error: ‘Load’ is not a member of ‘mlpack::data’

[nikhilsingh13]

The problem is with loading and saving the data. Am I doing it correctly or are there some changes that I am supposed to make. Kindly advise.

The code I am using is mentioned below.

library("inline")
library("RcppMLPACK")

write.table(iris[,1:4] , "iris.csv" , col.names = FALSE , row.names = FALSE)

code <- '
  std::string inputFile = Rcpp::as<std::string>(input);
  std::string outputFile = Rcpp::as<std::string>(output);
  arma::mat data;
  mlpack::data::Load(inputFile ,data ,true);
  arma::mat cov = data * trans(data) / data.n_cols;
  mlpack::data::Save(outputFile, cov)
'
mlCov_2 <- cxxfunction(signature(input="iris.csv" , output = "cov.csv") , code, plugin="RcppMLPACK")

The error is mentioned below:

Error in compileCode(f, code, language = language, verbose = verbose) : 
  Compilation ERROR, function(s)/method(s) not created! file27c07f10c076.cpp: In function ‘SEXPREC* file27c07f10c076(SEXP, SEXP)’:
file27c07f10c076.cpp:34:3: error: ‘Load’ is not a member of ‘mlpack::data’
   mlpack::data::Load(inputFile ,data ,true);
   ^
file27c07f10c076.cpp:36:3: error: ‘Save’ is not a member of ‘mlpack::data’
   mlpack::data::Save(outputFile, cov)
   ^
In file included from /home/nikhil/R/x86_64-pc-linux-gnu-library/3.3/Rcpp/include/Rcpp/r/headers.h:50:0,
                 from /home/nikhil/R/x86_64-pc-linux-gnu-library/3.3/Rcpp/include/RcppCommon.h:29,
                 from /home/nikhil/R/x86_64-pc-linux-gnu-library/3.3/RcppArmadillo/include/RcppArmadilloForward.h:26,
                 from /home/nikhil/R/x86_64-pc-linux-gnu-library/3.3/RcppArmadillo/include/RcppArmadillo.h:31,
                 from /home/nikhil/R/x86_64-pc-linux-gnu-library/3.3/RcppMLPACK/include/mlpack/core/arma_extend/arma_extend.hpp:19,
                 from /home/nik
In addition: Warning message:
running command '/usr/lib/R/bin/R CMD SHLIB file27c07f10c076.cpp 2> file27c07f10c076.cpp.err.txt' had status 1

Version of

RcppMLPACK_1.0.10-6 
inline_0.3.14

[eddelbuettel]

Darn. I wrote a response this morning before leaving for work. Must not have pressed the button.

In short: yes, we are not including the Load and Save headers. But once you add those it falls over for lack of Log ...

[eddelbuettel]

Ok, I started to write a post this morning around this file, a slightly modified version of yours:

#include <RcppMLPACK.h>
#include <mlpack/core/data/load.hpp>
#include <mlpack/core/data/save.hpp>

// [[Rcpp::depends(RcppMLPACK)]]

// [[Rcpp::export]]
bool mlCov2(std::string inputFile, std::string outputFile) {
  arma::mat data;
  mlpack::data::Load(inputFile ,data ,true);
  arma::mat cov = data * trans(data) / data.n_cols;
  mlpack::data::Save(outputFile, cov);
  return true;
}
    

/***R
write.table(iris[,1:4] , "/tmp/iris.csv" , col.names = FALSE , row.names = FALSE)
mlCov2("/tmp/iris.csv", "/tmp/iris2.csv")
*/

We can sourceCpp("filename.cpp") this and it survives Load and Save now, but stumbles over Log.

That is the bad news. The good news is that it works for me with the under-development RcppMLPACK2 -- at least after I fixed the plugin there.

[talegari]

@eddelbuettel
Thanks for your response.

