[Rd] discrete-time survival model
Dear List, Last week, I asked on R-help if there was a package implementing the dynamic discrete time model of Fahrmeir (1994) - http://www.jstor.org/pss/2336962. Unfortunately, nobody answered. So I start to implement it. Does anyone interested in such development? Or is it useless because someone has already implemented this model? In the coming days, I plan to move my code on R-forge, so people can easily contribute. If you are interested please let me know. Thanks in advance Christophe -- Christophe Dutang Ph.D. student at ISFA, Lyon, France website: http://dutangc.free.fr [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] where is libRmath.a libRmath.so
On 14.05.2010 04:47, Dave Lubbers wrote: I see Rmath.h in include. Why can't I find libRmath.a and/or libRmath.so? Since you have not compiled them yet? See the R Installation and Administration manual. Uwe Ligges __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] Creating an S3 method when the generic function is defined in another (imported) package
Hi, In one of my packages (maxLik), I would like to add an S3 method, where the generic function (estfun) is defined in another package (sandwich). Everything works fine if my package Depends on the other package and I import the generic function estfun from the sandwich package and define the new method in the NAMESPACE file. However, I prefer not to load the entire other package just to use a single generic function from it. Therefore, I replaced Depends by Imports in the DESCRIPTION file and (re-)exported estfun in the NAMESPACE file. Now R CMD check warns about the undocumented code object estfun but I do not want to include a duplicate of the documentation of estfun. Do you know a better solution? Thanks a lot, Arne -- Arne Henningsen http://www.arne-henningsen.name __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Creating an S3 method when the generic function is defined in another (imported) package
On Fri, 14 May 2010, Arne Henningsen wrote: Hi, In one of my packages (maxLik), I would like to add an S3 method, where the generic function (estfun) is defined in another package (sandwich). Everything works fine if my package Depends on the other package and I import the generic function estfun from the sandwich package and define the new method in the NAMESPACE file. However, I prefer not to load the entire other package just to use a single generic function from it. Therefore, I replaced Depends by Imports in the DESCRIPTION file and (re-)exported estfun in the NAMESPACE file. Now R CMD check warns about the undocumented code object estfun but I do not want to include a duplicate of the documentation of estfun. Do you know a better solution? I probably wouldn't re-export the estfun() generic. People will probably load sandwich anyway before calling estfun(), in particular if they want to compute the sandwich() covariance matrix. hth, Z Thanks a lot, Arne -- Arne Henningsen http://www.arne-henningsen.name __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Creating an S3 method when the generic function is defined in another (imported) package
On 14 May 2010 10:37, Achim Zeileis achim.zeil...@uibk.ac.at wrote: On Fri, 14 May 2010, Arne Henningsen wrote: Hi, In one of my packages (maxLik), I would like to add an S3 method, where the generic function (estfun) is defined in another package (sandwich). Everything works fine if my package Depends on the other package and I import the generic function estfun from the sandwich package and define the new method in the NAMESPACE file. However, I prefer not to load the entire other package just to use a single generic function from it. Therefore, I replaced Depends by Imports in the DESCRIPTION file and (re-)exported estfun in the NAMESPACE file. Now R CMD check warns about the undocumented code object estfun but I do not want to include a duplicate of the documentation of estfun. Do you know a better solution? I probably wouldn't re-export the estfun() generic. People will probably load sandwich anyway before calling estfun(), in particular if they want to compute the sandwich() covariance matrix. Good point! Thanks a lot, Arne -- Arne Henningsen http://www.arne-henningsen.name __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] results of pnorm as either NaN or Inf
The answer to pnorm(-x, log.p=TRUE) is of course about -0.5*x*x for large x. So trying to compute it for -1e108 is a bit silly, both on your part and the C code used. I've altered the C code not to attempt it for x 1e170, which excludes the area where underflow occurs. On Fri, 14 May 2010, ted.hard...@manchester.ac.uk wrote: On 13-May-10 20:04:50, efree...@berkeley.edu wrote: I stumbled across this and I am wondering if this is unexpected behavior or if I am missing something. pnorm(-1.0e+307, log.p=TRUE) [1] -Inf pnorm(-1.0e+308, log.p=TRUE) [1] NaN Warning message: In pnorm(q, mean, sd, lower.tail, log.p) : NaNs produced pnorm(-1.0e+309, log.p=TRUE) [1] -Inf I don't know C and am not that skilled with R, so it would be hard for me to look into the code for pnorm. If I'm not just missing something, I thought it may be of interest. Details: I am using Mac OS X 10.5.8. I installed a precompiled binary version. Here is the output from sessionInfo(), requested in the posting guide: R version 2.11.0 (2010-04-22) i386-apple-darwin9.8.0 locale: [1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices utils datasets methods base loaded via a namespace (and not attached): [1] tools_2.11.0 Thank you very much, Eric Freeman UC Berkeley This is probably platform-independent. I get the same results with R on Linux. More to the point: You are clearly pushing the envelope here. First, have a look at what R makes of your inputs to pnorm(): -1.0e+307 # [1] -1e+307 -1.0e+308 # [1] -1e+308 -1.0e+309 # [1] -Inf So, somewhere beteen -1e+308 and -1.0e+309. the envelope burst! Given -1.0e+309, R returns -Inf (i.e. R can no longer represent this internally as a finite number). Now look at pnorm(-Inf,log.p=TRUE) # [1] -Inf So, R knows how to give the correct answer (an exact 0, or -Inf on the log scale) if you feed pnorm() with -Inf. So you're OK with -1e+N where N = 309. For smaller powers, e.g. -1e+(200:306), these give pnorm() much less than -1.0e+309, and presumably R's algorithm (which I haven't studied either ... ) returns 0 for pnorm(), as it should to the available internal accuracy. So, up to pnorm(-1.0e+307, log.p=TRUE) = -Inf. All is as it should be. However, at -1e+308, the envelope is about to burst, and something may occur within the algorithm which results in a NaN. So there is nothing anomalous about your results except at -1e+308, which is where R is at a critical point. That's how I see it, anway! Ted. E-Mail: (Ted Harding) ted.hard...@manchester.ac.uk Fax-to-email: +44 (0)870 094 0861 Date: 14-May-10 Time: 00:01:27 -- XFMail -- __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel -- Brian D. Ripley, rip...@stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UKFax: +44 1865 272595 __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Using Sweave in hostile environments
Just for the archives, a partial answer to my question:
I was unable to find a definition for the Sinput environment that would
work, but relatively minor changes to some of the RweaveLatex functions
cause Sweave to generate SaveVerbatim environments rather than Sinput
environments, and that's enough to get things to work. So the example
below would be handled using
chunkname, eval=FALSE, saveVerbatim=TRUE=
x - 1
@
somewhere safe, and then
\BUseVerbatim[fontshape=sl]{chunkname.in.1}
in the box. Each piece of input or output will be saved in a separate
verbatim piece, and they each need to be used explicitly: you'll
probably want to look at the .tex file to figure out the names.
This is pretty rough and I won't commit it to R, but I've attached the
patch file (against R trunk rev 51988) in case anyone else wants to use it.
Duncan Murdoch
On 13/05/2010 10:19 AM, Duncan Murdoch wrote:
I'm trying to put together a poster using the LaTeX a0poster package and
including some things from pstricks to get gradient shading, etc.
The problem is that the default environments used by Sweave don't work
where I need them. A simple code chunk like
eval=FALSE=
x - 1
@
buried in a minipage within a psshadowbox gives the error
Runaway argument?
x - 1 \end {Sinput} \end {Schunk} and the resulting variable would\ETC.
D:/test.tex:302: Paragraph ended before \FV@
BeginScanning was complete
Has anyone solved this problem? One way I see that might work is to put
the code chunk outside the environment and use SaveVerbatim to save it,
then just put \BUseVerbatim or \LUseVerbatim in place where I need it.
But because I'm using Sweave, I somehow need to redefine the Sinput,
Soutput and maybe the Schunk environments to do that, and I don't see
how to do so.
Duncan Murdoch
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Index: Sweave.R
===
--- Sweave.R(revision 51988)
+++ Sweave.R(working copy)
@@ -368,7 +368,8 @@
split=FALSE, strip.white=true, include=TRUE,
pdf.version=grDevices::pdf.options()$version,
pdf.encoding=grDevices::pdf.options()$encoding,
-concordance=FALSE, expand=TRUE)
+concordance=FALSE, expand=TRUE,
+saveVerbatim=FALSE)
options[names(dots)] - dots
## to be on the safe side: see if defaults pass the check
@@ -482,8 +483,12 @@
thisline - thisline + 1
openSchunk - TRUE
}
-cat(\\begin{Sinput},
-file=chunkout, append=TRUE)
+if (options$saveVerbatim)
+cat(\\begin{SaveVerbatim}{, options$label,
.in., nce, }, sep=,
+file=chunkout, append=TRUE)
+else
+cat(\\begin{Sinput},
+ file=chunkout, append=TRUE)
openSinput - TRUE
}
cat(\n, paste(getOption(prompt), dce[1L:leading],
sep=, collapse=\n),
@@ -516,7 +521,10 @@
if(length(output) (options$results != hide)){
if(openSinput){
-cat(\n\\end{Sinput}\n, file=chunkout, append=TRUE)
+ if (options$saveVerbatim)
+ cat(\n\\end{SaveVerbatim}\n, file=chunkout,
append=TRUE)
+ else
+cat(\n\\end{Sinput}\n, file=chunkout,
append=TRUE)
linesout[thisline + 1L:2L] - srcline
thisline - thisline + 2L
openSinput - FALSE
@@ -529,8 +537,12 @@
thisline - thisline + 1L
openSchunk - TRUE
}
-cat(\\begin{Soutput}\n,
-file=chunkout, append=TRUE)
+if(options$saveVerbatim)
+cat(\\begin{SaveVerbatim}{, options$label,
.out., nce, }\n,
+ sep=, file=chunkout, append=TRUE)
+else
+cat(\\begin{Soutput}\n,
+ file=chunkout, append=TRUE)
linesout[thisline + 1L] - srcline
thisline - thisline + 1L
}
@@ -552,7 +564,10 @@
remove(output)
if(options$results==verbatim){
-cat(\n\\end{Soutput}\n, file=chunkout, append=TRUE)
+ if (options$saveVerbatim)
+ cat(\n\\end{SaveVerbatim}\n, file=chunkout,
[Rd] Compiling R with ATLAS
Hello. I know almost nothing about Linux, so I apologize if the answer to this is obvious. I am trying to compile R with ATLAS (successfully compiled 3.9.24 on Ubuntu 10.04 64 bit). I configured with: ./configure --enable-R-shlib --enable-BLAS-shlib --with-blas=-L/usr/local/atlas/lib -lf77blas -latlas --with-lapack=-L/usr/local/atlaslib -llapack -lcblas --with-x and I get this error (from config.log): configure:28566: checking for dgemm_ in -L/usr/local/atlas/lib -lf77blas -latlas configure:28587: gcc -std=gnu99 -o conftest -g -O2 -fpic -I/usr/local/include -L/usr/local/lib64 conftest.c -L/usr/local/atlas/lib -lf77blas -latlas -lgfortran -lm -ldl -lm 5 conftest.c: In function 'main': conftest.c:193: warning: implicit declaration of function 'dgemm_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2u_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2u_' collect2: ld returned 1 exit status configure:28587: $? = 1 I found the following e-mail in the archives http://www.mail-archive.com/r-devel@r-project.org/msg16853.html, but it has a different error, and the suggestion of switching out the BLAS later does not seem to apply to ATLAS which requires two files, I think. Any advice would be very appreciated. Thank you. [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Compiling R with ATLAS
There is a Debian/Ubuntu specific list, r-sig-deb...@r-project.org, which I am cc:'ing on this reply, that is a better location for this discussion. It appears that you may be going about things the hard way. There are Ubuntu packages for atlas and for R that can handle all of this for you. Take a look at http://cran.r-project.org/bin/linux/ubuntu for information on how to get started. It will probably be much easier to install the pre-compiled packages first and worry about compilation for customization later. On Thu, May 13, 2010 at 11:16 PM, Avraham Adler avraham.ad...