Re: [Rd] getConnection, R_outpstream_st
Hello,
Maybe I should give more background about what we are trying to do with
this. We (Dirk, Saptarshi and I) are interfacing R with the protocol
buffer C++ library from google [1,2] .
protocol buffers are google's data interchange format, designed to be
efficient (much more efficient than XML) and language neutral.
Google supports bindings to java, c++ and python, and there are dozens
of third party bindings (perl, C, C#, ...) available. We are doing the
one for R, allowing to create and manipulate protocol buffer messages
from R. For example with this message type :
package tutorial ;
message Person {
required int32 id = 1;
required string name = 2;
optional string email = 3;
}
you can create a message like this :
message <- new( tutorial.Person, id = 0, name = "Romain", email =
"francoisrom...@free.fr" )
We can also read/write messages from/to files :
tutorial.Person$read( "somefile" )
serialize( message, "somefile" )
Now we'd like to be able to use the extensive IO support R provides
(connections) and essentially read and write messages from and to
arbitrary binary connections. Practically, it means calling this method
of the google::protobuf::Message c++ class [3,4]
bool Message::ParseFromIstream(istream * input)
bool Message::SerializeToOstream( ostream * output) const
This means we need to be able to build istream and ostream pointers
working together with a binary R connection. It seemed to me from
reading the code of serialize.c in R that we would be able to achieve
this by sort of wrapping R_inpstream_st and R_outpstream_st as c++
streams.
The only thing I cannot figure out at that point is how to access a
connection pointer (Rconnection). From this :
#ifdef NEED_CONNECTION_PSTREAMS
/* The connection interface is not yet available to packages. To
allow limited use of connection pointers this defines the opaque
pointer type. */
#ifndef HAVE_RCONNECTION_TYPEDEF
typedef struct Rconn *Rconnection;
#define HAVE_RCONNECTION_TYPEDEF
#endif
void R_InitConnOutPStream(R_outpstream_t stream, Rconnection con,
R_pstream_format_t type, int version,
SEXP (*phook)(SEXP, SEXP), SEXP pdata);
void R_InitConnInPStream(R_inpstream_t stream, Rconnection con,
R_pstream_format_t type,
SEXP (*phook)(SEXP, SEXP), SEXP pdata);
#endif
I get that if I do have a Rconnection, then I can initialize a
R_outpstream_t and just use this, but I can't find a way to get one. The
internal "getConnection" seems to do just this, but it is not accessible
from packages.
We'd appreciate any help.
Romain
[1] http://r-forge.r-project.org/projects/rprotobuf/
[2] http://code.google.com/p/protobuf/
[3]
http://code.google.com/apis/protocolbuffers/docs/reference/cpp/google.protobuf.message.html#Message.ParseFromIstream.details
[4]
http://code.google.com/apis/protocolbuffers/docs/reference/cpp/google.protobuf.message.html#Message.SerializeToOstream.details
On 11/07/2009 11:24 AM, Romain Francois wrote:
Hello,
I'm trying to use the limited connections api defined in
Rinternals.h. I have code that looks like this (inspired from
do_serializeToConn) :
SEXP serialize_to_connection( SEXP xp, SEXP connection ){
Rconnection con ; struct R_outpstream_st out; R_pstream_format_t type
= R_pstream_binary_format ; SEXP (*hook)(SEXP, SEXP) = NULL ; con =
getConnection(Rf_asInteger(connection)); R_InitConnOutPStream(&out,
con, type, 0, hook, R_NilValue );
return R_NilValue ; }
The problem I have is that I cannot actually call getConnection since
it is not part of the api. Is there another way to get the
Rconnection that is associated with a number.
Many thanks.