I tried to reproduce your solution with rcppmlpack2, but no awail:

Sys.setenv("PKG_CXXFLAGS"="-std=c++11")

library("Rcpp")
library("RcppMLPACK")

writeLines("#include <RcppMLPACK.h>
#include <mlpack/core/data/load.hpp>
#include <mlpack/core/data/save.hpp>

// [[Rcpp::depends(RcppMLPACK)]]

// [[Rcpp::export]]
bool mlCov2(std::string inputFile, std::string outputFile) {
  arma::mat data;
  mlpack::data::Load(inputFile ,data ,true);
  arma::mat cov = data * trans(data) / data.n_cols;
  mlpack::data::Save(outputFile, cov);
  return true;
}
"
, con = "dirk_cpp.cpp")

Rcpp::sourceCpp("dirk_cpp.cpp")
write.table(iris[,1:4] , "iris.csv" , col.names = FALSE , row.names = FALSE)
mlCov2("iris.csv", "iris2.csv")

I get the following error at Rcpp::sourceCpp("dirk_cpp.cpp") line:

In file included from /usr/include/mlpack/prereqs.hpp:65:0,
                 from /usr/include/mlpack/core.hpp:207,
                 from /home/adminn/R/x86_64-pc-linux-gnu-library/3.3/RcppMLPACK/include/RcppMLPACK.h:22,
                 from dirk_cpp.cpp:1:
/usr/include/mlpack/core/boost_backport/boost_backport.hpp:28:51: fatal error: boost/serialization/unordered_map.hpp: No such file or directory
   #include <boost/serialization/unordered_map.hpp>
                                                   ^
compilation terminated.

I built rcppmlpack2 on ubuntu using this

[eddelbuettel]

I am on Ubuntu too so we should be able to make this work.

Start with my PPA to get a pre-made and current .deb for MLPACK 2.1.0.

The test file I used is

#include <RcppMLPACK.h>
#include <mlpack/core/data/load.hpp>
#include <mlpack/core/data/save.hpp>

// [[Rcpp::depends(RcppMLPACK)]]

// [[Rcpp::export]]
bool mlCov2(std::string inputFile, std::string outputFile) {
  arma::mat data;
  mlpack::data::Load(inputFile ,data ,true);
  arma::mat cov = data * trans(data) / data.n_cols;
  mlpack::data::Save(outputFile, cov);
  return true;
}
    

/***R
write.table(iris[,1:4] , "/tmp/iris.csv" , col.names = FALSE , row.names = FALSE)
mlCov2("/tmp/iris.csv", "/tmp/iris2.csv")
*/

which you can just sourceCpp() -- it will also execute the R block at the bottom.

[eddelbuettel]

And you probably missed one more commit reflecting changes I had on my laptop from last evening. I just committed that this morning...

[eddelbuettel]

For what it is worth, with the updated RcppMLPACK2 I can build and run this file

#include <RcppMLPACK.h>

// [[Rcpp::plugins(cpp11)]]
// [[Rcpp::depends(RcppMLPACK)]]

// [[Rcpp::export]]
bool mlCov2(std::string inputFile, std::string outputFile) {
  arma::mat data;
  mlpack::data::Load(inputFile ,data ,true);
  arma::mat cov = data * trans(data) / data.n_cols;
  mlpack::data::Save(outputFile, cov);
  return true;
}
    
/***R
write.table(iris[,1:4] , "/tmp/iris.csv" , col.names = FALSE , row.names = FALSE)
mlCov2("/tmp/iris.csv", "/tmp/iris2.csv")
*/

This turns on C++11 and tells R to use headers from RcppMLPACK (based on the RcppMLPACK2 repo). Then:

R> sourceCpp("/tmp/mlpackError.cpp")

R> write.table(iris[,1:4] , "/tmp/iris.csv" , col.names = FALSE , row.names = FALSE)

R> mlCov2("/tmp/iris.csv", "/tmp/iris2.csv")
[WARN ] '/tmp/iris.csv' is not a standard csv file.
[1] TRUE
R> 

[talegari]

@eddelbuettel

Thanks for your quick reply.

I installed via ppa again and new commit of rcppmlpack2. But I still end up with the same error:

/usr/include/mlpack/core/boost_backport/boost_backport.hpp:28:51: fatal error: boost/serialization/unordered_map.hpp: No such file or directory #include <boost/serialization/unordered_map.hpp>

Would you confirm if you can find this file in the directory on your computer?

[eddelbuettel]

You may need to install libboost-serialization-dev -- this is generic: when headers foo.h are missing you often simply need libfoo-dev.

[talegari]

Strangely, its libboost-serialization-dev is already the newest version

[eddelbuettel]

Then I cannot help you beyond repeating that it works for me. The information you supply is not sufficient to say anything more.