Question

我是编写C代码的初学者。但是，这是从FNN库中提取的，用于测试。

这是我的KNN_ANN.cpp代码：

#include <ANN.h> // ANN header
#include <R.h>       // R header

    using namespace std;



void Rvector2ANNarray(ANNpointArray data_pts,
                            double *data,
                            const int n, const int d)
      {
    //copy data point address into n x d matrix from d x n R vector column by column 
    for (int i = 0; i < n; i++) {
      data_pts[i] = data+i*d;
    }               
  } 


extern "C" {


  void get_KNNX_kd(double *data, double* query,
                   int *k, int *dim,
                   int *n_pts, int *m_pts,
                   int *nn_idx, double *nn_dist)
  {
    const int               d=*dim;     // Actual Dimension
    const int               n=*n_pts;   // Number of Base Data points
    const int               m=*m_pts;   // Number of Query Data points
    const int               K = *k;     // Max. num of NN   
    const double        error_bound = 0.0;

    ANNidxArray     index = new ANNidx[K];      // Near neighbor indices
    ANNdistArray    dist = new ANNdist[K];      // Near neighbor squared distance       
    ANNpointArray   data_pts  = new ANNpoint[n];
    ANNpointArray   query_pts  = new ANNpoint[m];

    //copy data into  matrix from a vector column by column
    Rvector2ANNarray(data_pts, data, n, d);
    Rvector2ANNarray(query_pts, query, m, d);

    ANNkd_tree  *kd_tree    = new ANNkd_tree(data_pts, n, d);   

    register int ptr = 0;   
    for(int i = 0; i < m; i++)  // read query points
    {
      kd_tree->annkSearch(  // search
                           query_pts[i],        // query point
                           K,               // number of near neighbors
                           index,               // nearest neighbors index
                           dist,                // squared distance
                           error_bound);        // error bound

      for (int j = 0; j < K; j++)      //return result row by row
      {
        nn_dist[ptr] = sqrt(dist[j]);   // unsquare distance
        nn_idx[ptr]  = index[j] + 1;    // put indexes in returned array
        ptr++;
      } // end inner for
    } // end for

    delete[] index;
    delete[] dist;  
    //  delete kd_tree;

    //  annDeallocPts(data_pts);
    //  annDeallocPts(query_pts);
    delete[] data_pts;
    delete[] query_pts;
    delete kd_tree;

    annClose();

  }//end get_KNNX_kd



}//end extern

我已将ANN.h和R.h放在正确的目录中： /Library/Frameworks/R.framework/Versions/3.5/Resources/include

然后我使用以下命令在终端中编译CPP代码以生成KNN_ANN.o和KNN_ANN.so： R CMD SHLIB KNN_ANN.cpp

当我回到R并致电

dyn.load("KNN_ANN.so")

我收到以下错误消息：

Error in dyn.load("KNN_ANN.so") :    unable to load shared object '/Users/ypeng/KNN_ANN.so':   dlopen(/Users/ypeng/KNN_ANN.so, 6):
找不到符号：__ Z8annClosev引用自： /Users/ypeng/KNN_ANN.so预期输入位置： /Users/ypeng/KNN_ANN.so

任何人，如果我的CPP代码有问题，请告诉我。.

编辑：这是我提到的基于kmeans集群进行KNN搜索的代码。

Cluster_Scoring <- function(score_table, newdata){ centers <- score_table n_centers <- nrow(centers) #for() #n_1=1:(nrow(newdata)/100) dist_mat <- as.matrix(dist(rbind(centers, newdata[n_1,]))) dist_mat <- dist_mat[-seq(n_centers), seq(n_centers)] a=max.col(-dist_mat) }

再次，此代码仅适用于小型数据集，并且在像我的大型数据集（1.5亿行/ 200列）上运行时返回错误

dyn.load（“ KNN_ANN.so”）转向错误找不到符号，预期出现在：调用C函数时的平面名称空间

0 个答案: