智能优化 之 下山单纯形法 C++

单纯形法简介在其他网站上都可以查到,我就不多说了

我们主要说方法

它主要解决的是局部最优解的问题

利用多边形进行求解的,若有n个变量,则利用n+1边形

我们这里以两个变量为例,求解第三维度的最优解

例如解决

min  f(x,y)=x2 - 4*x + y2 - y - x*y

matlab 图

智能优化 之 下山单纯形法 C++

可以看出,差不多是(3,2)附近取得最小

我们来用下山单纯形求解

我们设立三个初始点 (0,0),(1.2,0),(0,0.8)

我们把它们分别带入f中,函数值越小的越接近解,我们把它称为最好点,反之,函数值最大的点,我们称之为最坏点

我们要做的是,利用已知点,寻找更加接近解的点

我们需要了解几种寻找下一个点的思想

反射 reflect

智能优化 之 下山单纯形法 C++

假设三角形的三个点是ABP,其中P是最坏点,那么我们寻找一个Q点,使得APBQ是一个平行四边形

设向量α为p->A,β为p->B    (假设1)

那么Q = p + (α+β),其中p和Q是坐标

扩张 extern

智能优化 之 下山单纯形法 C++

假设,我们得到的新点Q,它比原来三角形中最好的点还要好,那么,我们可以假定这个探索方向是正确的,我们不妨再往前走一步!

其中Q->R = (p->R)/2,我们这里称扩张Q点

设向量α为Q->A, β为Q->B     (假设2)

于是,R =  Q - (α+β)/2

收缩 Shrink

我认为收缩有两种

因为我们一般先做反射点,所以,之后的操作如果针对反射点,那么就是对反射点进行收缩

智能优化 之 下山单纯形法 C++

基于(假设2),R = Q + (α+β)/4

还有一种是最优解本来就在三角形PAB中,我们对P做收缩

智能优化 之 下山单纯形法 C++

基于(假设1),则Q = P + (α+β)/4

压缩 compress

智能优化 之 下山单纯形法 C++

我们认为,如果上述操作均没有找到更好的点来替代最坏点,那么说明之前的三角形是非法的,那么我们进行压缩操作

即,取两边中点与最坏点构成新的三角形

我们用下山单纯形法求解步骤如下:

求出初始点的最坏点,构成三角形

重复下述,直到满足精度

先做一次反射

  如果反射点比最好点还要好(更加接近条件:min f(x0,y0))->做一次扩张

    如果扩张点比反射点还要好->扩张点代替之前的最坏点,形成新的三角形

    反之->反射点代替之前的最坏点

  反之,如果反射点比最坏点还要坏->反射点做收缩1

    如果收缩点1比最坏点好->收缩点1代替最坏点

    反之->最坏点做收缩2

      如果收缩点2比最坏点好->收缩点2代替最坏点

      反之->三角形做压缩

  反之,反射点代替最坏点,形成新的三角形

C++代码:

triangle.h

#pragma once

#define stds std::

