1,Boost -> Bind

<1>auto ，initializer_list<T>,auto指向函数指针的简易，和typdef 定义的类型执行函数指针有多复杂。

#include <iostream>
#include <initializer_list>
#include <string>
using namespace std;

template <typename  T>
T sum(initializer_list<T> rh)
{
    T val;
    for(auto p= rh.begin();p!=rh.end();p++)
    {
        val+= *p;
    }
    return val;
}

int main()
{
    // use init list
    cout << sum({,,,,}) <<endl;
    cout << sum({1.0,2.0,3.1,4.0,5.0}) <<endl;

    //
    cout << "use auto to point the function sum" <<endl;
    auto dadd_func = sum<double>;
    auto iadd_func = sum<int>;
    auto tadd_func = sum<string>;
    cout << dadd_func({,,,}) <<endl;
    cout << iadd_func({,,,,}) <<endl;
    cout << tadd_func({"houdini","maya","json"}) <<endl;

    cout << "\nuse the typedef to pointer\n";
    typedef int     (*td_int_sum)(initializer_list<int> rh);
    typedef string  (*td_str_sum)(initializer_list<string> rh);
    td_int_sum int_add = sum<int>;
    td_str_sum str_add = sum<string>;
    cout << int_add({,,,,,}) <<endl;
    cout << str_add({"s1","s2","s5"}) << endl;

    return ;
}

<2>funcional,std::generate,std::count_if

#include <iostream>
#include <math.h>
#include <vector>
#include <algorithm>
#include <functional>

using namespace std;

double square(double x){return x*x;}

int main()
{
    vector <int> vars();

    generate(vars.begin(),vars.end(),std::rand);
    for_each(vars.begin(),vars.end(),[](int v){cout << v <<endl;});

    // lambda can transfer local variable
    int sum = ;
    for_each(vars.begin(),vars.end(),[&sum](int v){sum+=v;});
    cout << "the sum is " << sum <<endl;

    // <100 num
    cout << "get <100 numbers" <<endl;
    cout << count_if(vars.begin(),vars.end(),[](int v){ return v<;}) <<endl;

    // functional
    function<double(double)> ef1 = square;
    cout << ef1() <<endl;      //

    function<void(int var)> ef2 = [](int val){cout << val <<endl;};
    ef2();                   //

    return ;
}

<3> remove_if,vector,min_element,max_element

include <iostream>
#include <algorithm>
#include <vector>
#include <string>
using namespace std;

void cppRemove_if()
{
    cout << "====cppRemove_if====\n";
    int myInts[]{,,,,,,,};        // 1 2 3 4 5 6 7 8
    int *pbegin = myInts;
    int *pend = myInts + sizeof(myInts)/ sizeof(int);
    pend = remove_if(pbegin,pend,
                     [](const int &val)->bool{return val% == ;});//2 4 5 8 ? ? ? ?
    for (int* p=pbegin; p!=pend; ++p)
        cout << ' ' << *p;
    cout << "\n";
    cout << "====cppRemove_if====\n";

}

int main()
{

    vector<int> va{,,,,,};

    // find, if not find elements,will return last *iter;
    auto va_find2  = find(va.begin(),va.end(),);
    auto va_find2e = find_if(va.begin(),va.end(),[](const int &x){return x==;});
    cout << *va_find2 <<endl;
    cout << *va_find2e <<endl;

    cout << *min_element(va.begin(),va.end()) <<endl;
    cout << *max_element(va.begin(),va.end()) <<endl;

    auto min_max = minmax_element(va.begin(),va.end());
    cout << "min val:" <<*(min_max.first)<<endl;
    cout << "max val:" <<*(min_max.second)<<endl;

    cout << "remove the second elements \n";
    va.erase(va.begin()+,va.begin()+);
    for_each(va.begin(),va.end(),[](const int &x){cout << x <<endl;});

    cout << "remove by condition <5 \n";
    va.erase(remove_if(va.begin(),va.end(),[](int x){return x <;}),va.end());
    for_each(va.begin(),va.end(),[](const int &x){cout << x <<endl;});

    cppRemove_if();
    return ;
}

<4>binary_search,sort更加详细的用法:

#include <iostream>
#include <algorithm>
#include <vector>
#include <functional>
using namespace std;

template <typename T1>
void ShowIntArray(const T1 begin, const T1 end)
{
    for_each(begin,end,[](const int &x){cout << x <<" "; });
    cout << "\n";
}
template <typename T>
void ShowSTLArray(const T&cont)
{
    auto iter = cont.begin();
    auto end = cont.end();
    for(;iter!=end;iter++)
    {
        cout << *iter <<" ";
    }
    cout <<endl;
};

void cpp_sort()
{
    int a[]= {,,,,,};
    int *begin = a;
    int *end = a + ;

    cout << "before sort:\n";
    ShowIntArray(begin, end);

    sort(begin,end);
    cout << "after sort:\n";
    ShowIntArray(begin, end);

    cout << "from large to small:\n";
    sort(begin,end,[](const int &x,const int &y){return x>y;});
    ShowIntArray(begin, end);

    cout << "from small to large use less<int>():\n";
    sort(begin,end,less<int>());
    ShowIntArray(begin, end);

    cout << "from large to small use greater<int>():\n";
    sort(begin,end,greater<int>());
    ShowIntArray(begin, end);

    vector<string> vecStr{"Got","cool","features"};
    cout << "sort the sting array:\n";
    ShowSTLArray(vecStr);
    auto strCmp = [](string &a,string &b)
    {
        return a.length() > b.length();
    };
    cout << "sort the array results:\n";
    sort(vecStr.begin(),vecStr.end(),strCmp);
    ShowSTLArray(vecStr);

}

void cpp_binary_search()
{

    cout << "=======search 01:==========\n";
    std::vector<int> haystack {, , , , };
    std::vector<int> needles {, , };
    sort(haystack.begin(),haystack.end());
    for (auto needle : needles)
    {
        cout << "Searching for " << needle << '\n';
        if (binary_search(haystack.begin(), haystack.end(), needle))
        {

            cout << "Found " << needle << '\n';
        }
        else
        {
            cout << "no dice!\n";
        }
    }
    cout << "=======search 01:==========\n";

    std::vector<int> haystack2 {, , , , ,,,};
    sort(haystack2.begin(),haystack2.end(),[](int &x,int &y){return x<y;});
    ShowSTLArray(haystack2);

    auto func =[](int i,int j)->bool{cout<< "i:" << i; cout << " j:"<<j;cout<<"\n";return (i<j);};
    if (binary_search(haystack2.begin(),haystack2.end(),,func))
    {
        cout << "found 5" <<endl;
    }

}

int main()
{
    //cpp_sort();
    cpp_binary_search();
    return ;
}

binarySearch结果:

Sort结果:

 <5> 线程大法

(1) hello world thread:

#include <iostream>
#include <thread>
#include <memory>
using namespace std;

void thread_task()
{
    cout << "thread hello world\n";
}

int main()
{
    shared_ptr<thread> t(new thread(thread_task));
    t->join();

    return ;
}

(2)带参数的函数(bind方法，直接使用thread构造也可以)

void thread_parm(const int &n,const string& name)
{
    for(int i=;i<n;i++)
    {
        cout << name <<":thread loop in " << i <<endl;
    }
}

void withParam()
{
    thread t0(thread_parm,,"houdini");
    thread t1(bind(thread_parm,,"maya"));
    t0.join();
    t1.join();
}

int main()
{
    withParam();
    return ;
}

(3)成员对象函数执行在线程中（也可以作用到智能指针对象）

class HelloObject
{
public:
    void sayHello(const string& name,int n)
    {
        for(int i=;i<n;i++)
        {
            cout << name << " thread: " << i <<endl;
        }
    }
};

void objectFunction()
{
    HelloObject obj;
    thread t(&HelloObject::sayHello,&obj,"Json",);
    t.join();

    // work with shared_ptr
    shared_ptr<HelloObject> objPtr(new HelloObject());
    thread tptr(&HelloObject::sayHello,objPtr,"Houdini",);
    tptr.join();
}

int main()
{
    objectFunction();
    return ;
}

(4)传递引用，头文件functional，std::ref()

class FuncObj
{
public:
    void operator()()const
    {
        cout << this <<endl;
    }
};
void passRef()
{
    auto obj = FuncObj();
    obj();

    //pass by value
    cout << "thread will pass by value\n";
    thread t1(obj);
    t1.join();

    //pass by ref
    cout << "thread will pass by ref\n";
    thread t2(ref(obj));
    t2.join();

}

int main()
{
    passRef();
    return ;
}

结果:

