[Rcpp-devel] list of matrices to Eigen matrices
I have a list of matrices in R which I want to turn into a bunch of
Eigen matrices using RcppEigen. I found a way to do this, but it seems
unnecessarily tortured. Here's what I have:
// [[Rcpp::export]]
void listMat(Rcpp::List mat_list) {
int n_mats = mat_list.size();
std::vector vec_of_mats(n_mats);
for (int i=0; i < n_mats; i++) {
Rcpp::NumericMatrix tmpcv = mat_list[i];
double *pcv = &tmpcv(0,0);
Eigen::Map tmpmapd(pcv, tmpcv.nrow(), tmpcv.ncol());
vec_of_mats[i] = Eigen::MatrixXf(tmpmapd.cast());
}
std::cout << vec_of_mats[0] << std::endl; // Sanity check.
return;
}
Note that I'm putting each matrix in my R list into a std::vector
where each element is a float matrix (ie MatrixXf). This does assume
that Rcpp::NumericMatrix stores the matrix data contiguously and in
column-major order, which is true, but will it always necessarily be
so? (I get a little bit nervous about reaching into a class like this
and just pulling out an address). Is there a more direct way of doing
this?
Thanks,
Mark
___
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
Re: [Rcpp-devel] list of matrices to Eigen matrices
Mark,
On 20 March 2013 at 13:03, M A wrote:
| I have a list of matrices in R which I want to turn into a bunch of
| Eigen matrices using RcppEigen. I found a way to do this, but it seems
| unnecessarily tortured. Here's what I have:
Nope, that is fine. (Apart from maybe doing with *pcv and sticking that into
the ctor directly). Someone somewhere needs to fill the vector; we don't
have a function for it so you had to fill this in.
| // [[Rcpp::export]]
| void listMat(Rcpp::List mat_list) {
| int n_mats = mat_list.size();
| std::vector vec_of_mats(n_mats);
|
| for (int i=0; i < n_mats; i++) {
| Rcpp::NumericMatrix tmpcv = mat_list[i];
| double *pcv = &tmpcv(0,0);
|
| Eigen::Map tmpmapd(pcv, tmpcv.nrow(), tmpcv.ncol());
| vec_of_mats[i] = Eigen::MatrixXf(tmpmapd.cast());
| }
| std::cout << vec_of_mats[0] << std::endl; // Sanity check.
| return;
| }
|
| Note that I'm putting each matrix in my R list into a std::vector
| where each element is a float matrix (ie MatrixXf). This does assume
| that Rcpp::NumericMatrix stores the matrix data contiguously and in
| column-major order, which is true, but will it always necessarily be
| so? (I get a little bit nervous about reaching into a class like this
No need to be nervous as we get the matrix from R which (presumably) won't
change these internals anytime soo.
| and just pulling out an address). Is there a more direct way of doing
| this?
It looks fine to me. I'd write very similar code.
Dirk
--
Dirk Eddelbuettel | e...@debian.org | http://dirk.eddelbuettel.com
___
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
[Rcpp-devel] Installing my own package in the system directory
Hi, I use an Intel desktop under Ubuntu 11.04, and I have super-user privileges. I wrote a utility i/o function to use in R. I am able to install it in my account's R/site-library, and it works fine. I cannot install it on the system R/site-library using either R CMD INSTALL or install.packages; I get an "invalid package" warning followed by an "Error: ERROR: no package specified", etc. All of that goes away if I specify to install in my account's site library. I read about setting up a local repository, but before I plunge into setting something like that for little things, I'd like to know I haven't missed something obvious. Is there something else I may do? Best, Mario ___ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
Re: [Rcpp-devel] Installing my own package in the system directory
Mario, That email had nothing to do with Rcpp and is there unsuitable for this mailing list. Please do not abuse the list. Dirk -- Dirk Eddelbuettel | e...@debian.org | http://dirk.eddelbuettel.com ___ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
Re: [Rcpp-devel] Installing my own package in the system directory
Sorry. Sent from my iPhone On Mar 20, 2013, at 5:43 PM, Dirk Eddelbuettel wrote: > > Mario, > > That email had nothing to do with Rcpp and is there unsuitable for this > mailing list. Please do not abuse the list. > > Dirk > > -- > Dirk Eddelbuettel | e...@debian.org | http://dirk.eddelbuettel.com ___ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
Re: [Rcpp-devel] Rcpp Google summer of code 2013?
On 03/18/2013 12:50 AM, Dirk Eddelbuettel wrote: On 17 March 2013 at 22:34, Michael Weylandt wrote: | There's an active R-GSOC bunch and I'm sure a good Rcpp project would get | support that way. Dirk has sponsored GSOC projects in the past so I bet you | wouldn't have too much trouble finding a mentor. ;-) Yep, I was quite involved for three years and even acted as admin one year, but I stopped mentoring after GSoC 2011. For one I don't have the bandwidth for it (but come to think about, I think I offered to be a back-up mentor for some thing using Rcpp but being primairly BioC) and Google also encourages rotation of personel so that things don't get stuck in the same rut. So if someone wants to have a go at a Rcpp-based GSoC project: by all means, go for it. Thank you for your reply Dirk. If someone would feel tempted to mentor the project I would be more then willing to show interest as a student! Berry | http://rwiki.sciviews.org/doku.php?id=developers:projects:gsoc2013 | And a mailing list off google groups. Yup. And there are of course other organizations too (eg I mentored for R as well as within Debian). Dirk ___ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
[Rcpp-devel] Simple project segfaults
Hi everybody,
Just now I was playing around with Rcpp in combination with Qt (I have
found an excellent example on this). And decided to have a go(small
dummyproject) at this myself. Problem is that the application segfaults
when I run it.
QtR.h
http://pastebin.com/AXT144SK
QtR.cpp
http://pastebin.com/JUUuCFN2
main.cpp
http://pastebin.com/yWUmDL17
mainwindow.cpp
http://pastebin.com/gE84ZtxH
I create an instance of my QtR class called qtr which creates an RInside
session in the constructor. In my mainwindow.cpp I reference to my
instance of qtr and then invoke the loadData method. In the loadData
method I try to use
m_R.parseEvalQ("cat("+fname_no_ext+"',file='outfile.txt',sep='\n')");
This results in a segfault.
However when I remove this line from the loadData method and add the
m_R.parseEvalQ("cat("+fname_no_ext+"',file='outfile.txt',sep='\n')");
line to the constructor of QtR which is invoked in main.cpp it runs
without trouble.
I probably am making a very obvious mistake. But I seem unable to see
the issue at this point myself.
Berry
PS
I have used pastebin for readability, or would everybody prefer the code
within the e-mail?
___
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
