对于那些喜欢在IPS中测量速度的人;)

require 'benchmark/ips'

ary = []
1000.times { 
  ary << {:bar => rand(1000)} 
}

Benchmark.ips do |x|
  x.report("sort")               { ary.sort{ |a,b| b[:bar] <=> a[:bar] } }
  x.report("sort reverse")       { ary.sort{ |a,b| a[:bar] <=> b[:bar] }.reverse }
  x.report("sort_by -a[:bar]")   { ary.sort_by{ |a| -a[:bar] } }
  x.report("sort_by a[:bar]*-1") { ary.sort_by{ |a| a[:bar]*-1 } }
  x.report("sort_by.reverse!")   { ary.sort_by{ |a| a[:bar] }.reverse }
  x.compare!
end

和结果:

Warming up --------------------------------------
                sort    93.000  i/100ms
        sort reverse    91.000  i/100ms
    sort_by -a[:bar]   382.000  i/100ms
  sort_by a[:bar]*-1   398.000  i/100ms
    sort_by.reverse!   397.000  i/100ms
Calculating -------------------------------------
                sort    938.530  (± 1.8%) i/s -      4.743k in   5.055290s
        sort reverse    901.157  (± 6.1%) i/s -      4.550k in   5.075351s
    sort_by -a[:bar]      3.814k (± 4.4%) i/s -     19.100k in   5.019260s
  sort_by a[:bar]*-1      3.732k (± 4.3%) i/s -     18.706k in   5.021720s
    sort_by.reverse!      3.928k (± 3.6%) i/s -     19.850k in   5.060202s

Comparison:
    sort_by.reverse!:     3927.8 i/s
    sort_by -a[:bar]:     3813.9 i/s - same-ish: difference falls within error
  sort_by a[:bar]*-1:     3732.3 i/s - same-ish: difference falls within error
                sort:      938.5 i/s - 4.19x  slower
        sort reverse:      901.2 i/s - 4.36x  slower

简单说一下，这表示降序的意图。

descending = -1
a.sort_by { |h| h[:bar] * descending }

(同时会想出更好的办法);)

a.sort_by { |h| h[:bar] }.reverse!

从升序到降序的简单解是:

字符串

str = ['ravi', 'aravind', 'joker', 'poker']
asc_string = str.sort # => ["aravind", "joker", "poker", "ravi"]
asc_string.reverse # => ["ravi", "poker", "joker", "aravind"]

数字

digit = [234,45,1,5,78,45,34,9]
asc_digit = digit.sort # => [1, 5, 9, 34, 45, 45, 78, 234]
asc_digit.reverse # => [234, 78, 45, 45, 34, 9, 5, 1]

是什么:

 a.sort {|x,y| y[:bar]<=>x[:bar]}

它的工作原理!

irb
>> a = [
?>   { :foo => 'foo', :bar => 2 },
?>   { :foo => 'foo', :bar => 3 },
?>   { :foo => 'foo', :bar => 5 },
?> ]
=> [{:bar=>2, :foo=>"foo"}, {:bar=>3, :foo=>"foo"}, {:bar=>5, :foo=>"foo"}]

>>  a.sort {|x,y| y[:bar]<=>x[:bar]}
=> [{:bar=>5, :foo=>"foo"}, {:bar=>3, :foo=>"foo"}, {:bar=>2, :foo=>"foo"}]

我看到我们(和其他国家)基本上有两个选择:

a.sort_by { |h| -h[:bar] }

and

a.sort_by { |h| h[:bar] }.reverse

虽然当排序键是唯一的时，两种方法都会得到相同的结果，但请记住，相反的方法会颠倒相等的键的顺序。

例子:

a = [{foo: 1, bar: 1},{foo: 2,bar: 1}]
a.sort_by {|h| -h[:bar]}
 => [{:foo=>1, :bar=>1}, {:foo=>2, :bar=>1}]
a.sort_by {|h| h[:bar]}.reverse
 => [{:foo=>2, :bar=>1}, {:foo=>1, :bar=>1}]

虽然你通常不需要关心这个，但有时你会关心。为了避免这种行为，你可以引入第二个排序键(至少对于所有具有相同排序键的项来说，它必须是唯一的):

a.sort_by {|h| [-h[:bar],-h[:foo]]}
 => [{:foo=>2, :bar=>1}, {:foo=>1, :bar=>1}]
a.sort_by {|h| [h[:bar],h[:foo]]}.reverse
 => [{:foo=>2, :bar=>1}, {:foo=>1, :bar=>1}]

对于上面提到的基准测试套件，这些结果也适用于排序数组。

Sort_by /reverse它是:

# foo.rb
require 'benchmark'

NUM_RUNS = 1000

# arr = []
arr1 = 3000.times.map { { num: rand(1000) } }
arr2 = 3000.times.map { |n| { num: n } }.reverse

Benchmark.bm(20) do |x|
  { 'randomized'     => arr1,
    'sorted'         => arr2 }.each do |label, arr|
    puts '---------------------------------------------------'
    puts label

    x.report('sort_by / reverse') {
      NUM_RUNS.times { arr.sort_by { |h| h[:num] }.reverse }
    }
    x.report('sort_by -') {
      NUM_RUNS.times { arr.sort_by { |h| -h[:num] } }
    }
  end
end

结果是:

$: ruby foo.rb
                           user     system      total        real
---------------------------------------------------
randomized
sort_by / reverse      1.680000   0.010000   1.690000 (  1.682051)
sort_by -              1.830000   0.000000   1.830000 (  1.830359)
---------------------------------------------------
sorted
sort_by / reverse      0.400000   0.000000   0.400000 (  0.402990)
sort_by -              0.500000   0.000000   0.500000 (  0.499350)