0x22fdff
thread will pass by value
0x7c6150
thread will pass by ref
0x22fdff

普通的函数也可以

void increment(int &value)
{
    value++ ;
    cout << "value :" << value <<endl;

}
void passRef2()
{
    int a = ;
    thread t(increment,ref(a));
    t.join();
}

int main()
{
    passRef2();
    return ;
}

(5)基本功能:

匿名函数:get_id() 区分线程

void lambdaTest()
{
    vector <thread> threads;
    for(int i=;i<;i++)
    {
        threads.emplace_back(thread([](){cout << "thread id " << this_thread::get_id() << endl;}));
    }

    for(auto &t : threads)
    {
        t.join();
    }
}

int main()
{
    lambdaTest();
    return ;
}

总线程数:

cout << thread::hardware_concurrency() <<endl;

(6)异常与线程

标准处理方法

struct Counter2
{
    int value;
    Counter2():value(){}
    void increment()
    {
        ++value;
    }
    void decrement()
    {
        if(value == )
        {
            throw string("value cannot be less than 0");
        }

        --value;
    }
};

struct Wrapper
{
    Counter2 ct;
    mutex m;

    void increment()
    {

    }
    void decrement()
    {
        m.lock();
        try
        {
            ct.decrement();
        }
        catch (const string &e)
        {
            m.unlock();
            cout << e <<endl;
            throw e;
        }

        m.unlock();
    }
};

void exceptionLock()
{
    Wrapper wap;
    wap.ct.value = ;

    vector<thread> threads;
    for(int i=;i<;i++)
    {
        threads.emplace_back(thread([&wap](){
            wap.decrement();

        }));
    }
    for(auto &t:threads)
    {
        t.join();
    }
    cout << wap.ct.value << endl;
}

(7)模仿Inter TBB parallel_for

串行时间:87

并行时间:19

struct BlockRange
{
    BlockRange():begin(),end()
    {

    }
    int begin;
    int end;
};

class ApplyFoo
{
public:
    ApplyFoo(vector<int> *data):mData(data)
    {
    }
    void operator()(const BlockRange &range)const
    {
        for(int i=range.begin;i<range.end;i++)
        {
            (*mData)[i] += ;
        }
    }
private:
    vector<int> *mData;
};

template <typename T>
void parallel_for(const T &body,int size,int begin)
{
    auto nThreads = thread::hardware_concurrency();
    auto nValuesSize = size;
    auto perBlockSize =nValuesSize / nThreads;
    if(nValuesSize < nThreads)
    {
        BlockRange range;
        range.begin = begin;
        range.end = nValuesSize;
        body(range);
        return;
    }
    // building blocks
    vector<BlockRange> blocks;
    int index = begin;
    while(index <= nValuesSize)
    {
        BlockRange range;
        range.begin = index;
        range.end   = index+ perBlockSize;
        blocks.push_back(range);
        index += (perBlockSize) ;
    }
    // fix last block end size;
    blocks[blocks.size()-].end = nValuesSize;
    // thread pool to run
    typedef shared_ptr<thread> thread_ptr;
    vector<thread_ptr> pools;
    for(BlockRange&r:blocks)
    {
        pools.emplace_back(new thread(body,r));
    }
    for(auto &t:pools)
    {
        t->join();
    }

}

void parallel()
{
    vector<int> values();
    fill(values.begin(),values.end(),);

    double start,end,cost;
    start=clock();
    parallel_for(ApplyFoo(&values),values.size(),);
    end= clock();
    cost = end -start;
    cout << "parallel for cost time:" << cost <<endl;

}
void serial()
{
    vector<int> values();
    fill(values.begin(),values.end(),);

    double start,end,cost;
    start=clock();
    for(int i=;i<values.size();i++)
    {
        values[i] += ;
    }
    end= clock();
    cost = end -start;
    cout << "serial for cost time:" << cost <<endl;
}

int main()
{

    parallel();
    serial();
    return ;
}

并行accumulation:

10亿个元素相加:简直他妈的快飞起来了。

串行时间:13063

并行时间:1023

#include <vector>
#include <time.h>
#include <iostream>
#include <thread>
#include <algorithm>
using namespace std;

struct BlockRange
{
    BlockRange():begin(),end(),id()
    {

    }
    int begin;
    int end;
    int id;
};

class ApplyFoo
{
public:
    ApplyFoo(vector<int> *data):mData(data)
    {
    }
    void operator()(const BlockRange &range,vector<int> *des)const
    {
        auto value = int();
        for(int i=range.begin;i<range.end;i++)
        {
            value +=(*mData)[i];
        }
        (*des)[range.id] = value;
    }
private:
    vector<int> *mData;

