE_P(beg)) &&
                (FIXNUM_P(end) || RB_BIGNUM_TYPE_P(end))) {
            return int_range_sum(beg, end, excl, memo.v);
        }
    }

    if (RB_TYPE_P(obj, T_HASH) &&
            rb_method_basic_definition_p(CLASS_OF(obj), id_each))
        hash_sum(obj, &memo);
    else
        rb_block_call(obj, id_each, 0, 0, enum_sum_i, (VALUE)&memo);

    if (memo.float_value) {
        return DBL2NUM(memo.f + memo.c);
    }
    else {
        if (memo.n != 0)
            memo.v = rb_fix_plus(LONG2FIX(memo.n), memo.v);
        if (!UNDEF_P(memo.r)) {
            memo.v = rb_rational_plus(memo.r, memo.v);
        }
        return memo.v;
    }
}

With no block given, returns the sum of initial_value and the elements:

(1..100).sum          # => 5050
(1..100).sum(1)       # => 5051
('a'..'d').sum('foo') # => "fooabcd"

Generally, the sum is computed using methods + and each; for performance optimizations, those methods may not be used, and so any redefinition of those methods may not have effect here.

One such optimization: When possible, computes using Gauss’s summation formula n(n+1)/2:

100 * (100 + 1) / 2 # => 5050

With a block given, calls the block with each element; returns the sum of initial_value and the block return values:

(1..4).sum {|i| i*i }                        # => 30
(1..4).sum(100) {|i| i*i }                   # => 130
h = {a: 0, b: 1, c: 2, d: 3, e: 4, f: 5}
h.sum {|key, value| value.odd? ? value : 0 } # => 9
('a'..'f').sum('x') {|c| c < 'd' ? c : '' }  # => "xabc"

static VALUE
enum_take(VALUE obj, VALUE n)
{
    struct MEMO *memo;
    VALUE result;
    long len = NUM2LONG(n);

    if (len < 0) {
        rb_raise(rb_eArgError, "attempt to take negative size");
    }

    if (len == 0) return rb_ary_new2(0);
    result = rb_ary_new2(len);
    memo = MEMO_NEW(result, 0, len);
    rb_block_call(obj, id_each, 0, 0, take_i, (VALUE)memo);
    return result;
}

For non-negative integer n, returns the first n elements:

r = (1..4)
r.take(2) # => [1, 2]
r.take(0) # => []

h = {foo: 0, bar: 1, baz: 2, bat: 3}
h.take(2) # => [[:foo, 0], [:bar, 1]]

static VALUE
enum_take_while(VALUE obj)
{
    VALUE ary;

    RETURN_ENUMERATOR(obj, 0, 0);
    ary = rb_ary_new();
    rb_block_call(obj, id_each, 0, 0, take_while_i, ary);
    return ary;
}

Calls the block with successive elements as long as the block returns a truthy value; returns an array of all elements up to that point:

(1..4).take_while{|i| i < 3 } # => [1, 2]
h = {foo: 0, bar: 1, baz: 2}
h.take_while{|element| key, value = *element; value < 2 }
# => [[:foo, 0], [:bar, 1]]

With no block given, returns an Enumerator.

static VALUE
enum_tally(int argc, VALUE *argv, VALUE obj)
{
    VALUE hash;
    if (rb_check_arity(argc, 0, 1)) {
        hash = rb_to_hash_type(argv[0]);
        rb_check_frozen(hash);
    }
    else {
        hash = rb_hash_new();
    }

    return enum_hashify_into(obj, 0, 0, tally_i, hash);
}

When argument hash is not given, returns a new hash whose keys are the distinct elements in self; each integer value is the count of occurrences of each element:

%w[a b c b c a c b].tally # => {"a"=>2, "b"=>3, "c"=>3}

When argument hash is given, returns hash, possibly augmented; for each element ele in self:

• Adds it as a key with a zero value if that key does not already exist:

  hash[ele] = 0 unless hash.include?(ele)
  

• Increments the value of key ele:

  hash[ele] += 1
  

This is useful for accumulating tallies across multiple enumerables:

h = {}                   # => {}
%w[a c d b c a].tally(h) # => {"a"=>2, "c"=>2, "d"=>1, "b"=>1}
%w[b a z].tally(h)       # => {"a"=>3, "c"=>2, "d"=>1, "b"=>2, "z"=>1}
%w[b a m].tally(h)       # => {"a"=>4, "c"=>2, "d"=>1, "b"=>3, "z"=>1, "m"=>1}

The key to be added or found for an element depends on the class of self; see Enumerable in Ruby Classes.

Examples:

• Array (and certain array-like classes): the key is the element (as above).

• Hash (and certain hash-like classes): the key is the 2-element array formed from the key-value pair:

  h = {}                        # => {}
  {foo: 'a', bar: 'b'}.tally(h) # => {[:foo, "a"]=>1, [:bar, "b"]=>1}
  {foo: 'c', bar: 'd'}.tally(h) # => {[:foo, "a"]=>1, [:bar, "b"]=>1, [:foo, "c"]=>1, [:bar, "d"]=>1}
  {foo: 'a', bar: 'b'}.tally(h) # => {[:foo, "a"]=>2, [:bar, "b"]=>2, [:foo, "c"]=>1, [:bar, "d"]=>1}
  {foo: 'c', bar: 'd'}.tally(h) # => {[:foo, "a"]=>2, [:bar, "b"]=>2, [:foo, "c"]=>2, [:bar, "d"]=>2}
  

static VALUE
enum_to_a(int argc, VALUE *argv, VALUE obj)
{
    VALUE ary = rb_ary_new();

    rb_block_call_kw(obj, id_each, argc, argv, collect_all, ary, RB_PASS_CALLED_KEYWORDS);

    return ary;
}

Returns an array containing the items in self:

