kmp算法

kmp 算法

kmp算法要点在于在dfa上模拟运行输入，若到达接受状态，则接受并返回匹配位置，否则返回不匹配。
其dfa的模拟代码如下

int kmp(string pat,string txt)
{
    int n=txt.size();
    int m=txt.size();
    int i,j; // j is current state,txt[i] is current input
    auto dfa=buildDFA(string pat)
    for(j=0,i=0;i<n && j<m;i++)
    {
        j=dfa[j][txt[i]];
    }
    if(j==m) return i-m; // if reachs accept state then accept 
    else return n; // refuse
}

其dfa的构造代码如下：
+ Match: if state j and next char c==pat[j],goto j+1
+ Mismatch: if next char c !=pat[j],simulate pat[1,j-1] on dfa,if maintain state X,which is simulate pat[1,j-1] on dfa will get X.

vector<vector<int>> buildDFA(string pat)
{
    int m=pat.size();
    const int R=256;
    vector<vector<int> > dfa(m,vector<int>(R));
    dfa[0][pat[0]]=1;
    // X simulate pat[1,j-1] on dfa
    // X0=0
    // Xj=dfa[Xj-1][txt[j-1]]
    for(int X=0,j=1;j<m;j++)
    {
        // copy transition function at state X to state j
        copy(dfa[X].begin(),dfa[X].end(),dfa[j].begin)；
        dfa[j][pat[j]]=j+1;
        X=dfa[X][j]; //update state X
    }
}
代码合并如下
``` C++
int kmp(string pat,string txt)
{
  int n=txt.size();
  int m=pat.size();
  int i,j;
  const int R=256;
  //dfa dfa[state][input]
  // construct dfa
  vector<vector<int> > dfa(m,vector<int>(R));
  dfa[0][pat[0]]=1;
  for(int X=0,j=1;j<m;j++)
  {
    copy(dfa[X].begin(),dfa[X].end(),dfa[j].begin());
    dfa[j][pat[j]]=j+1;
    X=dfa[X][pat[j]];
  }
  // simulate dfa
  // txt[i] is current input
  // dfa[i][j]: j is state,i is input
  for(i=0,j=0;i<n&&j<m;i++)
    j=dfa[j][txt[i]];
  if(j==m) return i-m; // m is final state
  else return n;
}