[Rcpp-devel] Joining each row of CharacterMatrix to return a CharacterVector?

I preface this by stating that I'm very much a Rcpp beginner who is
comfortable in R but I've never before used C++. I'm working through the
Rcpp documentation but haven't been able to answer my question.

I've written an Rcpp (v0.10.1) function f that takes as input a
CharacterMatrix X. X has 20 million rows and 100 columns. For each row of X
the function alters certain entries of that row according to rules governed
by some other input variables. f returns the updated version of X. This
function works as I'd like it to:
# a toy example with nrow = 2, ncol = 2

X <- matrix('A', ncol = 2, nrow = 2)
X

     [,1] [,2]
[1,] "A"  "A"
[2,] "A"  "A"

X <- f(X, other_input_variables)
X

     [,1] [,2]
[1,] "Z"  "A"
[2,] "z"  "A"

However, instead of f returning a CharacterMatrix as it currently does, I'd
like to return a CharacterVector Y, where each element of Y is a "collapsed"
row of the updated X.

I can achieve the desired result in R by using:
Y <- apply(X=X, MARGIN = 1, FUN = function(x){paste0(x, collapse = '')})

Y

[1] "ZA" "zA"

Hi Pete,

On 11 December 2012 at 09:43, hickey at wehi.EDU.AU wrote:
| I preface this by stating that I'm very much a Rcpp beginner who is comfortable
| in R but I've never before used C++. I'm working through the Rcpp documentation
| but haven't been able to answer my question.
| 
| I've written an Rcpp (v0.10.1) function f that takes as input a CharacterMatrix
| X. X has 20 million rows and 100 columns. For each row of X the function alters
| certain entries of that row according to rules governed by some other input
| variables. f returns the updated version of X. This function works as I'd like
| it to: 
| # a toy example with nrow = 2, ncol = 2
| > X <- matrix('A', ncol = 2, nrow = 2)
| > X
|      [,1] [,2]
| [1,] "A"  "A" 
| [2,] "A"  "A" 
| > X <- f(X, other_input_variables)
| > X
|      [,1] [,2]
| [1,] "Z"  "A" 
| [2,] "z"  "A" 
| 
| However, instead of f returning a CharacterMatrix as it currently does, I'd
| like to return a CharacterVector Y, where each element of Y is a "collapsed"
| row of the updated X.
| 
| I can achieve the desired result in R by using: 
| Y <- apply(X=X, MARGIN = 1, FUN = function(x){paste0(x, collapse = '')}) 
| > Y
| [1] "ZA" "zA"
| 
| but I wondered whether this "joining" is likely to be more efficiently
| performed within my function f? If so, how do I join the 100 individual
| character entries of a row of the CharacterMatrix X into a single string that
| will then comprise an element of the returned CharacterVector Y?

Ah, the joy of working with character strings/vectors/pointers :)  

You certainly can. And there will be a lot of old, bad, ... tutorials out
there.  I can't right now think of a good tutorial to point you to -- other
than the perennial "C++ Annotations" by Brokken which is at the same time
good, current, up-to-date and free (!!) -- so maybe you shoud continue with
the little 2 x 2 and 3 x 3 examples:

i)   loop over a row, first init the target string to be ""
ii)  assign each element of the matrix to a string
iii) append, which can be as easy as using the   +   for two strings
iv)  accumulate the result strings in a vector of strings

That should work, does not require pointers, free, malloc, ...  You can
optimize later.

Hope this helps,  Dirk

-- 
Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com  

I preface this by stating that I'm very much a Rcpp beginner who is
comfortable in R but I've never before used C++. I'm working through the
Rcpp documentation but haven't been able to answer my question.

