erlang下lists模块sort(排序)方法源码解析(一)

排序算法一直是各种语言最简单也是最复杂的算法,例如十大经典排序算法(动图演示)里面讲的那样

第一次看lists的sort方法的时候,蒙了,几百行的代码,我心想要这么复杂么(因为C语言的冒泡排序我记得不超过30行),于是自己就实现了下

结果更蒙了

bubble_sort(L)->
	bubble_sort(L,length(L)).
 
bubble_sort(L,0)->
	L;
bubble_sort(L,N)->
	bubble_sort(do_bubble_sort(L),N-1).
 
do_bubble_sort([A])->
	[A];
do_bubble_sort([A,B|R])->
case A<B of
	true ->
		[A|do_bubble_sort([B|R])];
	false ->
		[B|do_bubble_sort([A|R])]
end.

对比结果如下

6> timer:tc(tt1,bubble_sort,[B]).
{21130,
 [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,
  23,24,25,26,27|...]}
7> timer:tc(lists,sort,[B]).     
{162,
 [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,
  23,24,25,26,27|...]}
8> 

B是一个打乱顺序的1到1000的序列,我X,这不是一个数量级的算法啊~~~~,不是说好越简单的代码越快么,三观被刷新了。

还是老实读lists的源码,一共250+行,摘录于lists.erl

  1 -spec sort(List1) -> List2 when
  2       List1 :: [T],
  3       List2 :: [T],
  4       T :: term().
  5 
  6 sort([X, Y | L] = L0) when X =< Y ->
  7     case L of
  8     [] -> 
  9         L0;
 10     [Z] when Y =< Z ->
 11         L0;
 12     [Z] when X =< Z ->
 13         [X, Z, Y];
 14     [Z] ->
 15         [Z, X, Y];
 16     _ when X == Y ->
 17         sort_1(Y, L, [X]);
 18     _ ->
 19         split_1(X, Y, L, [], [])
 20     end;
 21 sort([X, Y | L]) ->
 22     case L of
 23     [] ->
 24         [Y, X];
 25     [Z] when X =< Z ->
 26         [Y, X | L];
 27     [Z] when Y =< Z ->
 28         [Y, Z, X];
 29     [Z] ->
 30         [Z, Y, X];
 31     _ ->
 32         split_2(X, Y, L, [], [])
 33     end;
 34 sort([_] = L) ->
 35     L;
 36 sort([] = L) ->
 37     L.
 38 
 39 sort_1(X, [Y | L], R) when X == Y ->
 40     sort_1(Y, L, [X | R]);
 41 sort_1(X, [Y | L], R) when X < Y ->
 42     split_1(X, Y, L, R, []);
 43 sort_1(X, [Y | L], R) ->
 44     split_2(X, Y, L, R, []);
 45 sort_1(X, [], R) ->
 46     lists:reverse(R, [X]).
 47 
 48 %% Ascending.
 49 split_1(X, Y, [Z | L], R, Rs) when Z >= Y ->
 50     split_1(Y, Z, L, [X | R], Rs);
 51 split_1(X, Y, [Z | L], R, Rs) when Z >= X ->
 52     split_1(Z, Y, L, [X | R], Rs);
 53 split_1(X, Y, [Z | L], [], Rs) ->
 54     split_1(X, Y, L, [Z], Rs);
 55 split_1(X, Y, [Z | L], R, Rs) ->
 56     split_1_1(X, Y, L, R, Rs, Z);
 57 split_1(X, Y, [], R, Rs) ->
 58     rmergel([[Y, X | R] | Rs], []).
 59 
 60 split_1_1(X, Y, [Z | L], R, Rs, S) when Z >= Y ->
 61     split_1_1(Y, Z, L, [X | R], Rs, S);