(0..4).to_a # => [0, 1, 2, 3, 4]

static VALUE
enum_to_h(int argc, VALUE *argv, VALUE obj)
{
    rb_block_call_func *iter = rb_block_given_p() ? enum_to_h_ii : enum_to_h_i;
    return enum_hashify(obj, argc, argv, iter);
}

When self consists of 2-element arrays, returns a hash each of whose entries is the key-value pair formed from one of those arrays:

[[:foo, 0], [:bar, 1], [:baz, 2]].to_h # => {:foo=>0, :bar=>1, :baz=>2}

When a block is given, the block is called with each element of self; the block should return a 2-element array which becomes a key-value pair in the returned hash:

(0..3).to_h {|i| [i, i ** 2]} # => {0=>0, 1=>1, 2=>4, 3=>9}

Raises an exception if an element of self is not a 2-element array, and a block is not passed.

# File lib/set.rb, line 847
def to_set(klass = Set, *args, &block)
  klass.new(self, *args, &block)
end

Makes a set from the enumerable object with given arguments. Needs to require "set" to use this method.

static VALUE
enum_uniq(VALUE obj)
{
    VALUE hash, ret;
    rb_block_call_func *const func =
        rb_block_given_p() ? uniq_iter : uniq_func;

    hash = rb_obj_hide(rb_hash_new());
    rb_block_call(obj, id_each, 0, 0, func, hash);
    ret = rb_hash_values(hash);
    rb_hash_clear(hash);
    return ret;
}

With no block, returns a new array containing only unique elements; the array has no two elements e0 and e1 such that e0.eql?(e1):

%w[a b c c b a a b c].uniq       # => ["a", "b", "c"]
[0, 1, 2, 2, 1, 0, 0, 1, 2].uniq # => [0, 1, 2]

With a block, returns a new array containing elements only for which the block returns a unique value:

a = [0, 1, 2, 3, 4, 5, 5, 4, 3, 2, 1]
a.uniq {|i| i.even? ? i : 0 } # => [0, 2, 4]
a = %w[a b c d e e d c b a a b c d e]
a.uniq {|c| c < 'c' }         # => ["a", "c"]

static VALUE
enum_zip(int argc, VALUE *argv, VALUE obj)
{
    int i;
    ID conv;
    struct MEMO *memo;
    VALUE result = Qnil;
    VALUE args = rb_ary_new4(argc, argv);
    int allary = TRUE;

    argv = RARRAY_PTR(args);
    for (i=0; i<argc; i++) {
        VALUE ary = rb_check_array_type(argv[i]);
        if (NIL_P(ary)) {
            allary = FALSE;
            break;
        }
        argv[i] = ary;
    }
    if (!allary) {
        static const VALUE sym_each = STATIC_ID2SYM(id_each);
        CONST_ID(conv, "to_enum");
        for (i=0; i<argc; i++) {
            if (!rb_respond_to(argv[i], id_each)) {
                rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (must respond to :each)",
                         rb_obj_class(argv[i]));
            }
            argv[i] = rb_funcallv(argv[i], conv, 1, &sym_each);
        }
    }
    if (!rb_block_given_p()) {
        result = rb_ary_new();
    }

    /* TODO: use NODE_DOT2 as memo(v, v, -) */
    memo = MEMO_NEW(result, args, 0);
    rb_block_call(obj, id_each, 0, 0, allary ? zip_ary : zip_i, (VALUE)memo);

    return result;
}

With no block given, returns a new array new_array of size self.size whose elements are arrays. Each nested array new_array[n] is of size other_enums.size+1, and contains:

• The n-th element of self.

• The n-th element of each of the other_enums.

If all other_enums and self are the same size, all elements are included in the result, and there is no nil-filling:

a = [:a0, :a1, :a2, :a3]
b = [:b0, :b1, :b2, :b3]
c = [:c0, :c1, :c2, :c3]
d = a.zip(b, c)
d # => [[:a0, :b0, :c0], [:a1, :b1, :c1], [:a2, :b2, :c2], [:a3, :b3, :c3]]

f = {foo: 0, bar: 1, baz: 2}
g = {goo: 3, gar: 4, gaz: 5}
h = {hoo: 6, har: 7, haz: 8}
d = f.zip(g, h)
d # => [
  #      [[:foo, 0], [:goo, 3], [:hoo, 6]],
  #      [[:bar, 1], [:gar, 4], [:har, 7]],
  #      [[:baz, 2], [:gaz, 5], [:haz, 8]]
  #    ]

If any enumerable in other_enums is smaller than self, fills to self.size with nil:

a = [:a0, :a1, :a2, :a3]
b = [:b0, :b1, :b2]
c = [:c0, :c1]
d = a.zip(b, c)
d # => [[:a0, :b0, :c0], [:a1, :b1, :c1], [:a2, :b2, nil], [:a3, nil, nil]]

If any enumerable in other_enums is larger than self, its trailing elements are ignored:

a = [:a0, :a1, :a2, :a3]
b = [:b0, :b1, :b2, :b3, :b4]
c = [:c0, :c1, :c2, :c3, :c4, :c5]
d = a.zip(b, c)
d # => [[:a0, :b0, :c0], [:a1, :b1, :c1], [:a2, :b2, :c2], [:a3, :b3, :c3]]

When a block is given, calls the block with each of the sub-arrays (formed as above); returns nil:

a = [:a0, :a1, :a2, :a3]
b = [:b0, :b1, :b2, :b3]
c = [:c0, :c1, :c2, :c3]
a.zip(b, c) {|sub_array| p sub_array} # => nil

Output:

[:a0, :b0, :c0]
[:a1, :b1, :c1]
[:a2, :b2, :c2]
[:a3, :b3, :c3]