Question

当我使用下面的代码将库data.table加载到cpu集群时，

R抛出一个错误。但是data.table包安装在R上，并且在并行代码之外使用时工作正常。

#include <iostream>
#include <vector>
#include <map>

class Base
{
  private:
    std::map<int, std::string> m_base1, m_base2;
    std::vector<std::string> m_str1 = {"one", "two", "three"};
    std::vector<std::string> m_str2 = {"four", "five", "six"};

  public:
    std::vector<std::string> &str1() { return m_str1; }
    std::vector<std::string> &str2() { return m_str2; }

    std::map<int, std::string> &base1() { return m_base1; }
    std::map<int, std::string> &base2() { return m_base2; }
};

template <typename T>
void fill_vec(T *b)
{
    size_t counter = 0;
    for (const auto &str_iter : b->str1())
        (b->base1())[counter++] = str_iter;

    counter=0;
    for (const auto &str_iter : b->str2())
        (b->base2())[counter++] = str_iter;
}

int main(int argc, char *argv[])
{
    Base *b = new Base;
    fill_vec(b);

    return 0;
}

错误： -

clusterEvalQ（cl，library（data.table）） checkForRemoteErrors出错（lapply（cl，recvResult））： 3个节点产生错误;第一个错误：没有名为＆＃39; data.table＆＃39;
的包

Answer 1

基于HenrikB在评论中所说的内容，我通过将我的.libPaths（）调用添加到clusterEvalQ（）来解决这个问题：

.libPaths("C:/programs/rlib")
library(parallel)
no_cores<-detectCores()-1

cl<-makeCluster(no_cores)
#this is needed to see the package
clusterEvalQ(cl, .libPaths("C:/programs/rlib"))

# I'm using a function that uses the stringdist library
clusterEvalQ(cl, library(stringdist))

#You need to load your data into the cluster also
clusterExport(cl, "unmatched")
clusterExport(cl, "matched")

#now we're going to run it, amatch is a function in the stringdist lib

parLapply(cl, unmatched,function(x) amatch(x,matched, maxDist = Inf))

无法将库data.table加载到R parallel

1 个答案: