两个列表A和B，如果它们具有相同的长度，则是全等的，而在中A比较相等的元素在中比较相等B。

换句话说，给定任意两个有效索引x和y：

• 如果A[x] = A[y]是的话B[x] = B[y]。
• 如果A[x] != A[y]是的话B[x] != B[y]。

例如，清单[1, 2, 1, 4, 5]和[0, 1, 0, 2, 3]是一致的。

任务

鉴于非空目录非负整数A，返回一个新的列表非负整数，B这样是要一致A，同时尽量减少在整数的总和B。

可能有许多可能的有效输出。例如，在列表中[12, 700, 3]，的任何排列都[0, 1, 2]将被视为有效输出。

测试用例

Format:
input ->
one possible valid output

[1 2 1 4 5] ->
[0 1 0 2 3] (this is the example given above)

[3 2 2 1 5 7 2] ->
[1 0 0 2 3 4 0]

[8 8 8 8 8] ->
[0 0 0 0 0]

[2] ->
[0]

[8 6 8 4 6 8 2 4 6 8 0 2 4 6 8] ->
[0 1 0 2 1 0 3 2 1 0 4 3 2 1 0]

[14 1] ->
[1 0]

[19 6 4 9 14 17 10 9 6 14 8 14 6 15] ->
[8 0 3 2 1 7 5 2 0 1 4 1 0 6]

[15] ->
[0]

[1 18 4 8 6 19 12 17 6 13 7 6 8 1 6] ->
[1 8 3 2 0 9 5 7 0 6 4 0 2 1 0]

[9 10 11 9 7 11 16 17 11 8 7] ->
[2 4 0 2 1 0 5 6 0 3 1]

[1 3 16 19 14] ->
[0 1 3 4 2]

[18 8] ->
[1 0]

[13 4 9 6] ->
[3 0 2 1]

[16 16 18 6 12 10 4 6] ->
[1 1 5 0 4 3 2 0]

[11 18] ->
[0 1]

[14 18 18 11 9 8 13 3 3 4] ->
[7 1 1 5 4 3 6 0 0 2]

[20 19 1 1 13] ->
[3 2 0 0 1]

[12] ->
[0]

[1 14 20 4 18 15 19] ->
[0 2 6 1 4 3 5]

[13 18 20] ->
[0 1 2]

[9 1 12 2] ->
[2 0 3 1]

[15 11 2 9 10 19 17 10 19 11 16 5 13 2] ->
[7 2 0 5 1 3 9 1 3 2 8 4 6 0]

[5 4 2 2 19 14 18 11 3 12 20 14 2 19 7] ->
[5 4 0 0 2 1 9 7 3 8 10 1 0 2 6]

[9 11 13 13 13 12 17 8 4] ->
[3 4 0 0 0 5 6 2 1]

[10 14 16 17 7 4 3] ->
[3 4 5 6 2 1 0]

[2 4 8 7 8 19 16 11 10 19 4 7 8] ->
[4 1 0 2 0 3 7 6 5 3 1 2 0]

[15 17 20 18 20 13 6 10 4 19 9 15 18 17 5] ->
[0 1 3 2 3 9 6 8 4 10 7 0 2 1 5]

[15 14 4 5 5 5 3 3 19 12 4] ->
[5 4 2 0 0 0 1 1 6 3 2]

[7 12] ->
[0 1]

[18 5 18 2 5 20 8 8] ->
[2 0 2 3 0 4 1 1]

[4 6 10 7 3 1] ->
[2 3 5 4 1 0]

[5] ->
[0]

[6 12 14 18] ->
[0 1 2 3]

[7 15 13 3 4 7 20] ->
[0 4 3 1 2 0 5]

[10 15 19 14] ->
[0 2 3 1]

[14] ->
[0]

[19 10 20 12 17 3 6 16] ->
[6 2 7 3 5 0 1 4]

[9 4 7 18 18 15 3] ->
[4 2 3 0 0 5 1]

[7 4 13 7] ->
[0 1 2 0]

[19 1 10 3 1] ->
[3 0 2 1 0]

[8 14 20 4] ->
[1 2 3 0]

[17 20 18 11 1 15 7 2] ->
[5 7 6 3 0 4 2 1]

[11 4 3 17] ->
[2 1 0 3]

[1 9 15 1 20 8 6] ->
[0 3 4 0 5 2 1]

[16 13 10] ->
[2 1 0]

[17 20 20 12 19 10 19 7 8 5 12 19] ->
[7 2 2 1 0 6 0 4 5 3 1 0]

[18 11] ->
[1 0]

[2 16 7 12 10 18 4 14 14 7 15 4 8 3 14] ->
[3 9 2 7 6 10 1 0 0 2 8 1 5 4 0]

[5 7 2 2 16 14 7 7 18 19 16] ->
[3 0 1 1 2 4 0 0 5 6 2]

[8 6 17 5 10 2 14] ->
[3 2 6 1 4 0 5]

这是代码高尔夫球，因此最短的有效提交（以字节为单位）获胜。

Python 2中，62个 54字节

lambda L:map(sorted(set(L),key=L.count)[::-1].index,L)

在线尝试！

编辑：通过地图thanx保存8个字节到Maltysen

Pyth- 12 11 10个字节

XQ_o/QN{QU

测试套件。

Japt，11个字节

£â ñ@è¦XÃbX

在线测试！

说明

 £   â ñ@  è¦ Xà bX
UmX{Uâ ñX{Uè!=X} bX}   Ungolfed
                       Implicit: U = input array
UmX{               }   Map each item X in the input to:
    Uâ                   Take the unique items of U.
       ñX{     }         Sort each item X in this by
          Uè!=X            how many items in U are not equal to X.
                         This sorts the items that occur most to the front of the list.
                 bX      Return the index of X in this list.
                       Implicit: output result of last expression

J，11个字节

i.~~.\:#/.~

在线尝试！

说明

i.~~.\:#/.~  Input: array A
       #/.~  Frequency of each unique character, sorted by first appearance
   ~.        Unique, sorted by first appearance
     \:      Sort down the uniques using their frequencies
i.~          First index in that for each element of A

Röda，55个字节

{|l|l|indexOf _,[sort(l)|count|[[-_2,_1]]|sort|_|[_2]]}

在线尝试！

Haskell，93 91 85字节

import Data.List
f a=[i|x<-a,(i,y:_)<-zip[0..]$sortOn((0-).length)$group$sort a,x==y]

在线尝试！

编辑：感谢@Laikoni起飞6个字节！

不太短，但我想不出其他任何东西。这个想法是遍历数组（x<-a）并在元组列表（(i,y:_)<-... ,x==y）中执行查找，该数组根据输入的常见程度为输入中的每个唯一元素分配一个非负整数。该元组列表是通过首先sort将输入，将group其分为相等元素的子列表，按子列表的长度对该列表sortOn((0-).length)进行排序（;长度被取反以按“降序”排序），最后用一个无限列表进行压缩而生成的从0开始递增。我们使用模式匹配将子列表中的实际元素提取到中y。

Mathematica，94个字节

(s=First/@Reverse@SortBy[Tally[j=#],Last];For[i=1,i<=Length@s,j=j//.s[[i]]->i+5!;i++];j-5!-1)&

在线尝试！

CJam，17个 14字节

^{-3个字节感谢Peter Taylor}

这是我用来生成测试用例的程序的简化版本。

{_$e`$W%1f=f#}

这是一个匿名块，期望输入作为堆栈顶部的数组，并在堆栈顶部输出数组。

说明：

{_$e`$W%1f=f#} Stack:                  [1 2 1 4 5]
 _             Duplicate:              [1 2 1 4 5] [1 2 1 4 5]
  $            Sort:                   [1 2 1 4 5] [1 1 2 4 5]
   e`          Run-length encode:      [1 2 1 4 5] [[2 1] [1 2] [1 4] [1 5]]
     $         Sort lexicographically: [1 2 1 4 5] [[1 2] [1 4] [1 5] [2 1]]
      W%       Reverse:                [1 2 1 4 5] [[2 1] [1 5] [1 4] [1 2]]
        1f=    Second element of each: [1 2 1 4 5] [1 5 4 2]
           f#  Vectorized indexing:    [0 3 0 2 1]

TI-BASIC，66字节

Ans+max(Ans+1)seq(sum(Ans=Ans(I)),I,1,dim(Ans→A
cumSum(Ans→B
SortD(∟A,∟B
cumSum(0≠ΔList(augment({0},∟A→A
SortA(∟B,∟A
∟A-1

说明

seq(sum(Ans=Ans(I)),I,1,dim(Ans    Calculates the frequency of each element of Ans.
                                   Comparing a value to a list returns a list of booleans,
                                   so taking the sum will produce the number of matches.

Ans+max(Ans+1)                     Multiplies each frequency by one more than the max element,
                                   then adds each original value.
                                   This ensures that identical values with the same frequency
                                   will be grouped together when sorting.
                                   Additionally, all resulting values will be positive.

→A                                 Stores to ∟A.

cumSum(Ans→B                       Stores the prefix sum of the result into ∟B.
                                   Since ∟A has only positive values, ∟B is guaranteed
                                   to be strictly increasing.

SortD(∟A,∟B                        Sort ∟A in descending order (by frequency), grouping
                                   identical values together. Also, dependently sort ∟B
                                   so the original ordering can be restored.

       0≠ΔList(augment({0},∟A      Prepends a 0 to ∟A and compares each consecutive difference
                                   to 0. This places a 1 at each element that is different
                                   from the previous element, and 0 everywhere else.
                                   The first element is never 0, so it is considered different.

cumSum(                      →A    Takes the prefix sum of this list and stores to ∟A.
                                   Since there is a 1 at each element with a new value,
                                   the running sum will increase by 1 at each value change.
                                   As a result, we've created a unique mapping.

SortA(∟B,∟A                        Sorts ∟B in ascending order with ∟A as a dependent,
                                   restoring the original element ordering.

∟A-1                               Since we started counting up at 1 instead of 0,
                                   subtract 1 from each element in ∟A and return it.

Perl 5，77 +1（-a）= 78字节

$c{$_}++for@F;%r=map{$_=>$i++.$"}sort{$c{$b}<=>$c{$a}}keys%c;print$r{$_}for@F

在线尝试！

JavaScript（ES6），91字节

使用唯一值列表，按频率排序。

x=>x.map(x=>Object.keys(C).sort((a,b)=>C[b]-C[a]).indexOf(x+''),C={},x.map(v=>C[v]=-~C[v]))

测试

var F=
x=>x.map(x=>Object.keys(C).sort((a,b)=>C[b]-C[a]).indexOf(x+''),C={},x.map(v=>C[v]=-~C[v]))

Test=`[1 2 1 4 5] -> [0 1 0 2 3]
[3 2 2 1 5 7 2] -> [1 0 0 2 3 4 0]
[8 8 8 8 8] -> [0 0 0 0 0]
[2] -> [0]
[8 6 8 4 6 8 2 4 6 8 0 2 4 6 8] -> [0 1 0 2 1 0 3 2 1 0 4 3 2 1 0]
[14 1] -> [1 0]
[19 6 4 9 14 17 10 9 6 14 8 14 6 15] -> [8 0 3 2 1 7 5 2 0 1 4 1 0 6]
[15] -> [0]
[1 18 4 8 6 19 12 17 6 13 7 6 8 1 6] -> [1 8 3 2 0 9 5 7 0 6 4 0 2 1 0]
[9 10 11 9 7 11 16 17 11 8 7] -> [2 4 0 2 1 0 5 6 0 3 1]
[1 3 16 19 14] -> [0 1 3 4 2]
[18 8] -> [1 0]
[13 4 9 6] -> [3 0 2 1]
[16 16 18 6 12 10 4 6] -> [1 1 5 0 4 3 2 0]
[11 18] -> [0 1]
[14 18 18 11 9 8 13 3 3 4] -> [7 1 1 5 4 3 6 0 0 2]
[20 19 1 1 13] -> [3 2 0 0 1]
[12] -> [0]
[1 14 20 4 18 15 19] -> [0 2 6 1 4 3 5]
[13 18 20] -> [0 1 2]
[9 1 12 2] -> [2 0 3 1]
[15 11 2 9 10 19 17 10 19 11 16 5 13 2] -> [7 2 0 5 1 3 9 1 3 2 8 4 6 0]
[5 4 2 2 19 14 18 11 3 12 20 14 2 19 7] -> [5 4 0 0 2 1 9 7 3 8 10 1 0 2 6]
[9 11 13 13 13 12 17 8 4] -> [3 4 0 0 0 5 6 2 1]
[10 14 16 17 7 4 3] -> [3 4 5 6 2 1 0]
[2 4 8 7 8 19 16 11 10 19 4 7 8] -> [4 1 0 2 0 3 7 6 5 3 1 2 0]
[15 17 20 18 20 13 6 10 4 19 9 15 18 17 5] -> [0 1 3 2 3 9 6 8 4 10 7 0 2 1 5]
[15 14 4 5 5 5 3 3 19 12 4] -> [5 4 2 0 0 0 1 1 6 3 2]
[7 12] -> [0 1]
[18 5 18 2 5 20 8 8] -> [2 0 2 3 0 4 1 1]
[4 6 10 7 3 1] -> [2 3 5 4 1 0]
[5] -> [0]
[6 12 14 18] -> [0 1 2 3]
[7 15 13 3 4 7 20] -> [0 4 3 1 2 0 5]
[10 15 19 14] -> [0 2 3 1]
[14] -> [0]
[19 10 20 12 17 3 6 16] -> [6 2 7 3 5 0 1 4]
[9 4 7 18 18 15 3] -> [4 2 3 0 0 5 1]
[7 4 13 7] -> [0 1 2 0]
[19 1 10 3 1] -> [3 0 2 1 0]
[8 14 20 4] -> [1 2 3 0]
[17 20 18 11 1 15 7 2] -> [5 7 6 3 0 4 2 1]
[11 4 3 17] -> [2 1 0 3]
[1 9 15 1 20 8 6] -> [0 3 4 0 5 2 1]
[16 13 10] -> [2 1 0]
[17 20 20 12 19 10 19 7 8 5 12 19] -> [7 2 2 1 0 6 0 4 5 3 1 0]
[18 11] -> [1 0]
[2 16 7 12 10 18 4 14 14 7 15 4 8 3 14] -> [3 9 2 7 6 10 1 0 0 2 8 1 5 4 0]
[5 7 2 2 16 14 7 7 18 19 16] -> [3 0 1 1 2 4 0 0 5 6 2]
[8 6 17 5 10 2 14] -> [3 2 6 1 4 0 5]`

Test.split(`\n`).forEach(row => {
  row=row.match(/\d+/g)
  var nv = row.length/2
  var tc = row.slice(0,nv)
  var exp = row.slice(nv)
  var xsum = eval(exp.join`+`)
  var result = F(tc)
  var rsum = eval(result.join`+`)
  var ok = xsum == rsum
  console.log('Test ' + (ok ? 'OK':'KO')
  + '\nInput [' + tc 
  + ']\nExpected (sum ' + xsum + ') ['+ exp 
  + ']\nResult (sum ' + rsum + ') [' + result + ']')
  
})

PHP，92字节

$b=array_unique($a);print_r(array_map(function($v)use($b){return array_search($v,$b);},$a));

在线尝试！

R，58个字节

x=scan();cat(match(x,names(z<-table(x))[rev(order(z))])-1)

在线尝试！

Chas Brown的Python解答端口。

table计算每个元素的计数x（存储值作为names属性），order返回的索引的排列z，并且match返回的第一个匹配的索引x中names(z)。然后减去，1因为R索引基于1。