Romain
--
Romain Francois
Professional R Enthusiast
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
|- http://tr.im/BcPw : celebrating R commit #5
|- http://tr.im/ztCu : RGG #158:161: examples of package IDPmisc
`- http://tr.im/yw8E : New R package : sos
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] RSiteSearch and my R site
On 08/11/2009 7:23 AM, Jonathan Baron wrote: Duncan, This is not a problem with anything you just did, but I'm writing you and ccing r-devel, in case I am allowed to post there (which I think I'm not). R-devel is a public list, no problem posting here. The problem (in case this gets to r-devel) is that I need to have static html files for all packages, and static html ONLY for new packages. You have fixed the latter problem in the latest version of R-devel (11/08/09), in "R CMD INSTALL". Actually that fix broke some other things, so it's not in place at the moment. We'll get something there, but it might be different from what I did yesterday. > But when I install R-devel from source I cannot get all the html files. The configure script has an option for --enable-prebuilt-html, and this does build the static html files for about half of the standard packages! These include base, utils, tools, ..., but they do not include datasets, boot, cluster, ... I can of course work around this by using your new version of R CMD INSTALL after installing everything else. It does not wipe out the executables that are already installed (I think). But this might actually be a bug that was not intended. Yes, it sounds like it. Duncan __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] RSiteSearch and my R site
Duncan, This is not a problem with anything you just did, but I'm writing you and ccing r-devel, in case I am allowed to post there (which I think I'm not). The problem (in case this gets to r-devel) is that I need to have static html files for all packages, and static html ONLY for new packages. You have fixed the latter problem in the latest version of R-devel (11/08/09), in "R CMD INSTALL". But when I install R-devel from source I cannot get all the html files. The configure script has an option for --enable-prebuilt-html, and this does build the static html files for about half of the standard packages! These include base, utils, tools, ..., but they do not include datasets, boot, cluster, ... I can of course work around this by using your new version of R CMD INSTALL after installing everything else. It does not wipe out the executables that are already installed (I think). But this might actually be a bug that was not intended. Jon -- Jonathan Baron, Professor of Psychology, University of Pennsylvania Home page: http://www.sas.upenn.edu/~baron __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] Summary methods
I've defined the following for objects of a class called jml
summary.jml <- function(object, ...){
tab <- cbind(Estimate = coef(object),
StdError = object$se,
Infit = object$Infit,
Outfit = object$Outfit)
res <- list(call = object$call, coefficients = tab,
N = nrow(object$Data), iter = object$Iterations)
class(res) <- "summary.jml"
res
}
print.summary.jml <- function(x, ...){
cat("Call:\n")
print(x$call)
cat("\n")
cat("Number of iterations to completion:", x$iter, "\n")
cat("Number of individuals used in estimation:", x$N, "\n")
cat("\n")
printCoefmat(x$coefficients)
}
Use of the methods on a fitted jml object yields:
> summary(aa)
Call:
jml2.formula(formula = ~V1 + V2 + V3 + V4 + V5 + V6 + V7 + V8 +
V9 + V10, data = itemDat, bias = F)
Number of iterations to completion: 6
Number of individuals used in estimation: 486
StdError Infit Outfit
[1,] -0.380346 0.103002 1.007466 0.9935
[2,] 0.025939 0.104052 1.003050 1.0083
[3,] 2.563784 0.171174 0.941453 0.9414
[4,] -2.930519 0.156923 1.010786 1.0515
[5,] 1.139241 0.118932 0.978101 1.1424
[6,] -1.461751 0.111563 1.030612 1.2709
[7,] 0.486202 0.107986 1.008374 1.0394
[8,] -0.497102 0.103117 0.961431 0.9012
[9,] -0.486478 0.103099 1.001752 0.9829
[10,] 1.541029 0.129214 1.010011 0.9150
Two questions. First, why is the name of the first column empty instead of
"Estimate" as I have specified in the summary method? Second, I am struggling
to get the row names of the coefficients to align with the formula call. For
example, instead of
StdError Infit Outfit
[1,] -0.380346 0.103002 1.007466 0.9935
I would prefer
StdError Infit Outfit
V1 -0.380346 0.103002 1.007466 0.9935
This also occurs in my print method
print.jml <- function(x, digits = 2, ...){
cat("\nCall:\n", deparse(x$call), "\n\n", sep = "")
cat("Coefficients:\n")
print.default(format(coef(x), digits = digits), print.gap=2,
quote = FALSE)
invisible(x)
}
Which produces
> win
Call:
jml2.default(dat = itemDat[, 1:10])
Coefficients:
[,1]
[1,] -0.38034638
[2,] 0.02593937
[3,] 2.56378422
[4,] -2.93051899
[5,] 1.13924076
[6,] -1.46175119
[7,] 0.48620247
[8,] -0.49710150
[9,] -0.48647770
[10,] 1.54102895
Thank you
Harold
> sessionInfo()
R version 2.10.0 (2009-10-26)
i386-pc-mingw32
locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
[5] LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] MASS_7.3-3 lme4_0.999375-32 Matrix_0.999375-31 lattice_0.17-26
[5] MiscPsycho_1.4 statmod_1.4.1
loaded via a namespace (and not attached):
[1] grid_2.10.0 plink_1.2-2 tools_2.10.0
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] RcppTemplate
The Rcpp R/C++ object mapping library and package template has been updated on CRAN in package RcppTemplate. It allows you to work with R objects like data frames and zoo time series in C++ programs. R can call C++ functions, and C++ objects can call R functions, with parameters and return values of any of the most commonly used R data types (data frames, factors, time series, etc.). By using as.zoo, as.irts, as.xts, etc. it is possible to work with the other commonly used time series on the C++ side. The main design goal was improved performance. Dominick __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
