Rcpp Data Frame Return with a List Column (Where is the AsIs?)

Question

I would like to generate a data frame in an Rcpp function which contains a list column. I have tried several things and have been unable to find a working solution. The following is a Rcpp c++ file where I have attempted this:

#include <Rcpp.h>
#include <vector>

using namespace Rcpp;
using namespace std;

// [[Rcpp::export]]
DataFrame makeListColumn() {

  vector<RawVector> the_future_list;

  the_future_list.push_back(RawVector::create(0, 1, 2));
  the_future_list.push_back(RawVector::create(3, 4));
  the_future_list.push_back(RawVector::create(5, 6, 7, 8, 9, 10));



  vector<int> another_regular_column;
  another_regular_column.push_back(42);
  another_regular_column.push_back(24);
  another_regular_column.push_back(4242);

  DataFrame ret = DataFrame::create(Named("another_regular_column") = another_regular_column, Named("thelistcol") = the_future_list);

  return ret;
}

/*** R
a = makeListColumn()
dput(a)
*/

The output from this is the following:

a = makeListColumn()

structure(list(another_regular_column = c(42L, 24L, 4242L, 42L, 24L, 4242L), thelistcol.as.raw.c.0x00..0x01..0x02.. = as.raw(c(0x00, 0x01, 0x02, 0x00, 0x01, 0x02)), thelistcol.as.raw.c.0x03..0x04.. = as.raw(c(0x03, 0x04, 0x03, 0x04, 0x03, 0x04)), thelistcol.as.raw.c.0x05..0x06..0x07..0x08..0x09..0x0a.. = as.raw(c(0x05, 0x06, 0x07, 0x08, 0x09, 0x0a))), class = "data.frame", row.names = c(NA, -6L))

What I am looking for is the following (done in a regular R script):

what_i_wanted = data.frame(
  another_regular_column = c(42, 24, 4242),  
  thelistcol = I(list(as.raw(c(0,1,2)), as.raw(c(3, 4)), as.raw(c(5, 6, 7, 8, 9, 10))))
)

This produces the output:

structure(list(another_regular_column = c(42, 24, 4242), thelistcol = structure(list( as.raw(c(0x00, 0x01, 0x02)), as.raw(c(0x03, 0x04)), as.raw(c(0x05, 0x06, 0x07, 0x08, 0x09, 0x0a))), class = "AsIs")), class = "data.frame", row.names = c(NA, -3L))

The primary difference is between the R and the Rcpp approach is the I() call in the R code. If I remove that, the R code generates the same structure as the Rcpp code. I did some looking in the Rcpp documentation and did some google searches, but have come up empty handed.

Can somebody help me understand what I need to do in Rcpp to get this to work?

EDIT:

I did try to do something like:

List the_list = List::create(the_future_list);
the_list.attr("class") = CharacterVector::create("AsIs");

This unfortunately resulted in the following error:

Error in makeListColumn() : Could not convert using R function: as.data.frame.

score 4 · Accepted Answer · answered Oct 18 '19 at 15:05

AsIs isn't implemented.

The best way to handle working with list columns in a data.frame within C++ is to use Rcpp::List to handle the construction. Recall that a data.frame is a list with an observation count enforcement. In addition, we can modify the Rcpp::List object attributes -- unlike a std data structure -- to include the AsIs flag.

In short, this looks like:

#include <Rcpp.h>

// [[Rcpp::export]]
SEXP makeListColumn() {
// ^ Changed from Rcpp::DataFrame to a general SEXP object. 

  // Store inside of an Rcpp List
  Rcpp::List the_future_list(3);
  the_future_list[0] = Rcpp::RawVector::create(0, 1, 2);
  the_future_list[1] = Rcpp::RawVector::create(3, 4);
  the_future_list[2] = Rcpp::RawVector::create(5, 6, 7, 8, 9, 10);

  // Mark with AsIs
  the_future_list.attr("class") = "AsIs";

  // Store inside of a regular vector
  std::vector<int> another_regular_column;
  another_regular_column.push_back(42);
  another_regular_column.push_back(24);
  another_regular_column.push_back(4242);

  // Construct a list 
  Rcpp::List ret = Rcpp::List::create(
      Rcpp::Named("another_regular_column") = another_regular_column,
      Rcpp::Named("thelistcol") = the_future_list);

  // Coerce to a data.frame
  ret.attr("class") = "data.frame";
  ret.attr("row.names") = Rcpp::seq(1, another_regular_column.size());

  // Return the data.frame
  return ret;
}

Most importantly, note we forgo the Rcpp::DataFrame class and return a SEXP object. Moreover, we coerce an Rcpp::List to an Rcpp::DataFrame by changing its class and assigning row.names to the object.

In practice, the code returns:

a = makeListColumn()
str(a)
# 'data.frame': 3 obs. of  2 variables:
# $ another_regular_column: int  42 24 4242
# $ thelistcol            :List of 3
#  ..$ : raw  00 01 02
#  ..$ : raw  03 04
#  ..$ : raw  05 06 07 08 ...
#  ..- attr(*, "class")= chr "AsIs"

Compared to the desired result:

what_i_wanted = data.frame(
    another_regular_column = c(42, 24, 4242),  
    thelistcol = I(list(as.raw(c(0,1,2)), as.raw(c(3, 4)), as.raw(c(5, 6, 7, 8, 9, 10))))
)

str(what_i_wanted)
# 'data.frame': 3 obs. of  2 variables:
# $ another_regular_column: num  42 24 4242
# $ thelistcol            :List of 3
#  ..$ : raw  00 01 02
#  ..$ : raw  03 04
#  ..$ : raw  05 06 07 08 ...
#  ..- attr(*, "class")= chr "AsIs"

all.equal(a, what_i_wanted)
# [1] TRUE

Rcpp Data Frame Return with a List Column (Where is the AsIs?)

1 Answers1

Linked