@gmail.com wrote: Hello. I know almost nothing about Linux, so I apologize if the answer to this is obvious. I am trying to compile R with ATLAS (successfully compiled 3.9.24 on Ubuntu 10.04 64 bit). I configured with: ./configure --enable-R-shlib --enable-BLAS-shlib --with-blas=-L/usr/local/atlas/lib -lf77blas -latlas --with-lapack=-L/usr/local/atlaslib -llapack -lcblas --with-x and I get this error (from config.log): configure:28566: checking for dgemm_ in -L/usr/local/atlas/lib -lf77blas -latlas configure:28587: gcc -std=gnu99 -o conftest -g -O2 -fpic -I/usr/local/include -L/usr/local/lib64 conftest.c -L/usr/local/atlas/lib -lf77blas -latlas -lgfortran -lm -ldl -lm 5 conftest.c: In function 'main': conftest.c:193: warning: implicit declaration of function 'dgemm_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2u_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2u_' collect2: ld returned 1 exit status configure:28587: $? = 1 I found the following e-mail in the archives http://www.mail-archive.com/r-devel@r-project.org/msg16853.html, but it has a different error, and the suggestion of switching out the BLAS later does not seem to apply to ATLAS which requires two files, I think. Any advice would be very appreciated. Thank you. [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] update.packages fails with directory not found
Bill, Thanks. The problem is common but not consistent on update. I'll wait a few days for packages to be updated at CRAN, then try it. Mike On Wed, 12 May 2010 08:49:42 -0700, William Dunlap wdun...@tibco.com wrote: If you have the time and can consistently reproduce this problem you might try using Sysinternal's Process Monitor program to capture the details of what R and other programs were doing in that directory around the time of the problem. http://technet.microsoft.com/en-us/sysinternals/bb896645.aspx __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
This isn't like a local variable though, since any function above the
baseenv() in the search path will also not be found.
f
function(a) { rnorm(b) }
environment: base
f()
Error in f() : could not find function rnorm
Jeff
On Fri, May 14, 2010 at 9:20 AM, Hadley Wickham had...@rice.edu wrote:
Hello All,
By default, a reference of a variable in a function cause R to look
for the variable in the parent environment if it is not available in
the current environment (without generating any errors or warnings).
I'm wondering if there is a way to revert this behaviors, such that it
will not look for the parent environment and will generate an error if
the variable is not available in the current environment. Is this
tuning has to be done at the C level?
Try this:
b - 1
f - function(a) {
b
}
environment(f) - baseenv()
f()
Hadley
--
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Jeffrey Ryan
jeffrey.r...@insightalgo.com
ia: insight algorithmics
www.insightalgo.com
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
You can get around that by using this instead:
environment(f) - as.environment(2)
provided it is done after you have loaded all your packages.
On Fri, May 14, 2010 at 10:29 AM, Jeff Ryan jeff.a.r...@gmail.com wrote:
This isn't like a local variable though, since any function above the
baseenv() in the search path will also not be found.
f
function(a) { rnorm(b) }
environment: base
f()
Error in f() : could not find function rnorm
Jeff
On Fri, May 14, 2010 at 9:20 AM, Hadley Wickham had...@rice.edu wrote:
Hello All,
By default, a reference of a variable in a function cause R to look
for the variable in the parent environment if it is not available in
the current environment (without generating any errors or warnings).
I'm wondering if there is a way to revert this behaviors, such that it
will not look for the parent environment and will generate an error if
the variable is not available in the current environment. Is this
tuning has to be done at the C level?
Try this:
b - 1
f - function(a) {
b
}
environment(f) - baseenv()
f()
Hadley
--
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Jeffrey Ryan
jeffrey.r...@insightalgo.com
ia: insight algorithmics
www.insightalgo.com
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
On Fri, May 14, 2010 at 10:10 AM, Gabor Grothendieck ggrothendi...@gmail.com wrote: You can get around that by using this instead: environment(f) - as.environment(2) provided it is done after you have loaded all your packages. Another approach is to inspect function by function: library(codetools) findGlobals(f, F)$variables Hadley -- Assistant Professor / Dobelman Family Junior Chair Department of Statistics / Rice University http://had.co.nz/ __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
On May 14, 2010, at 10:29 AM, Jeff Ryan wrote:
This isn't like a local variable though, since any function above the
baseenv() in the search path will also not be found.
Yes, but that is a consequence of the request and hence intended. You can
always either specify the full path to the function or assign locally any
functions you'll need:
f = function(a) stats::rnorm(b)
environment(f) - baseenv()
f()
Error in stats::rnorm(b) : object 'b' not found
f = function(a) { rnorm=stats::rnorm; rnorm(b) }
environment(f) - baseenv()
f()
Error in rnorm(b) : object 'b' not found
Alternatively you can use any environment up the search path hoping that those
won't define further variables.