#define VEC2_OUT

#include "lvgm\lvgm.h"    //本人博客: https://www.cnblogs.com/lv-anchoret/category/1367052.html
#include <vector>
#include <algorithm>
using namespace lvgm; class Mountain
{
public:
typedef dvec2 valtype; typedef double(*_Fun)(const valtype&); Mountain() { } /*
p: three position coordinates(in ordered or not)
f: the function Ptr
δ: the solution precision
*/
Mountain(const valtype& p1, const valtype& p2, const valtype& p3, const double δ)
: _δ(δ)
{
_positions.resize();
_positions[] = p1;
_positions[] = p2;
_positions[] = p3;
sort();
} static void setF(_Fun f)
{
_f = f;
} void setδ(double delt)
{
_δ = delt;
} public: /*
origion: the bad position
vec1: bad position -> min position
vec2: bad position -> mid position
*/
valtype reflect(const valtype& origion, const valtype& vec1, const valtype& vec2)
{
return origion + (vec1 + vec2);
} /*
origion: the change position
vec1: change position -> left position
vec2: change position -> right position
*/
valtype shrink(const valtype& origion, const valtype& vec1, const valtype& vec2)
{
return origion + (vec1 + vec2) / ;
} /*
origion: the origion position
vec1: origion position -> left position
vec2: origion position -> right position
*/
void compression(const valtype& origion, const valtype& vec1, const valtype& vec2)
{
_positions[] = origion + min(vec1, vec2) / ;
_positions[] = origion + max(vec1, vec2) / ;
} /*
origion: the change position
vec1: change position -> left position
vec2: change position -> right position
*/
valtype exter(const valtype& origion, const valtype& vec1, const valtype& vec2)
{
return origion - (vec1 + vec2) / ;
} void go()
{
double delt = (_positions[] - _positions[]).normal();
static int i = ;
while (delt > _δ)
{
stds cout << ++i << "次 " << _positions[] << "\t" << _positions[] << "\t" << _positions[] << stds endl;
valtype t = reflect(_positions[], _positions[] - _positions[], _positions[] - _positions[]);
if (_f(t) < _f(_positions[]))
{
valtype ex = exter(t, _positions[] - t, _positions[] - t);
if (_f(ex) < _f(t))
_positions[] = ex;
else
_positions[] = t;
}
else if (_f(t) > _f(_positions[]))
{
valtype sh = shrink(t, _positions[] - t, _positions[] - t);
if (_f(sh) < _f(_positions[])) //反射点收缩
_positions[] = sh;
else //三角内部内缩
{
sh = reflect(sh, _positions[] - sh, _positions[] - sh);
if (_f(sh) < _f(_positions[]))
_positions[] = sh;
else //针对原始点内缩,针对反射点收缩,都不管用,那么选择压缩
compression(_positions[0], _positions[] - _positions[], _positions[] - _positions[]);
}
}
else
_positions[] = t;
sort();
delt = (_positions[] - _positions[]).normal();
}
stds cout << "\n最好点为" << _positions[] << "\t精度为:" << _δ << stds endl << "函数值为:" << _f(_positions[]) << stds endl << stds endl;
} protected: const valtype& min(const valtype& vec1, const valtype& vec2)
{
return _f(vec1) < _f(vec2) ? vec1 : vec2;
} const valtype& max(const valtype& vec1, const valtype& vec2)
{
return _f(vec1) > _f(vec2) ? vec1 : vec2;
} friend bool cmp(const valtype& pos1, const valtype& pos2)
{
return Mountain::_f(pos1) < Mountain::_f(pos2);
} void sort()
{
stds sort(_positions.begin(), _positions.end(), cmp);
} private: stds vector<valtype> _positions; //min, mid, max or good, mid, bad double _δ; static _Fun _f;
};

main.cpp(原错误版本

#include "triangle.h"

Mountain::_Fun Mountain::_f=[](const Mountain::valtype& v)->double    {    return .;    };

int main()
{
auto fun = [](const Mountain::valtype& v)->double
{
return v.x()*v.x() - * v.x() + v.y()*v.y() - v.y() - v.x()*v.y();
}; Mountain m(Mountain::valtype(, ), Mountain::valtype(1.2, ), Mountain::valtype(, 0.8), 0.1); m.setF(fun); m.go(); m.setδ(0.01); m.go(); m.setδ(0.001); m.go(); m.setδ(0.0001); m.go(); m.setδ(0.00001); m.go(); }

error:错在初始化的时候写了一个默认函数,创建对象之后才进行setF设置内部函数,导致第一个三角形在构造函数中第一次sort的时候,并没有正确排序

我们稍微改动一下:

main.cpp

#include "triangle.h"

Mountain::_Fun Mountain::_f
{ [](const Mountain::valtype& v)->double { return v.x()*v.x() - * v.x() + v.y()*v.y() - v.y() - v.x()*v.y(); } }; int main()
{
Mountain m(Mountain::valtype(, ), Mountain::valtype(1.2, ), Mountain::valtype(, 0.8), 0.1); m.go(); m.setδ(0.01); m.go(); m.setδ(0.001); m.go(); m.setδ(0.0001); m.go(); m.setδ(0.00001); m.go(); m.setδ(0.000001); m.go();
}

这样我们收敛地更好:

结果:

迭代次数    good    medium        bad
1次 [ 1.2, ] [ , 0.8 ] [ , ]
2次 [ 1.8, 1.2 ] [ 1.2, ] [ , 0.8 ]
3次 [ 1.8, 1.2 ] [ , 0.4 ] [ 1.2, ]
4次 [ 3.6, 1.6 ] [ 1.8, 1.2 ] [ , 0.4 ]
5次 [ 3.6, 1.6 ] [ 2.4, 2.4 ] [ 1.8, 1.2 ]
6次 [ 3.6, 1.6 ] [ 2.4, 2.4 ] [ 4.2, 2.8 ]
7次 [ 2.4, 1.6 ] [ 3.6, 1.6 ] [ 2.4, 2.4 ]
8次 [ 2.7, ] [ 2.4, 1.6 ] [ 3.6, 1.6 ]
9次 [ 2.7, ] [ 3.075, 1.7 ] [ 2.4, 1.6 ]
10次 [ 2.7, ] [ 3.375, 2.1 ] [ 3.075, 1.7 ]
11次 [ 3.01875, 2.225 ] [ 2.7, ] [ 3.375, 2.1 ]
12次 [ 3.11719, 2.10625 ] [ 3.01875, 2.225 ] [ 2.7, ]
13次 [ 3.11719, 2.10625 ] [ 3.01875, 2.225 ] [ 3.25195, 2.24844 ]
14次 [ 3.11719, 2.10625 ] [ 2.88398, 2.08281 ] [ 3.01875, 2.225 ]
15次 [ 2.98242, 1.96406 ] [ 3.11719, 2.10625 ] [ 2.88398, 2.08281 ]
16次 [ 2.98242, 1.96406 ] [ 3.11719, 2.10625 ] [ 3.13271, 2.01133 ]
17次 [ 2.98242, 1.96406 ] [ 2.96689, 2.05898 ] [ 3.11719, 2.10625 ] 最好点为[ 2.98242, 1.96406 ] 精度为:0.1
函数值为:-6.99903 18次 [ 2.98242, 1.96406 ] [ 2.96689, 2.05898 ] [ 2.90339, 1.96416 ]
19次 [ 2.98242, 1.96406 ] [ 3.04592, 2.05889 ] [ 2.96689, 2.05898 ]
20次 [ 2.98242, 1.96406 ] [ 3.03781, 1.98772 ] [ 3.04592, 2.05889 ]
21次 [ 3.02802, 2.01739 ] [ 2.98242, 1.96406 ] [ 3.03781, 1.98772 ]
22次 [ 3.02802, 2.01739 ] [ 2.97263, 1.99373 ] [ 2.98242, 1.96406 ]
23次 [ 3.00928, 2.02631 ] [ 3.02802, 2.01739 ] [ 2.97263, 1.99373 ]
24次 [ 2.99564, 2.00779 ] [ 3.00928, 2.02631 ] [ 3.02802, 2.01739 ]
25次 [ 2.99564, 2.00779 ] [ 3.00928, 2.02631 ] [ 2.98968, 2.01688 ]
26次 [ 2.99564, 2.00779 ] [ 3.01524, 2.01722 ] [ 3.00928, 2.02631 ]
27次 [ 3.0016, 1.9987 ] [ 2.99564, 2.00779 ] [ 3.01524, 2.01722 ]
28次 [ 3.0016, 1.9987 ] [ 2.99564, 2.00779 ] [ 2.982, 1.98927 ]
29次 [ 3.0016, 1.9987 ] [ 3.00693, 2.01023 ] [ 2.99564, 2.00779 ]
30次 [ 3.0016, 1.9987 ] [ 3.00858, 2.0028 ] [ 3.00693, 2.01023 ] 最好点为[ 3.0016, 1.9987 ] 精度为:0.01
函数值为:-6.99999 31次 [ 3.0016, 1.9987 ] [ 3.00417, 1.99601 ] [ 3.00858, 2.0028 ]
32次 [ 3.0016, 1.9987 ] [ 3.00417, 1.99601 ] [ 2.99719, 1.99191 ]
33次 [ 3.0016, 1.9987 ] [ 3.00573, 2.00008 ] [ 3.00417, 1.99601 ]
34次 [ 3.0016, 1.9987 ] [ 3.00316, 2.00277 ] [ 3.00573, 2.00008 ]
35次 [ 2.99903, 2.00139 ] [ 3.0016, 1.9987 ] [ 3.00316, 2.00277 ]
36次 [ 2.99903, 2.00139 ] [ 3.0016, 1.9987 ] [ 2.99747, 1.99732 ]
37次 [ 3.00174, 2.00141 ] [ 2.99903, 2.00139 ] [ 3.0016, 1.9987 ]
38次 [ 3.00099, 2.00005 ] [ 3.00174, 2.00141 ] [ 2.99903, 2.00139 ]
39次 [ 3.00099, 2.00005 ] [ 3.0002, 2.00106 ] [ 3.00174, 2.00141 ]
40次 [ 2.99945, 1.9997 ] [ 3.00099, 2.00005 ] [ 3.0002, 2.00106 ]
41次 [ 2.99945, 1.9997 ] [ 3.00024, 1.99928 ] [ 3.00099, 2.00005 ] 最好点为[ 2.99945, 1.9997 ] 精度为:0.001
函数值为:- 42次 [ 2.99945, 1.9997 ] [ 2.99927, 1.99921 ] [ 3.00024, 1.99928 ]
43次 [ 2.99945, 1.9997 ] [ 2.9998, 1.99937 ] [ 2.99927, 1.99921 ]
44次 [ 2.99998, 1.99986 ] [ 2.99945, 1.9997 ] [ 2.9998, 1.99937 ]
45次 [ 2.99998, 1.99986 ] [ 2.99945, 1.9997 ] [ 2.99964, 2.00019 ]
46次 [ 2.99998, 1.99986 ] [ 2.99976, 1.99957 ] [ 2.99945, 1.9997 ]
47次 [ 2.99998, 1.99986 ] [ 3.00008, 1.99972 ] [ 2.99976, 1.99957 ]
48次 [ 2.99998, 1.99986 ] [ 3.0003, 2.00001 ] [ 3.00008, 1.99972 ]
49次 [ 2.99998, 1.99986 ] [ 3.00021, 2.00014 ] [ 3.0003, 2.00001 ]
50次 [ 2.99989, 1.99999 ] [ 2.99998, 1.99986 ] [ 3.00021, 2.00014 ]
51次 [ 3.00007, 2.00003 ] [ 2.99989, 1.99999 ] [ 2.99998, 1.99986 ]
52次 [ 3.00007, 2.00003 ] [ 2.99998, 2.00009 ] [ 2.99989, 1.99999 ]
53次 [ 3.00007, 2.00003 ] [ 3.00009, 2.0001 ] [ 2.99998, 2.00009 ] 最好点为[ 3.00007, 2.00003 ] 精度为:0.0001
函数值为:- 54次 [ 3.00007, 2.00003 ] [ 3.00003, 2.00008 ] [ 3.00009, 2.0001 ]
55次 [ 3.00001, 2.00001 ] [ 3.00007, 2.00003 ] [ 3.00003, 2.00008 ]
56次 [ 3.00001, 2.00001 ] [ 3.00004, ] [ 3.00007, 2.00003 ]
57次 [ 3.00001, 2.00001 ] [ 2.99998, 1.99998 ] [ 3.00004, ]
58次 [ 3.00001, 2.00001 ] [ 2.99998, 1.99998 ] [ 2.99997, 1.99999 ]
59次 [ 3.00001, 2.00001 ] [ 3.00002, ] [ 2.99998, 1.99998 ]
60次 [ , 1.99999 ] [ 3.00001, 2.00001 ] [ 3.00002, ]
61次 [ , 1.99999 ] [ 3.00001, 2.00001 ] [ 2.99998, 2.00001 ]
62次 [ , 1.99999 ] [ 3.00001, ] [ 3.00001, 2.00001 ] 最好点为[ 3.00001, ] 精度为:1e-
函数值为:- 63次 [ 3.00001, ] [ , 1.99999 ] [ 3.00001, ]
64次 [ 3.00001, ] [ 2.99999, ] [ , 1.99999 ]
65次 [ 3.00001, ] [ , ] [ 2.99999, ]
66次 [ , ] [ 3.00001, ] [ , ]
67次 [ , ] [ , ] [ 3.00001, ]
68次 [ , ] [ , ] [ , ]
69次 [ , ] [ , ] [ , ]
70次 [ , ] [ , ] [ , ]
71次 [ , ] [ , ] [ , ]
72次 [ , ] [ , ] [ , ]
73次 [ , ] [ , ] [ , ]
74次 [ , ] [ , ] [ , ] 最好点为[ , ] 精度为:1e-
函数值为:-

