Does Enumerable#group_by preserve the original order within each value? When I get this:
[1, 2, 3, 4, 5].group_by{|i| i % 2}
# => {1=>[1, 3, 5], 0=>
Yes, Enumerable#group_by
preserves input order.
Here's the implementation of that method in MRI, from https://github.com/ruby/ruby/blob/trunk/enum.c:
static VALUE
enum_group_by(VALUE obj)
{
VALUE hash;
RETURN_SIZED_ENUMERATOR(obj, 0, 0, enum_size);
hash = rb_hash_new();
rb_block_call(obj, id_each, 0, 0, group_by_i, hash);
OBJ_INFECT(hash, obj);
return hash;
}
static VALUE
group_by_i(RB_BLOCK_CALL_FUNC_ARGLIST(i, hash))
{
VALUE group;
VALUE values;
ENUM_WANT_SVALUE();
group = rb_yield(i);
values = rb_hash_aref(hash, group);
if (!RB_TYPE_P(values, T_ARRAY)) {
values = rb_ary_new3(1, i);
rb_hash_aset(hash, group, values);
}
else {
rb_ary_push(values, i);
}
return Qnil;
}
enum_group_by
calls group_by_i
on each array (obj
) element in order. group_by_i
creates a one-element array (rb_ary_new3(1, i)
) the first time a group is encountered, and pushes on to the array thereafter (rb_ary_push(values, i)
). So the input order is preserved.
Also, RubySpec requires it. From https://github.com/rubyspec/rubyspec/blob/master/core/enumerable/group_by_spec.rb:
it "returns a hash with values grouped according to the block" do
e = EnumerableSpecs::Numerous.new("foo", "bar", "baz")
h = e.group_by { |word| word[0..0].to_sym }
h.should == { :f => ["foo"], :b => ["bar", "baz"]}
end
More specifically, Enumerable
calls each
so it depends on how each
is implemented and whether each
yields the elements in the original order:
class ReverseArray < Array
def each(&block)
reverse_each(&block)
end
end
array = ReverseArray.new([1,2,3,4])
#=> [1, 2, 3, 4]
array.group_by { |i| i % 2 }
#=> {0=>[4, 2], 1=>[3, 1]}