 62 split_1_1(X, Y, [Z | L], R, Rs, S) when Z >= X ->
 63     split_1_1(Z, Y, L, [X | R], Rs, S);
 64 split_1_1(X, Y, [Z | L], R, Rs, S) when S =< Z ->
 65     split_1(S, Z, L, [], [[Y, X | R] | Rs]);
 66 split_1_1(X, Y, [Z | L], R, Rs, S) ->
 67     split_1(Z, S, L, [], [[Y, X | R] | Rs]);
 68 split_1_1(X, Y, [], R, Rs, S) ->    
 69     rmergel([[S], [Y, X | R] | Rs], []).
 70 
 71 %% Descending.
 72 split_2(X, Y, [Z | L], R, Rs) when Z =< Y ->
 73     split_2(Y, Z, L, [X | R], Rs);
 74 split_2(X, Y, [Z | L], R, Rs) when Z =< X ->
 75     split_2(Z, Y, L, [X | R], Rs);
 76 split_2(X, Y, [Z | L], [], Rs) ->
 77     split_2(X, Y, L, [Z], Rs);
 78 split_2(X, Y, [Z | L], R, Rs) ->
 79     split_2_1(X, Y, L, R, Rs, Z);
 80 split_2(X, Y, [], R, Rs) ->
 81     mergel([[Y, X | R] | Rs], []).
 82 
 83 split_2_1(X, Y, [Z | L], R, Rs, S) when Z =< Y ->
 84     split_2_1(Y, Z, L, [X | R], Rs, S);
 85 split_2_1(X, Y, [Z | L], R, Rs, S) when Z =< X ->
 86     split_2_1(Z, Y, L, [X | R], Rs, S);
 87 split_2_1(X, Y, [Z | L], R, Rs, S) when S > Z ->
 88     split_2(S, Z, L, [], [[Y, X | R] | Rs]);
 89 split_2_1(X, Y, [Z | L], R, Rs, S) ->
 90     split_2(Z, S, L, [], [[Y, X | R] | Rs]);
 91 split_2_1(X, Y, [], R, Rs, S) ->
 92     mergel([[S], [Y, X | R] | Rs], []).
 93 
 94 %% merge/1
 95 
 96 mergel([[] | L], Acc) ->
 97     mergel(L, Acc);
 98 mergel([T1, [H2 | T2], [H3 | T3] | L], Acc) ->
 99     mergel(L, [merge3_1(T1, [], H2, T2, H3, T3) | Acc]);
100 mergel([T1, [H2 | T2]], Acc) ->
101     rmergel([merge2_1(T1, H2, T2, []) | Acc], []);
102 mergel([L], []) ->
103     L;
104 mergel([L], Acc) ->
105     rmergel([lists:reverse(L, []) | Acc], []);
106 mergel([], []) ->
107     [];
108 mergel([], Acc) ->
109     rmergel(Acc, []);
110 mergel([A, [] | L], Acc) ->
111     mergel([A | L], Acc);
112 mergel([A, B, [] | L], Acc) ->
113     mergel([A, B | L], Acc).
114 
115 rmergel([[H3 | T3], [H2 | T2], T1 | L], Acc) ->
116     rmergel(L, [rmerge3_1(T1, [], H2, T2, H3, T3) | Acc]);
117 rmergel([[H2 | T2], T1], Acc) ->
118     mergel([rmerge2_1(T1, H2, T2, []) | Acc], []);
119 rmergel([L], Acc) ->
120     mergel([lists:reverse(L, []) | Acc], []);
121 rmergel([], Acc) ->
122     mergel(Acc, []).
123 
124 %% merge3/3
125 
126 %% Take L1 apart.
127 merge3_1([H1 | T1], M, H2, T2, H3, T3) when H1 =< H2 ->
128     merge3_12(T1, H1, H2, T2, H3, T3, M);
129 merge3_1([H1 | T1], M, H2, T2, H3, T3) ->
130     merge3_21(T1, H1, H2, T2, H3, T3, M);
131 merge3_1([], M, H2, T2, H3, T3) when H2 =< H3 ->
132     merge2_1(T2, H3, T3, [H2 | M]);
133 merge3_1([], M, H2, T2, H3, T3) ->
134     merge2_2(T2, H3, T3, M, H2).
135 
136 %% Take L2 apart.
137 merge3_2(T1, H1, M, [H2 | T2], H3, T3) when H1 =< H2 ->
138     merge3_12(T1, H1, H2, T2, H3, T3, M);
139 merge3_2(T1, H1, M, [H2 | T2], H3, T3) ->
140     merge3_21(T1, H1, H2, T2, H3, T3, M);
141 merge3_2(T1, H1, M, [], H3, T3) when H1 =< H3 ->
142     merge2_1(T1, H3, T3, [H1 | M]);
143 merge3_2(T1, H1, M, [], H3, T3) ->
144     merge2_2(T1, H3, T3, M, H1).
145 
146 % H1 =< H2. Inlined.
147 merge3_12(T1, H1, H2, T2, H3, T3, M) when H1 =< H3 ->
148     merge3_1(T1, [H1 | M], H2, T2, H3, T3);
149 merge3_12(T1, H1, H2, T2, H3, T3, M) ->
150     merge3_12_3(T1, H1, H2, T2, [H3 | M], T3).
151 
152 % H1 =< H2, take L3 apart.
153 merge3_12_3(T1, H1, H2, T2, M, [H3 | T3]) when H1 =< H3 ->
154     merge3_1(T1, [H1 | M], H2, T2, H3, T3);
155 merge3_12_3(T1, H1, H2, T2, M, [H3 | T3]) ->
156     merge3_12_3(T1, H1, H2, T2, [H3 | M], T3);
157 merge3_12_3(T1, H1, H2, T2, M, []) ->
158     merge2_1(T1, H2, T2, [H1 | M]).
159 
160 % H1 > H2. Inlined.
161 merge3_21(T1, H1, H2, T2, H3, T3, M) when H2 =< H3 ->
162     merge3_2(T1, H1, [H2 | M], T2, H3, T3);
163 merge3_21(T1, H1, H2, T2, H3, T3, M) ->
164     merge3_21_3(T1, H1, H2, T2, [H3 | M], T3).
165 
166 % H1 > H2, take L3 apart.
167 merge3_21_3(T1, H1, H2, T2, M, [H3 | T3]) when H2 =< H3 ->
168     merge3_2(T1, H1, [H2 | M], T2, H3, T3);
169 merge3_21_3(T1, H1, H2, T2, M, [H3 | T3]) ->
170     merge3_21_3(T1, H1, H2, T2, [H3 | M], T3);
171 merge3_21_3(T1, H1, H2, T2, M, []) ->
172     merge2_2(T1, H2, T2, M, H1).
173 
174 %% rmerge/3
175 
176 %% Take L1 apart.
177 rmerge3_1([H1 | T1], M, H2, T2, H3, T3) when H1 =< H2 ->
178     rmerge3_12(T1, H1, H2, T2, H3, T3, M);
179 rmerge3_1([H1 | T1], M, H2, T2, H3, T3) ->
180     rmerge3_21(T1, H1, H2, T2, H3, T3, M);
181 rmerge3_1([], M, H2, T2, H3, T3) when H2 =< H3 ->
182     rmerge2_2(T2, H3, T3, M, H2);
183 rmerge3_1([], M, H2, T2, H3, T3) ->
184     rmerge2_1(T2, H3, T3, [H2 | M]).
185 
186 %% Take L2 apart.
187 rmerge3_2(T1, H1, M, [H2 | T2], H3, T3) when H1 =< H2 ->
188     rmerge3_12(T1, H1, H2, T2, H3, T3, M);
189 rmerge3_2(T1, H1, M, [H2 | T2], H3, T3) ->
190     rmerge3_21(T1, H1, H2, T2, H3, T3, M);
191 rmerge3_2(T1, H1, M, [], H3, T3) when H1 =< H3 ->
192     rmerge2_2(T1, H3, T3, M, H1);
193 rmerge3_2(T1, H1, M, [], H3, T3) ->
194     rmerge2_1(T1, H3, T3, [H1 | M]).
195 
196 % H1 =< H2. Inlined.
197 rmerge3_12(T1, H1, H2, T2, H3, T3, M) when H2 =< H3 ->
198     rmerge3_12_3(T1, H1, H2, T2, [H3 | M], T3);
199 rmerge3_12(T1, H1, H2, T2, H3, T3, M) ->
200     rmerge3_2(T1, H1, [H2 | M], T2, H3, T3).
201 
202 % H1 =< H2, take L3 apart.
203 rmerge3_12_3(T1, H1, H2, T2, M, [H3 | T3]) when H2 =< H3 ->
204     rmerge3_12_3(T1, H1, H2, T2, [H3 | M], T3);
205 rmerge3_12_3(T1, H1, H2, T2, M, [H3 | T3]) ->
206     rmerge3_2(T1, H1, [H2 | M], T2, H3, T3);
207 rmerge3_12_3(T1, H1, H2, T2, M, []) ->
208     rmerge2_2(T1, H2, T2, M, H1).
209 
210 % H1 > H2. Inlined.
211 rmerge3_21(T1, H1, H2, T2, H3, T3, M) when H1 =< H3 ->
212     rmerge3_21_3(T1, H1, H2, T2, [H3 | M], T3);
213 rmerge3_21(T1, H1, H2, T2, H3, T3, M) ->
214     rmerge3_1(T1, [H1 | M], H2, T2, H3, T3).
215 
216 % H1 > H2, take L3 apart.
217 rmerge3_21_3(T1, H1, H2, T2, M, [H3 | T3]) when H1 =< H3 ->
218     rmerge3_21_3(T1, H1, H2, T2, [H3 | M], T3);
219 rmerge3_21_3(T1, H1, H2, T2, M, [H3 | T3]) ->
220     rmerge3_1(T1, [H1 | M], H2, T2, H3, T3);
221 rmerge3_21_3(T1, H1, H2, T2, M, []) ->
222     rmerge2_1(T1, H2, T2, [H1 | M]).
223 
224 %% merge/2
225 
226 merge2_1([H1 | T1], H2, T2, M) when H1 =< H2 ->
227     merge2_1(T1, H2, T2, [H1 | M]);
228 merge2_1([H1 | T1], H2, T2, M) ->
229     merge2_2(T1, H2, T2, M, H1);
230 merge2_1([], H2, T2, M) ->
231     lists:reverse(T2, [H2 | M]).
232 
233 merge2_2(T1, HdM, [H2 | T2], M, H1) when H1 =< H2 ->
234     merge2_1(T1, H2, T2, [H1, HdM | M]);
235 merge2_2(T1, HdM, [H2 | T2], M, H1) ->
236     merge2_2(T1, H2, T2, [HdM | M], H1);
237 merge2_2(T1, HdM, [], M, H1) ->
238     lists:reverse(T1, [H1, HdM | M]).
239 
240 %% rmerge/2
241 
242 rmerge2_1([H1 | T1], H2, T2, M) when H1 =< H2 ->
243     rmerge2_2(T1, H2, T2, M, H1);
244 rmerge2_1([H1 | T1], H2, T2, M) ->
245     rmerge2_1(T1, H2, T2, [H1 | M]);
246 rmerge2_1([], H2, T2, M) ->
247     lists:reverse(T2, [H2 | M]).
248 
249 rmerge2_2(T1, HdM, [H2 | T2], M, H1) when H1 =< H2 ->
250     rmerge2_2(T1, H2, T2, [HdM | M], H1);
251 rmerge2_2(T1, HdM, [H2 | T2], M, H1) ->
252     rmerge2_1(T1, H2, T2, [H1, HdM | M]);
253 rmerge2_2(T1, HdM, [], M, H1) ->
254     lists:reverse(T1, [H1, HdM | M]).

好,这是我见过最复杂的排序算法了。

这个算法和归并排序有点像,可是由于erlang的特性,变量不能变,使得和大部分的排序方法有很大的区别,这个算法的复杂度应该是0(2n)

这个算法可以份3大块,第一块是sort_*函数,第二块是split_*,第3块是rmergel和mergel

首先

sort([X, Y | L] = L0) when X =< Y ->  %当list是3个对比会返回,当list超过3个进入sort_1或者splite_*函数
..........
sort([X, Y | L]) ->            %分了2种情况,第一个元素大于第二个 或者 第一个元素小于等于第二个
.......
sort([_] = L) ->         %list只有1个也直接返回   
    L;
sort([] = L) ->          %list为空直接返回
    L.

sort_1(X, [Y | L], R) when X == Y ->
    sort_1(Y, L, [X | R]);
sort_1(X, [Y | L], R) when X < Y ->
    split_1(X, Y, L, R, []);
sort_1(X, [Y | L], R) ->
    split_2(X, Y, L, R, []);
sort_1(X, [], R) ->
    lists:reverse(R, [X]).

当这段代码还是比较清晰的,就说把超过3个元素的list传入split_*

下面看split_1系列

%% Ascending.

split_1(X, Y, [Z | L], R, Rs) when Z >= Y ->  %这里的时候是X<Y,也就是Z>=Y就是说这时X<Y<=Z,我们把最小X的放到R里面,而且Y,Z替换X,Y

  split_1(Y, Z, L, [X | R], Rs);

split_1(X, Y, [Z | L], R, Rs) when Z >= X ->  %这里的时候Z>=X,也就是X<=Z<Y,我们把最小的X放到R里面,而且Z替代X成了Z,Y

  split_1(Z, Y, L, [X | R], Rs);

split_1(X, Y, [Z | L], [], Rs) ->        %这里的时候Z<X,也就是Z<X<Y,我们把最小的Z放到R里面(R目前为空)

  split_1(X, Y, L, [Z], Rs);

split_1(X, Y, [Z | L], R, Rs) ->         %这里的时候Z<X,也就是Z<X<Y,我们把最小的Z放到最后的参数(R不为空的时候),调用split_1_1,为什么???

  split_1_1(X, Y, L, R, Rs, Z);

split_1(X, Y, [], R, Rs) ->            %当列表完成后调用下个函数rmerge1,这个后面再讲

  rmergel([[Y, X | R] | Rs], []).

 

WTF,这些到底在干什么,erlang又没有调试跟踪,又没说明,完全就蒙了,仔细研究下终于明白了这2个函数的意义,不得说写源码的真是大神啊~~~

通过上面的分析,我们知道了一个规律,每次都会比较3个数的大小,而且还会处理其中最小的数

X:下桩  Y:上桩, Z:目前list的第一个元素 R:经过排序了的list,Rs和S是split_1_1使用的变量

split_1这个函数的作用是把X,Y,Z中最小的放到R中,同时要保证这个数比R中现有的元素都大,

这个怎么保证呢,当Z>X(包括Z>X和Z>Y两种情况)的时候把直接X放进去R,

原因就是X一直小于Y,而且R里面的元素都比X小才放进去的,而且整个过程X和Y的值都是增加的,所以X肯定大于R中的任何一个

开始是R代表R中任何一个),假设Z0>Y0

  1. R0<X0<Y0<Z0  开始R0为空,比较成立
  2. R1<X1<Y1<Z1   这时R1=[X0|R0],X1=Y0,Y1=Z0,当Z1>Y1,比较还是成立
  3. R2<X2<Y2<Z2   这时R2=[X1|R1],X2=Y1,Y2=Z1,当Z2>Y2,比较还是成立
  4. 。。。。。。。

当Z>X的时候也一样,于是当Z>X或者Z>Y的时候,只要把X的值放到R中就行,R里面的元素越来越大,是排好序的(从大到小),于是上面绿色的注释的代码就能理解了

蓝色的注释代码当R为空, Z<X<Y,当然R<Z<X<Y,于是也能理解了

主要是褐色的代码模块当R不为空,我们知道R<X<Y,而且Z<X<Y,可是R里面的元素和Z不能确定,

于是我们知道了当前最小的是Z,可是Z不一定大于R的所有元素,上面的split_1函数的逻辑就不通了,然后把Z存入到最后一个参数进入split_1_1

我们来查看split_1_1

split_1_1(X, Y, [Z | L], R, Rs, S) when Z >= Y ->    %这时候X<Y<=Z,R<X, S<X,我们这里不管S(S不变)于是R<X<Y<=Z,按照上面逻辑,X存入R,Y,Z替换X,Y
    split_1_1(Y, Z, L, [X | R], Rs, S);
split_1_1(X, Y, [Z | L], R, Rs, S) when Z >= X ->    %这时候X<=Z<Y,R<X, S<X,我们这里不管S(S不变)于是R<X<=Z<Y,按照上面逻辑,X存入R,Z替换X
    split_1_1(Z, Y, L, [X | R], Rs, S);
split_1_1(X, Y, [Z | L], R, Rs, S) when S =< Z ->     %这时候S<=Z<X<Y,R<X,在这里我们知道Y>X>R,这里S,Z设置为X,Y,因为X,Y被重新设置,所以后面没有比较性
    split_1(S, Z, L, [], [[Y, X | R] | Rs]);       %于是我们把Y,X存入R(R里面的还是有序的),然后把R存入RS,清空R,返回到开始split_1的函数
split_1_1(X, Y, [Z | L], R, Rs, S) ->           %当S>Z一样
    split_1(Z, S, L, [], [[Y, X | R] | Rs]);
split_1_1(X, Y, [], R, Rs, S) ->    
    rmergel([[S], [Y, X | R] | Rs], []).

我们可以看到,紫色注释的代码,当S<=Z<X<Y,R<X我们知道最小的数是S,然后是Z,可是我们不能比较R里面的元素与这2个数的大小,

如果按照上面函数的逻辑,可以在弄个函数split_1_1_1,可这样函数不是闭环的,于是大神直接把肯定比R大的2个元素存入R(保证了R的有序),再回到split_1,这里真是太厉害了

 1 X:12,Y:13,Z:54,L:[32,1,4521,32,214,541,1,12,3],R:[],Rs:[]
 2 X:13,Y:54,Z:32,L:[1,4521,32,214,541,1,12,3],R:"\f",Rs:[]
 3 X:32,Y:54,Z:1,L:[4521,32,214,541,1,12,3],R:"\r\f",Rs:[]
 4 X:32,Y:54,Z:4521,L:[32,214,541,1,12,3],R:"\r\f",Rs:[],S:1
 5 X:54,Y:4521,Z:32,L:[214,541,1,12,3],R:" \r\f",Rs:[],S:1
 6 X:1,Y:32,Z:214,L:[541,1,12,3],R:[],Rs:[[4521,54,32,13,12]]
 7 X:32,Y:214,Z:541,L:[1,12,3],R:[1],Rs:[[4521,54,32,13,12]]
 8 X:214,Y:541,Z:1,L:[12,3],R:[32,1],Rs:[[4521,54,32,13,12]]
 9 X:214,Y:541,Z:12,L:[3],R:[32,1],Rs:[[4521,54,32,13,12]],S:1
10 X:1,Y:12,Z:3,L:[],R:[],Rs:[[541,214,32,1],[4521,54,32,13,12]]
11 Rs:[[12,3,1],[541,214,32,1],[4521,54,32,13,12]]
我们看个简单的例子执行过程,大概就能明白这个逻辑了。
这里的List = [12,13,54,32,1,4521,32,214,541,1,12,3],这2个函数执行完成后的结果是[[12,3,1],[541,214,32,1],[4521,54,32,13,12]]
可以看到这里经过了N次循环(N是List长度),生成了几个子list,每个子list都是有序的,这样肯定没有完成,剩下的就是mergel和rmergel函数的作用了

 篇幅太长,不好排版,下面的函数分析放

erlang下lists模块sort(排序)方法源码解析(二)

未完待续。。。

posted @ 2018-05-22 17:36  土豆008  阅读(1531)  评论(0编辑  收藏  举报