But back to the original question -- if taken literally, it's quite useless
since it corresponds to using emptyenv() instead of baseenv() which means that
you can't do anything:
local(1+1,emptyenv())
Error in eval(expr, envir, enclos) : could not find function +
So likely the only useful(?) distinction would be to allow function lookup the
regular way but change only the variable lookup to not look beyond the current
environment. That is stretching it, though...
Cheers,
Simon
f
function(a) { rnorm(b) }
environment: base
f()
Error in f() : could not find function rnorm
Jeff
On Fri, May 14, 2010 at 9:20 AM, Hadley Wickham had...@rice.edu wrote:
Hello All,
By default, a reference of a variable in a function cause R to look
for the variable in the parent environment if it is not available in
the current environment (without generating any errors or warnings).
I'm wondering if there is a way to revert this behaviors, such that it
will not look for the parent environment and will generate an error if
the variable is not available in the current environment. Is this
tuning has to be done at the C level?
Try this:
b - 1
f - function(a) {
b
}
environment(f) - baseenv()
f()
Hadley
--
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Jeffrey Ryan
jeffrey.r...@insightalgo.com
ia: insight algorithmics
www.insightalgo.com
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
On Fri, May 14, 2010 at 10:13 AM, Simon Urbanek
simon.urba...@r-project.org wrote:
On May 14, 2010, at 10:29 AM, Jeff Ryan wrote:
This isn't like a local variable though, since any function above the
baseenv() in the search path will also not be found.
Yes, but that is a consequence of the request and hence intended. You can
always either specify the full path to the function or assign locally any
functions you'll need:
f = function(a) stats::rnorm(b)
environment(f) - baseenv()
f()
Error in stats::rnorm(b) : object 'b' not found
f = function(a) { rnorm=stats::rnorm; rnorm(b) }
environment(f) - baseenv()
f()
Error in rnorm(b) : object 'b' not found
Alternatively you can use any environment up the search path hoping that
those won't define further variables.
But back to the original question -- if taken literally, it's quite useless
since it corresponds to using emptyenv() instead of baseenv() which means
that you can't do anything:
local(1+1,emptyenv())
Error in eval(expr, envir, enclos) : could not find function +
So likely the only useful(?) distinction would be to allow function lookup
the regular way but change only the variable lookup to not look beyond the
current environment. That is stretching it, though...
I didn't think of the corner cases. I'm OK with stats::rnorm or
stats:::rnorm. But '{' is not recognized. How to make it be
recognized?
b=1
f=function() {
stats::rnorm(b)
}
f()
#environment(f)=base()
environment(f)=emptyenv()
f()
Cheers,
Simon
f
function(a) { rnorm(b) }
environment: base
f()
Error in f() : could not find function rnorm
Jeff
On Fri, May 14, 2010 at 9:20 AM, Hadley Wickham had...@rice.edu wrote:
Hello All,
By default, a reference of a variable in a function cause R to look
for the variable in the parent environment if it is not available in
the current environment (without generating any errors or warnings).
I'm wondering if there is a way to revert this behaviors, such that it
will not look for the parent environment and will generate an error if
the variable is not available in the current environment. Is this
tuning has to be done at the C level?
Try this:
b - 1
f - function(a) {
b
}
environment(f) - baseenv()
f()
Hadley
--
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Jeffrey Ryan
jeffrey.r...@insightalgo.com
ia: insight algorithmics
www.insightalgo.com
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Tom
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] results of pnorm as either NaN or Inf
Thank you for your responses. I presented a minimal example of the issue, but I should have explained that this came up in the context of maximizing a log likelihood function (with optim). I certainly agree that there would be no good reason for a human to evaluate the function pnorm(-x, log.p=TRUE) for large x. However, I would suggest that one may want the function to return a value of -Inf for large x, rather than issue an error or a warning, as I'm guessing your alteration of the C code would do. Returning a value of -Inf would allow an optimization routine (or a pre-optimization-routine grid search) to know that it's going into (or is currently in) the wrong region. I can definitely see the counterargument that someone using an optimization routine should think about the function they are optimizing more carefully, especially given that there could be numerical issues. I think I'm not the right person to judge, both because I don't know the programming issues nor much about numerical optimization, but I wanted to offer a possible argument for returning -Inf in case it hasn't already been considered, and others can evaluate the issue. (In general, my earlier purpose was just to point out what I thought might be an inconsistency in case people wanted to change it. I'm not concerned about it for my own usage at all, since I can just adjust my own programs.) Thank you, and to all, thank you for all of your work on R. Eric On Fri, 14 May 2010 11:50:09 +0100 (BST), Prof Brian Ripley rip...@stats.ox.ac.uk wrote: The answer to pnorm(-x, log.p=TRUE) is of course about -0.5*x*x for large x. So trying to compute it for -1e108 is a bit silly, both on your part and the C code used. I've altered the C code not to attempt it for x 1e170, which excludes the area where underflow occurs. On Fri, 14 May 2010, ted.hard...@manchester.ac.uk wrote: On 13-May-10 20:04:50, efree...@berkeley.edu wrote: I stumbled across this and I am wondering if this is unexpected behavior or if I am missing something. pnorm(-1.0e+307, log.p=TRUE) [1] -Inf pnorm(-1.0e+308, log.p=TRUE) [1] NaN Warning message: In pnorm(q, mean, sd, lower.tail, log.p) : NaNs produced pnorm(-1.0e+309, log.p=TRUE) [1] -Inf I don't know C and am not that skilled with R, so it would be hard for me to look into the code for pnorm. If I'm not just missing something, I thought it may be of interest. Details: I am using Mac OS X 10.5.8. I installed a precompiled binary version. Here is the output from sessionInfo(), requested in the posting guide: R version 2.11.0 (2010-04-22) i386-apple-darwin9.8.0 locale: [1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices utils datasets methods base loaded via a namespace (and not attached): [1] tools_2.11.0 Thank you very much, Eric Freeman UC Berkeley This is probably platform-independent. I get the same results with R on Linux. More to the point: You are clearly pushing the envelope here. First, have a look at what R makes of your inputs to pnorm(): -1.0e+307 # [1] -1e+307 -1.0e+308 # [1] -1e+308 -1.0e+309 # [1] -Inf So, somewhere beteen -1e+308 and -1.0e+309. the envelope burst! Given -1.0e+309, R returns -Inf (i.e. R can no longer represent this internally as a finite number). Now look at pnorm(-Inf,log.p=TRUE) # [1] -Inf So, R knows how to give the correct answer (an exact 0, or -Inf on the log scale) if you feed pnorm() with -Inf. So you're OK with -1e+N where N = 309. For smaller powers, e.g. -1e+(200:306), these give pnorm() much less than -1.0e+309, and presumably R's algorithm (which I haven't studied either ... ) returns 0 for pnorm(), as it should to the available internal accuracy. So, up to pnorm(-1.0e+307, log.p=TRUE) = -Inf. All is as it should be. However, at -1e+308, the envelope is about to burst, and something may occur within the algorithm which results in a NaN. So there is nothing anomalous about your results except at -1e+308, which is where R is at a critical point. That's how I see it, anway! Ted. E-Mail: (Ted Harding) ted.hard...@manchester.ac.uk Fax-to-email: +44 (0)870 094 0861 Date: 14-May-10 Time: 00:01:27 -- XFMail -- __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Compiling R with ATLAS
Thank you. I am doing it this way as I really do not know much about Linux, and I would not know how to re-compile Rblas and Rlapack with the ATLAS BLAS and tuned LAPACK after I install the precompiled packages through synaptic. This way, ostensibly, I should be compiling it from scratch so it should work. Also, this way, I can compile R itself with optimization and -march=native to squeeze out whatever speed I can from my machine. Any advice would be very much appreciated, and thank you for pointing me to the correct list. --Avi On Fri, May 14, 2010 at 9:00 AM, Douglas Bates ba...@stat.wisc.edu wrote: There is a Debian/Ubuntu specific list, r-sig-deb...@r-project.org, which I am cc:'ing on this reply, that is a better location for this discussion. It appears that you may be going about things the hard way. There are Ubuntu packages for atlas and for R that can handle all of this for you. Take a look at http://cran.r-project.org/bin/linux/ubuntu for information on how to get started. It will probably be much easier to install the pre-compiled packages first and worry about compilation for customization later. On Thu, May 13, 2010 at 11:16 PM, Avraham Adler avraham.ad...@gmail.com wrote: Hello. I know almost nothing about Linux, so I apologize if the answer to this is obvious. I am trying to compile R with ATLAS (successfully compiled 3.9.24 on Ubuntu 10.04 64 bit). I configured with: ./configure --enable-R-shlib --enable-BLAS-shlib --with-blas=-L/usr/local/atlas/lib -lf77blas -latlas --with-lapack=-L/usr/local/atlaslib -llapack -lcblas --with-x and I get this error (from config.log): configure:28566: checking for dgemm_ in -L/usr/local/atlas/lib -lf77blas -latlas configure:28587: gcc -std=gnu99 -o conftest -g -O2 -fpic -I/usr/local/include -L/usr/local/lib64 conftest.c -L/usr/local/atlas/lib -lf77blas -latlas -lgfortran -lm -ldl -lm 5 conftest.c: In function 'main': conftest.c:193: warning: implicit declaration of function 'dgemm_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2u_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2u_' collect2: ld returned 1 exit status configure:28587: $? = 1 I found the following e-mail in the archives http://www.mail-archive.com/r-devel@r-project.org/msg16853.html, but it has a different error, and the suggestion of switching out the BLAS later does not seem to apply to ATLAS which requires two files, I think. Any advice would be very appreciated. Thank you. [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] Fwd: Compiling R with ATLAS
(Resent as plain text, sorry) Thank you. I am doing it this way as I really do not know much about Linux, and I would not know how to re-compile Rblas and Rlapack with the ATLAS BLAS and tuned LAPACK after I install the precompiled packages through synaptic. This way, ostensibly, I should be compiling it from scratch so it should work. Also, this way, I can compile R itself with optimization and -march=native to squeeze out whatever speed I can from my machine. Any advice would be very much appreciated, and thank you for pointing me to the correct list. --Avi On Fri, May 14, 2010 at 9:00 AM, Douglas Bates ba...@stat.wisc.edu wrote: There is a Debian/Ubuntu specific list, r-sig-deb...@r-project.org, which I am cc:'ing on this reply, that is a better location for this discussion. It appears that you may be going about things the hard way. There are Ubuntu packages for atlas and for R that can handle all of this for you. Take a look at http://cran.r-project.org/bin/linux/ubuntu for information on how to get started. It will probably be much easier to install the pre-compiled packages first and worry about compilation for customization later. On Thu, May 13, 2010 at 11:16 PM, Avraham Adler avraham.ad...@gmail.com wrote: Hello. I know almost nothing about Linux, so I apologize if the answer to this is obvious. I am trying to compile R with ATLAS (successfully compiled 3.9.24 on Ubuntu 10.04 64 bit). I configured with: ./configure --enable-R-shlib --enable-BLAS-shlib --with-blas=-L/usr/local/atlas/lib -lf77blas -latlas --with-lapack=-L/usr/local/atlaslib -llapack -lcblas --with-x and I get this error (from config.log): configure:28566: checking for dgemm_ in -L/usr/local/atlas/lib -lf77blas -latlas configure:28587: gcc -std=gnu99 -o conftest -g -O2 -fpic -I/usr/local/include -L/usr/local/lib64 conftest.c -L/usr/local/atlas/lib -lf77blas -latlas -lgfortran -lm -ldl -lm 5 conftest.c: In function 'main': conftest.c:193: warning: implicit declaration of function 'dgemm_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2u_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_zger2c_' /usr/local/atlas/lib/libf77blas.so: undefined reference to `atl_f77wrap_cger2u_' collect2: ld returned 1 exit status configure:28587: $? = 1 I found the following e-mail in the archives http://www.mail-archive.com/r-devel@r-project.org/msg16853.html, but it has a different error, and the suggestion of switching out the BLAS later does not seem to apply to ATLAS which requires two files, I think. Any advice would be very appreciated. Thank you. [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] [R] Subscripting a matrix-like object
-Original Message-
From: r-help-boun...@r-project.org
[mailto:r-help-boun...@r-project.org] On Behalf Of Jeffrey J. Hallman
Sent: Friday, May 14, 2010 9:17 AM
To: r-h...@stat.math.ethz.ch
Subject: Re: [R] Subscripting a matrix-like object
jhall...@frb.gov writes:
Answering my own question here, so you can ignore this unless you are
really interested in some fairly obscure stuff. It turns out
that this works:
singleIndex - missing(j) (length(sys.call()) ==
length(match.call()))
since sys.call() has an element for the empty argument created by
x[i,] - value
and match.call() does not. But it is pretty obscure.