};

template <typename retType,typename T>
retType parallel_add(const T &body,int size,int begin)
{
    vector<retType> partial_accum;
    auto nThreads = thread::hardware_concurrency();
    auto nValuesSize = size;
    auto perBlockSize =nValuesSize / nThreads;
    if(nValuesSize < nThreads)
    {
        partial_accum.resize();
        BlockRange range;
        range.begin = begin;
        range.end = nValuesSize;
        range.id = ;
        body(range,&partial_accum);
        return accumulate(partial_accum.begin(),partial_accum.end(),retType());
    }
    // building blocks
    vector<BlockRange> blocks;
    int index = begin;
    int blockId = ;
    while(index <= nValuesSize)
    {
        BlockRange range;
        range.begin = index;
        range.end   = index+ perBlockSize;
        range.id    = blockId;
        blocks.push_back(range);
        index += (perBlockSize) ;
        blockId += ;
    }
    partial_accum.resize(blocks.size());

    // fix last block end size;
    blocks[blocks.size()-].end = nValuesSize;
    // thread pool to run
    typedef shared_ptr<thread> thread_ptr;
    vector<thread_ptr> pools;

    for(BlockRange&r:blocks)
    {
        pools.emplace_back(new thread(body,r,&partial_accum));
    }
    for(auto &t:pools)
    {
        t->join();
    }

    return accumulate(partial_accum.begin(),partial_accum.end(),retType());

}

void parallel()
{
    vector<int> values();
    fill(values.begin(),values.end(),);

    double start,end,cost;
    start=clock();
    cout << "get the result :" <<parallel_add<int>(ApplyFoo(&values),values.size(),) <<endl;
    end= clock();
    cost = end -start;
    cout << "parallel for cost time:" << cost <<endl;
}

void serial()
{
    vector<int> values();
    fill(values.begin(),values.end(),);

    double start,end,cost;
    start=clock();
    cout << "get the result :" <<accumulate(values.begin(),values.end(),) <<endl;
    end= clock();
    cost = end -start;
    cout << "parallel for cost time:" << cost <<endl;
}

int main()
{

    parallel();
    //serial();
    return ;
}

<n> boost bind

#include <boost/bind.hpp>
#include <boost/shared_ptr.hpp>

#include <iostream>
using namespace std;
void dprint(int x,int y)
{
    cout << x << " " <<y <<endl;
}

class Bind_test
{
public:
    void setData(int x,int y)
    {
        _x = x;
        _y = y;
    }
    void printData()
    {
        cout << _x << " " <<_y <<endl;
    }
private:
    int _x;
    int _y;

};

void increnum(int &dg)
{
    dg++;
}

int main()
{
    boost::bind(&dprint,,)(); // 5,5
    boost::bind(&dprint,,_1)();  // 3, 5
    boost::bind(&dprint,_1,_1)();  // 2, 2
    boost::bind(&dprint,_1,_2)(,);  // 1, 2
    boost::bind(&dprint,_2,_1)(,);  // 2, 1 ->函数参数对掉

    cout << "\nbind the class function\n";
    boost::shared_ptr<Bind_test> bclass(new Bind_test);
    boost::bind(&Bind_test::setData,bclass,,)();
    bclass->printData();

    Bind_test *bclass_02  = new Bind_test;
    boost::bind(&Bind_test::setData,bclass_02,,)();
    bclass_02->printData(); // 2 ,3
    delete  bclass_02;

    Bind_test bclass_03;
    boost::bind(&Bind_test::setData,&bclass_03,,)();
    bclass_03.printData(); // 4 ,5
    boost::bind(&Bind_test::setData,&bclass_03,_1,_1)();
    bclass_03.printData(); // 9 ,9
    boost::bind(&Bind_test::setData,&bclass_03,_1,_2)(,);
    bclass_03.printData(); // 9 ,10

    int dgNum = ;
    boost::bind(&increnum,boost::ref(dgNum))(); // 类似C++11 Thread 里要传递引用std::ref(x)
    cout << dgNum <<endl;

    cin.get();
    return ;
}

额外的:

static_assert 编译时候assertions

下面将输出:hello \n no

cout << R"(hello \n no)" <<endl;

。