I've written an Rcpp (v0.10.1) function f that takes as input a
CharacterMatrix X. X has 20 million rows and 100 columns. For each row
of X the function alters certain entries of that row according to rules
governed by some other input variables. f returns the updated version of
X. This function works as I'd like it to:
# a toy example with nrow = 2, ncol = 2

 > X <- matrix('A', ncol = 2, nrow = 2)
 > X

      [,1] [,2]
[1,] "A"  "A"
[2,] "A"  "A"

 > X <- f(X, other_input_variables)
 > X

      [,1] [,2]
[1,] "Z"  "A"
[2,] "z"  "A"

However, instead of f returning a CharacterMatrix as it currently does,
I'd like to return a CharacterVector Y, where each element of Y is a
"collapsed" row of the updated X.

I can achieve the desired result in R by using:
Y <- apply(X=X, MARGIN = 1, FUN = function(x){paste0(x, collapse = '')})

 > Y

[1] "ZA" "zA"

but I wondered whether this "joining" is likely to be more efficiently
performed within my function f? If so, how do I join the 100 individual
character entries of a row of the CharacterMatrix X into a single string
that will then comprise an element of the returned CharacterVector Y?

Many thanks,
Pete
--------------------------------
Peter Hickey,
PhD Student/Research Assistant,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
Ph: +613 9345 2324

hickey at wehi.edu.au <mailto:hickey at wehi.edu.au>
http://www.wehi.edu.au

Hello,

We don't know your function f, so this is hard to say. Anyway, this
below implements something similar to apply(.,1,paste0) in rcpp (current
devl version):

#include <Rcpp.h>
using namespace Rcpp ;

// [[Rcpp::export]]
CharacterVector pasteColumns(CharacterMatrix m){
     String buffer ;
     int nc = m.ncol(), nr = m.nrow() ;
     CharacterVector out(nr) ;
     for( int i=0; i<nr; i++){
         CharacterMatrix::Row row = m(i,_) ;
         buffer = "" ;
         for( int j=0; j<nc; j++){
             buffer += row[j] ;
         }
         out[i] = buffer ;
     }
     return out ;
}

With this, I get these timings:

     nc <- 100; nr <- 2e4
     M <- matrix( sample(letters, nc*nr, replace = TRUE) , ncol = nc )

     require(microbenchmark)
     microbenchmark(
         pasteColumns(M),
         apply(M, 1, paste0)
         )
     Unit: milliseconds
                  expr       min        lq    median        uq      max
     1 apply(M, 1, paste0) 451.39975 484.41435 495.92757 501.58728 714.1418
     2     pasteColumns(M)  67.91322  68.29269  70.34704  77.09383 145.9161



Le 10/12/12 23:43, hickey at wehi.EDU.AU a ?crit :

I preface this by stating that I'm very much a Rcpp beginner who is
comfortable in R but I've never before used C++. I'm working through the
Rcpp documentation but haven't been able to answer my question.

I've written an Rcpp (v0.10.1) function f that takes as input a
CharacterMatrix X. X has 20 million rows and 100 columns. For each row
of X the function alters certain entries of that row according to rules
governed by some other input variables. f returns the updated version of
X. This function works as I'd like it to:
# a toy example with nrow = 2, ncol = 2

 > X <- matrix('A', ncol = 2, nrow = 2)
 > X

      [,1] [,2]
[1,] "A"  "A"
[2,] "A"  "A"

 > X <- f(X, other_input_variables)
 > X

      [,1] [,2]
[1,] "Z"  "A"
[2,] "z"  "A"

However, instead of f returning a CharacterMatrix as it currently does,
I'd like to return a CharacterVector Y, where each element of Y is a
"collapsed" row of the updated X.

I can achieve the desired result in R by using:
Y <- apply(X=X, MARGIN = 1, FUN = function(x){paste0(x, collapse = '')})

 > Y

[1] "ZA" "zA"

but I wondered whether this "joining" is likely to be more efficiently
performed within my function f? If so, how do I join the 100 individual
character entries of a row of the CharacterMatrix X into a single string
that will then comprise an element of the returned CharacterVector Y?

Many thanks,
Pete
--------------------------------
Peter Hickey,
PhD Student/Research Assistant,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
Ph: +613 9345 2324

hickey at wehi.edu.au <mailto:hickey at wehi.edu.au>
http://www.wehi.edu.au