I was curious about this because in the version of R
I'm using (2.11.0, 2010-04-22), nargs() does distinguish
between x[i] and x[i,]. E.g., with the following
replacement function
`[-.foo` - function (x, i, j, ..., value)
{
cat([-.foo: nargs=, nargs(), \n, sep = )
cat( sys.call()=, deparse(sys.call()), \n, sep = )
cat( missing(i)=, missing(i), \n, sep = )
cat( missing(j)=, missing(j), \n, sep = )
x
}
I get
fooObject - structure(1:4, class=foo)
fooObject[1] - 12
[-.foo: nargs=3
sys.call()=`[-.foo`(`*tmp*`, 1, value = 12)
missing(i)=FALSE
missing(j)=TRUE
fooObject[1,] - 12
[-.foo: nargs=4
sys.call()=`[-.foo`(`*tmp*`, 1, , value = 12)
missing(i)=FALSE
missing(j)=TRUE
fooObject[,1] - 12
[-.foo: nargs=4
sys.call()=`[-.foo`(`*tmp*`, , 1, value = 12)
missing(i)=TRUE
missing(j)=FALSE
Is this a bug in nargs() that was recently fixed?
sys.call() and especially match.call() are
fairly expensive functions to call so it would be nice
to be able to avoid them.
Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com
Jeff
I have an S3 class called tis (Time Indexed Series) which
may or may
not have multiple columns. I have a function [-.tis that I've
reproduced below.
My question is this: inside of [-.tis, how can I
distinguish between
calls of the form
x[i] - someValue
and
x[i,] - someValue ?
In either case, nargs() is 3, and looking at the values
from sys.call()
and match.call() I could not tell them apart. Am I missing
something?
[-.tis - function(x, i, j, ..., value){
tif - tif(x)
xStart - start(x)
x - stripTis(x)
if(missing(i)){
if(missing(j)) x[] - value
else x[,j] - value
}
else {
i - i[!is.na(i)]
if(is.numeric(i)){
if(!is.ti(i) couldBeTi(i, tif = tif))
i - asTi(i)
if(is.ti(i)){
i - i + 1 - xStart
if(any(i 1)){
newRows - 1 - min(i)
xStart - xStart - newRows
if(is.null(m - ncol(x))) m - 1
i - i + newRows
if(is.matrix(x))
x - rbind(matrix(NA, newRows, m), x)
else
x - c(rep(NA, newRows), x)
}
}
}
else if(!is.logical(i)) stop(non-numeric, non-logical
row index)
if(is.matrix(x)){
if(any(i nrow(x))){
newRows - max(i) - nrow(x)
x - rbind(x, matrix(NA, newRows, ncol(x)))
}
if(missing(j)){
if(is.matrix(i)) x[i] - value
else {
if(is.logical(i))
x[i,] - rep(value, length = sum(i)*ncol(x))
else
x[i,] - rep(value, length = length(i)*ncol(x))
}
}
else x[i,j] - value
}
else x[i] - value
}
start(x) - xStart
class(x) - c(tis, oldClass(x))
x
}
--
Jeff
__
r-h...@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] S4 default/missing args
I am trying to understand the missing vs default value for an argument
to an S4 method. (I'm not sure if this is a bug or my confusion.) In S3
I can specify a default value for an argument and then both check if the
argument was missing in the call, and use it because it takes on the
default value. I do not seem to be able to do that in S4 as shown by
this simplified example:
setGeneric(TSdoc,
def= function(x, con, ...) standardGeneric(TSdoc))
setMethod(TSdoc, signature(x=character, con=missing),
definition= function(x, con=getOption(TSconnection), ...){
if(is.null(con))
stop(con should be specified or set with
options(TSconnection=con).)
else return('ok')} )
Then when I try
TSdoc(B103)
In my original problem I got
Error in is.null(con) : 'con' is missing
And in the simplified example I get
Error in TSdoc(B103) :
element 1 is empty;
the part of the args list of 'is.null' being evaluated was:
(con)
I'm not sure why there is a difference, but the implication is the same
in both, the default
getOption(TSconnection)
NULL
Is not getting through to the is.null(). I can fix (or work around)
this with
setMethod(TSdoc, signature(x=character, con=missing),
definition= function(x, con, ...){
if(is.null(con - getOption(TSconnection)))
stop(con should be specified or set with
options(TSconnection=con).)
else return('ok') } )
TSdoc(B103)
Error in TSdoc(B103) :
con should be specified or set with options(TSconnection=con).
So my question is, is this a bug or do I have to adjust my programming
style for the way missing/default arguments are handled in S4?
Paul Gilbert
La version française suit le texte anglais.
This email may contain privileged and/or confidential information, and the Bank
of
Canada does not waive any related rights. Any distribution, use, or copying of
this
email or the information it contains by other than the intended recipient is
unauthorized. If you received this email in error please delete it immediately
from
your system and notify the sender promptly by email that you have done so.
Le présent courriel peut contenir de l'information privilégiée ou
confidentielle.
La Banque du Canada ne renonce pas aux droits qui s'y rapportent. Toute
diffusion,
utilisation ou copie de ce courriel ou des renseignements qu'il contient par une
personne autre que le ou les destinataires désignés est interdite. Si vous
recevez
ce courriel par erreur, veuillez le supprimer immédiatement et envoyer sans
délai Ã
l'expéditeur un message électronique pour l'aviser que vous avez éliminé de
votre
ordinateur toute copie du courriel reçu.
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
On May 14, 2010, at 11:33 AM, thmsfuller...@gmail.com wrote:
On Fri, May 14, 2010 at 10:13 AM, Simon Urbanek
simon.urba...@r-project.org wrote:
On May 14, 2010, at 10:29 AM, Jeff Ryan wrote:
This isn't like a local variable though, since any function above the
baseenv() in the search path will also not be found.
Yes, but that is a consequence of the request and hence intended. You can
always either specify the full path to the function or assign locally any
functions you'll need:
f = function(a) stats::rnorm(b)
environment(f) - baseenv()
f()
Error in stats::rnorm(b) : object 'b' not found
f = function(a) { rnorm=stats::rnorm; rnorm(b) }
environment(f) - baseenv()
f()
Error in rnorm(b) : object 'b' not found
Alternatively you can use any environment up the search path hoping that
those won't define further variables.
But back to the original question -- if taken literally, it's quite useless
since it corresponds to using emptyenv() instead of baseenv() which means
that you can't do anything:
local(1+1,emptyenv())
Error in eval(expr, envir, enclos) : could not find function +
So likely the only useful(?) distinction would be to allow function lookup
the regular way but change only the variable lookup to not look beyond the
current environment. That is stretching it, though...
I didn't think of the corner cases. I'm OK with stats::rnorm or
stats:::rnorm. But '{' is not recognized. How to make it be
recognized?
You either include base or (if you don't want to) you have to create functions
for everything you'll need including `{` so for example:
f=function(a) base::rnorm(b)
environment(f)=mini.env
mini.env=new.env(parent=emptyenv())
import=c({,::,+,rnorm) # add all functions you want to use
for (fn in import) mini.env[[fn]] = get(fn)
f=function(a) rnorm(a+b)
environment(f)=mini.env
f(1)
Error in rnorm(a + b) : object 'b' not found
Cheers,
Simon
b=1
f=function() {
stats::rnorm(b)
}
f()
#environment(f)=base()
environment(f)=emptyenv()
f()
Cheers,
Simon
f
function(a) { rnorm(b) }
environment: base
f()
Error in f() : could not find function rnorm
Jeff
On Fri, May 14, 2010 at 9:20 AM, Hadley Wickham had...@rice.edu wrote:
Hello All,
By default, a reference of a variable in a function cause R to look
for the variable in the parent environment if it is not available in
the current environment (without generating any errors or warnings).
I'm wondering if there is a way to revert this behaviors, such that it
will not look for the parent environment and will generate an error if
the variable is not available in the current environment. Is this
tuning has to be done at the C level?
Try this:
b - 1
f - function(a) {
b
}
environment(f) - baseenv()
f()
Hadley
--
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Jeffrey Ryan
jeffrey.r...@insightalgo.com
ia: insight algorithmics
www.insightalgo.com
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Tom
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Restrict access to variables in parent environment
On 14/05/2010 10:14 AM, thmsfuller...@gmail.com wrote: Hello All, By default, a reference of a variable in a function cause R to look for the variable in the parent environment if it is not available in the current environment (without generating any errors or warnings). I'm wondering if there is a way to revert this behaviors, such that it will not look for the parent environment and will generate an error if the variable is not available in the current environment. Is this tuning has to be done at the C level? You could do that by setting the environment of the function to emptyenv(), but it will not be pretty. Remember that everything in R is an object, so you won't have access to the base level objects like +, or mean, or any other function. Duncan Murdoch __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] The parsing of '{' and a function that equal to '{'
Hello All,
I tried the sample code from the help. Although '{' is assigned to
'do', the call syntaxes for 'do' and '{' are not the same ('do' has
','s, but '{' has line breaks). I guess there is a difference in
parsing the code block of 'do' and the code block of '{'. Could you
please let me know some internal details so that I can understand the
differences?
do=get({)
do(x - 3, y - 2*x-3, 6-x-y)
[1] 0
{
+ x - 3
+ y - 2*x-3
+ 6-x-y
+ }
[1] 0
x
[1] 3
y
[1] 3
--
Tom
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] The parsing of '{' and a function that equal to '{'
On May 14, 2010, at 3:38 PM, thmsfuller...@gmail.com wrote:
Hello All,
I tried the sample code from the help. Although '{' is assigned to
'do', the call syntaxes for 'do' and '{' are not the same ('do' has
','s, but '{' has line breaks). I guess there is a difference in
parsing the code block of 'do' and the code block of '{'. Could you
please let me know some internal details so that I can understand the
differences?
{ foo; bar }
gets parsed into
`{`(foo, bar)
as you illustrated below. You can see that in the resulting expression:
e=parse(text={foo; bar})[[1]]
attributes(e)=NULL ## remove the original sources
e
{
foo
bar
}
.Internal(inspect(e))
@102182578 06 LANGSXP g0c0 [NAM(2)]
@10082c9b8 01 SYMSXP g0c0 [MARK,gp=0x4000] {
@100edfaa0 01 SYMSXP g0c0 [] foo
@100edfc98 01 SYMSXP g0c0 [] bar
and in case you are not familiar with the internal representation:
quote(`{`(foo,bar))
{
foo
bar
}
.Internal(inspect(quote(`{`(foo,bar
@100f99868 06 LANGSXP g0c0 []
@10082c9b8 01 SYMSXP g0c0 [MARK,gp=0x4000] {
@100edfaa0 01 SYMSXP g0c0 [] foo
@100edfc98 01 SYMSXP g0c0 [] bar
Cheers,
Simon
do=get({)
do(x - 3, y - 2*x-3, 6-x-y)
[1] 0
{
+ x - 3
+ y - 2*x-3
+ 6-x-y
+ }
[1] 0
x
[1] 3
y
[1] 3
--
Tom
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] where is libRmath.a libRmath.so
R 2.7.2 - the manual says configure make Which is what I did. So I did read the manual and followed the directions. The manual is too terse to get me there. -- View this message in context: http://r.789695.n4.nabble.com/where-is-libRmath-a-libRmath-so-tp2216048p2217224.html Sent from the R devel mailing list archive at Nabble.com. __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] S4 default/missing args
From the documentation for setMethod():
Method definitions can have default expressions for arguments, but a
current limitation is that the generic function must have some default
expression for the same argument in order for the method's defaults to
be used.
(It would be nice to fix this, but would require some significant change
to how methods are dispatched, with some possible inefficiencies. The
issue is that R processes formal arguments differently if there is no
default expression. Volunteers among the R hackers welcome, though.)
Meanwhile, if you own the generic function just give it a default, say:
setGeneric(TSdoc,
def= function(x, con = stop(Sorry, no generic default), ...)
standardGeneric(TSdoc))
John
On 5/14/10 9:50 AM, Paul Gilbert wrote:
I am trying to understand the missing vs default value for an argument
to an S4 method. (I'm not sure if this is a bug or my confusion.) In S3
I can specify a default value for an argument and then both check if the
argument was missing in the call, and use it because it takes on the
default value. I do not seem to be able to do that in S4 as shown by
this simplified example:
setGeneric(TSdoc,
def= function(x, con, ...) standardGeneric(TSdoc))
setMethod(TSdoc, signature(x=character, con=missing),
definition= function(x, con=getOption(TSconnection), ...){
if(is.null(con))
stop(con should be specified or set with
options(TSconnection=con).)
else return('ok')} )
Then when I try
TSdoc(B103)
In my original problem I got
Error in is.null(con) : 'con' is missing
And in the simplified example I get
Error in TSdoc(B103) :
element 1 is empty;
the part of the args list of 'is.null' being evaluated was:
(con)
I'm not sure why there is a difference, but the implication is the same
in both, the default
getOption(TSconnection)
NULL
Is not getting through to the is.null(). I can fix (or work around)
this with
setMethod(TSdoc, signature(x=character, con=missing),
definition= function(x, con, ...){
if(is.null(con- getOption(TSconnection)))
stop(con should be specified or set with
options(TSconnection=con).)
else return('ok') } )
TSdoc(B103)
Error in TSdoc(B103) :
con should be specified or set with options(TSconnection=con).
So my question is, is this a bug or do I have to adjust my programming
style for the way missing/default arguments are handled in S4?
Paul Gilbert
La version française suit le texte anglais.
This email may contain privileged and/or confidential information, and the Bank
of
Canada does not waive any related rights. Any distribution, use, or copying of
this
email or the information it contains by other than the intended recipient is
unauthorized. If you received this email in error please delete it immediately
from
your system and notify the sender promptly by email that you have done so.
Le présent courriel peut contenir de l'information privilégiée ou
confidentielle.
La Banque du Canada ne renonce pas aux droits qui s'y rapportent. Toute
diffusion,
utilisation ou copie de ce courriel ou des renseignements qu'il contient par une
personne autre que le ou les destinataires désignés est interdite. Si vous
recevez
ce courriel par erreur, veuillez le supprimer immédiatement et envoyer sans
délai Ã
l'expéditeur un message électronique pour l'aviser que vous avez éliminé de
votre
ordinateur toute copie du courriel reçu.
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] where is libRmath.a libRmath.so
On May 14, 2010, at 4:52 PM, Dave Lubbers wrote: R 2.7.2 - the manual says configure make Which is what I did. So I did read the manual and followed the directions. The manual is too terse to get me there. cd src/nmath/standalone make Cheers, Simon -- View this message in context: http://r.789695.n4.nabble.com/where-is-libRmath-a-libRmath-so-tp2216048p2217224.html Sent from the R devel mailing list archive at Nabble.com. __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] where is libRmath.a libRmath.so
On 14 May 2010 at 13:52, Dave Lubbers wrote: | | R 2.7.2 - the manual says | configure | make | | Which is what I did. So I did read the manual and followed the directions. | The manual is too terse to get me there. On Debian / Ubuntu, you can also do sudo apt-get install r-mathlib to install this library. -- Regards, Dirk __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] where is libRmath.a libRmath.so
On 14/05/2010 4:52 PM, Dave Lubbers wrote: R 2.7.2 - the manual says configure make Which is what I did. So I did read the manual and followed the directions. The manual is too terse to get me there. You used the wrong tense. In referring to R 2.7.2, only past tenses are grammatically correct. If you want to say the manual *is* too terse, then you need to install R 2.11.0. Duncan Murdoch __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