之前错误版本结果如下

结果:

1次  [ ,  ]      [ 1.2,  ]     [ , 0.8 ]
2次 [ 1.2, ] [ 1.2, -0.8 ] [ , ]
3次 [ 1.2, ] [ 1.2, -0.8 ] [ 2.4, -0.8 ]
4次 [ 1.2, ] [ 0.6, -0.2 ] [ 1.2, -0.8 ]
5次 [ 1.2, ] [ 0.6, 0.6 ] [ 0.6, -0.2 ]
6次 [ 1.5, 1.3 ] [ 1.2, ] [ 0.6, 0.6 ]
7次 [ 2.1, 0.7 ] [ 1.5, 1.3 ] [ 1.2, ]
8次 [ 2.4, ] [ 2.1, 0.7 ] [ 1.5, 1.3 ]
9次 [ 2.4, ] [ , 1.4 ] [ 2.1, 0.7 ]
10次 [ 2.4, ] [ , 1.4 ] [ 3.3, 2.7 ]
11次 [ , 2.2 ] [ 2.4, ] [ , 1.4 ]
12次 [ , 2.2 ] [ 2.85, 1.75 ] [ 2.4, ]
13次 [ , 2.2 ] [ 2.85, 1.75 ] [ 3.45, 1.95 ]
14次 [ , 2.2 ] [ 2.85, 1.75 ] [ 2.6625, 1.9875 ]
15次 [ , 2.2 ] [ 3.1875, 1.9625 ] [ 2.85, 1.75 ]
16次 [ 2.97188, 1.91563 ] [ , 2.2 ] [ 3.1875, 1.9625 ]
17次 [ 2.97188, 1.91563 ] [ 2.88516, 2.10547 ] [ , 2.2 ]
18次 [ 2.97188, 1.91563 ] [ 2.85703, 1.82109 ] [ 2.88516, 2.10547 ]
19次 [ 2.97188, 1.91563 ] [ 2.8998, 1.98691 ] [ 2.85703, 1.82109 ]
20次 [ 2.97188, 1.91563 ] [ 3.01465, 2.08145 ] [ 2.8998, 1.98691 ]
21次 [ 2.97188, 1.91563 ] [ 3.01465, 2.08145 ] [ 3.08672, 2.01016 ]
22次 [ 2.94653, 1.99272 ] [ 2.97188, 1.91563 ] [ 3.01465, 2.08145 ]
23次 [ 2.98693, 2.01781 ] [ 2.94653, 1.99272 ] [ 2.97188, 1.91563 ] 最好点为[ 2.98693, 2.01781 ] 精度为:0.1
函数值为:-6.99928 24次 [ 2.98693, 2.01781 ] [ 2.9693, 1.96045 ] [ 2.94653, 1.99272 ]
25次 [ 3.0097, 1.98553 ] [ 2.98693, 2.01781 ] [ 2.9693, 1.96045 ]
26次 [ 3.01282, 2.02228 ] [ 3.0097, 1.98553 ] [ 2.98693, 2.01781 ]
27次 [ 3.01282, 2.02228 ] [ 3.0097, 1.98553 ] [ 3.02342, 1.99696 ]
28次 [ 2.99909, 2.01086 ] [ 3.01282, 2.02228 ] [ 3.0097, 1.98553 ]
29次 [ 3.00782, 2.00105 ] [ 2.99909, 2.01086 ] [ 3.01282, 2.02228 ]
30次 [ 3.00782, 2.00105 ] [ 2.9941, 1.98963 ] [ 2.99909, 2.01086 ]
31次 [ 3.00003, 2.0031 ] [ 3.00782, 2.00105 ] [ 2.9941, 1.98963 ]
32次 [ 3.00003, 2.0031 ] [ 3.00782, 2.00105 ] [ 3.00884, 2.0083 ] 最好点为[ 3.00003, 2.0031 ] 精度为:0.01
函数值为:-6.99999 33次 [ 3.00003, 2.0031 ] [ 2.99901, 1.99585 ] [ 3.00782, 2.00105 ]
34次 [ 3.00003, 2.0031 ] [ 2.99901, 1.99585 ] [ 2.99537, 1.99869 ]
35次 [ 3.00003, 2.0031 ] [ 3.00367, 2.00026 ] [ 2.99901, 1.99585 ]
36次 [ 3.00043, 1.99877 ] [ 3.00003, 2.0031 ] [ 3.00367, 2.00026 ]
