Returns an array of substrings of +self+
that are the result of splitting +self+
at each occurrence of the given field separator +field_sep+.

When +field_sep+ is <tt>$;</tt>:

- If <tt>$;</tt> is +nil+ (its default value),
  the split occurs just as if +field_sep+ were given as a space character
  (see below).

- If <tt>$;</tt> is a string,
  the split occurs just as if +field_sep+ were given as that string
  (see below).

When +field_sep+ is <tt>' '</tt> and +limit+ is +0+ (its default value),
the split occurs at each sequence of whitespace:

  'abc def ghi'.split(' ')          # => ["abc", "def", "ghi"]
  "abc \n\tdef\t\n  ghi".split(' ') # => ["abc", "def", "ghi"]
  'abc  def   ghi'.split(' ')       # => ["abc", "def", "ghi"]
  ''.split(' ')                     # => []

When +field_sep+ is a string different from <tt>' '</tt>
and +limit+ is +0+,
the split occurs at each occurrence of +field_sep+;
trailing empty substrings are not returned:

  'abracadabra'.split('ab')   # => ["", "racad", "ra"]
  'aaabcdaaa'.split('a')      # => ["", "", "", "bcd"]
  ''.split('a')               # => []
  '3.14159'.split('1')        # => ["3.", "4", "59"]
  '!@#$%^$&*($)_+'.split('$') # => ["!@#", "%^", "&*(", ")_+"]
  'тест'.split('т')           # => ["", "ес"]
  'こんにちは'.split('に')      # => ["こん", "ちは"]

When +field_sep+ is a Regexp and +limit+ is +0+,
the split occurs at each occurrence of a match;
trailing empty substrings are not returned:

  'abracadabra'.split(/ab/) # => ["", "racad", "ra"]
  'aaabcdaaa'.split(/a/)    # => ["", "", "", "bcd"]
  'aaabcdaaa'.split(//)     # => ["a", "a", "a", "b", "c", "d", "a", "a", "a"]
  '1 + 1 == 2'.split(/\W+/) # => ["1", "1", "2"]

If the \Regexp contains groups, their matches are also included
in the returned array:

  '1:2:3'.split(/(:)()()/, 2) # => ["1", ":", "", "", "2:3"]

As seen above, if +limit+ is +0+,
trailing empty substrings are not returned:

  'aaabcdaaa'.split('a')    # => ["", "", "", "bcd"]

If +limit+ is positive integer +n+, no more than <tt>n - 1-</tt>
splits occur, so that at most +n+ substrings are returned,
and trailing empty substrings are included:

  'aaabcdaaa'.split('a', 1) # => ["aaabcdaaa"]
  'aaabcdaaa'.split('a', 2) # => ["", "aabcdaaa"]
  'aaabcdaaa'.split('a', 5) # => ["", "", "", "bcd", "aa"]
  'aaabcdaaa'.split('a', 7) # => ["", "", "", "bcd", "", "", ""]
  'aaabcdaaa'.split('a', 8) # => ["", "", "", "bcd", "", "", ""]

Note that if +field_sep+ is a \Regexp containing groups,
their matches are in the returned array, but do not count toward the limit.

If +limit+ is negative, it behaves the same as if +limit+ was zero,
meaning that there is no limit,
and trailing empty substrings are included:

  'aaabcdaaa'.split('a', -1) # => ["", "", "", "bcd", "", "", ""]

If a block is given, it is called with each substring and returns +self+:

  'abc def ghi'.split(' ') {|substring| p substring }

Output:

  "abc"
  "def"
  "ghi"
  => "abc def ghi"

Note that the above example is functionally the same as calling +#each+ after
+#split+ and giving the same block. However, the above example has better
performance because it avoids the creation of an intermediate array. Also,
note the different return values.

  'abc def ghi'.split(' ').each {|substring| p substring }

Output:

  "abc"
  "def"
  "ghi"
  => ["abc", "def", "ghi"]

Related: String#partition, String#rpartition.