37次 [ 3.00043, 1.99877 ] [ 2.99851, 2.00127 ] [ 3.00003, 2.0031 ]
38次 [ 3.00043, 1.99877 ] [ 2.99851, 2.00127 ] [ 2.99891, 1.99693 ]
39次 [ 3.00043, 1.99877 ] [ 2.99975, 2.00156 ] [ 2.99851, 2.00127 ]
40次 [ 3.00043, 1.99877 ] [ 2.99975, 2.00156 ] [ 3.00167, 1.99906 ] 最好点为[ 2.9993, 2.00071 ] 精度为:0.001
函数值为:- 41次 [ 2.9993, 2.00071 ] [ 3.00043, 1.99877 ] [ 2.99975, 2.00156 ]
42次 [ 2.99992, 1.99883 ] [ 2.9993, 2.00071 ] [ 3.00043, 1.99877 ]
43次 [ 2.9992, 2.00028 ] [ 2.99992, 1.99883 ] [ 2.9993, 2.00071 ]
44次 [ 2.99969, 1.99897 ] [ 2.9992, 2.00028 ] [ 2.99992, 1.99883 ]
45次 [ 2.99921, 2.00002 ] [ 2.99969, 1.99897 ] [ 2.9992, 2.00028 ]
46次 [ 2.99958, 1.99911 ] [ 2.99921, 2.00002 ] [ 2.99969, 1.99897 ]
47次 [ 2.99924, 1.99986 ] [ 2.99958, 1.99911 ] [ 2.99921, 2.00002 ]
48次 [ 2.99951, 1.99922 ] [ 2.99924, 1.99986 ] [ 2.99958, 1.99911 ]
49次 [ 2.99928, 1.99975 ] [ 2.99951, 1.99922 ] [ 2.99924, 1.99986 ] 最好点为[ 2.99947, 1.9993 ] 精度为:0.0001
函数值为:- 50次 [ 2.99947, 1.9993 ] [ 2.99928, 1.99975 ] [ 2.99951, 1.99922 ]
51次 [ 2.9993, 1.99968 ] [ 2.99947, 1.9993 ] [ 2.99928, 1.99975 ]
52次 [ 2.9993, 1.99968 ] [ 2.99944, 1.99936 ] [ 2.99947, 1.9993 ]
53次 [ 2.99932, 1.99963 ] [ 2.9993, 1.99968 ] [ 2.99944, 1.99936 ]
54次 [ 2.99938, 1.9995 ] [ 2.99932, 1.99963 ] [ 2.9993, 1.99968 ]
55次 [ 2.99938, 1.9995 ] [ 2.9994, 1.99945 ] [ 2.99932, 1.99963 ]
56次 [ 2.99938, 1.9995 ] [ 2.99936, 1.99955 ] [ 2.9994, 1.99945 ]
57次 [ 2.99938, 1.9995 ] [ 2.99936, 1.99955 ] [ 2.99935, 1.99957 ]
58次 [ 2.99938, 1.9995 ] [ 2.99938, 1.99949 ] [ 2.99936, 1.99955 ]
59次 [ 2.99938, 1.9995 ] [ 2.99937, 1.99953 ] [ 2.99938, 1.99949 ]
60次 [ 2.99938, 1.9995 ] [ 2.99936, 1.99954 ] [ 2.99937, 1.99953 ]
61次 [ 2.99937, 1.99951 ] [ 2.99938, 1.9995 ] [ 2.99936, 1.99954 ]
62次 [ 2.99937, 1.99951 ] [ 2.99938, 1.99949 ] [ 2.99938, 1.9995 ]
63次 [ 2.99938, 1.9995 ] [ 2.99937, 1.99951 ] [ 2.99938, 1.99949 ]
64次 [ 2.99937, 1.99954 ] [ 2.99938, 1.9995 ] [ 2.99937, 1.99951 ]
65次 [ 2.99938, 1.99954 ] [ 2.99937, 1.99954 ] [ 2.99938, 1.9995 ] 最好点为[ 2.99938, 1.99954 ] 精度为:1e-
函数值为:-

感谢您的阅读,生活愉快~

上一篇:Tomjson - 一个"短小精悍"的 json 解析库


下一篇:GTK+中的构件II(Widgets)