Re: [Rd] strange bug? with R CMD check
Yes, I managed to work around the problem. You can still reproduce the strange behavior with revision 5 of the package though. svn checkout svn://svn.r-forge.r-project.org/svnroot/sbr/p...@5 safeBinaryRegression Kjell On 22 juil. 09, at 07:31, Paul Johnson wrote: I am guessing you solved this, because I just downloaded your svn files and there's no more "sbr" and the check passes on my system. Right? pj On Tue, Jul 21, 2009 at 8:32 AM, Kjell Konis wrote: Hello, I am trying to get a package to pass R CMD check on an iMac running Mac OS X. When the package is named safeBinaryRegression I get the following warning from R CMD check: * checking whether the name space can be loaded with stated dependencies ... WARNING Error in dyn.load(file, DLLpath = DLLpath, ...) : function 'make_lp' not provided by package 'lpSolveAPI' Calls: loadNamespace -> library.dynam -> dyn.load Execution halted A namespace must be able to be loaded with just the base namespace loaded: otherwise if the namespace gets loaded by a saved object, the session will be unable to start. Probably some imports need to be declared in the NAMESPACE file. Here is the contents of the DESCRIPTION: Package: safeBinaryRegression Version: 0.1-2 Date: 2009-07-01 Title: Safe Binary Regression Author: Kjell Konis Maintainer: Kjell Konis Depends: R (>= 2.9.1), lpSolveAPI (>= 5.5.0.14) LinkingTo: lpSolveAPI Description: Overloads the stats::glm function so that a test for the existence of the maximum likelihood estimate is computed as part of the fitting procedure for binary regression models. License: GPL-2 Here is the contents of the NAMESPACE: export(glm) useDynLib(safeBinaryRegression, linprog, reducedLP) The strange thing is that if I change the length of the package name (for example safeBinaryRegression to safeBinaryRegress) the warning goes away and R CMD check completes successfully. Also, regardless of the package name, the package installs and runs as expected. I'm at a loss trying to figure out what's going on. To reproduce: (1) check out the sbr package (which passes R CMD check) svn checkout svn://svn.r-forge.r-project.org/svnroot/sbr/pkg safeBinaryRegression (2) replace sbr with safeBinaryRegression in DESCRIPTION and NAMESPACE (3) run R CMD check Session Info: sessionInfo() R version 2.9.1 (2009-06-26) i386-apple-darwin8.11.1 locale: fr_CH.UTF-8/en_US.UTF-8/C/C/fr_CH.UTF-8/fr_CH.UTF-8 attached base packages: [1] stats graphics grDevices utils datasets methods base Thanks, Kjell __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel -- Paul E. Johnson Professor, Political Science 1541 Lilac Lane, Room 504 University of Kansas __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Rcmd check fails on Windows Samba network path in R 2.9.1
See r-devel thread '[Rd] file.access() on network (mounted) drive on
Windows Vista' on Nov 26, 2008:
http://tolstoy.newcastle.edu.au/R/e5/devel/08/11/0806.html
where it was concluded that file.access() is "not 100%", e.g.
file.access() and file.info() can give different answers.
I added fileAccess(..., safe=TRUE) to R.utils, which imitates
file.access() but relies also on file.info() and file.exists() to
infer the rights. It is still not 100% but (hopefully) closer than
file.access().
/H
On Tue, Jul 21, 2009 at 1:55 PM, Kevin R.
Coombes wrote:
> Hi,
>
> The problem almost certainly has something to do with Samba. We also have a
> NetApp file system, and copying the package source to that drive and running
> Rcmd check from Windows works just fine.
>
> When running the commands from an interactive R session, file_test returns
> TRUE and file.access to test write permission indicates a failure, even
> though that information is incorrect.Here is a session transcript:
>
> -
>> getwd()
> [1] "n:/krc/Umpire/R-Package"
>> lib <- "Umpire.Rcheck"
>> file_test("-d", lib)
> [1] TRUE
>> file.access(lib, 2)
> Umpire.Rcheck
> -1
>> dir.create(paste(lib, "testdir", sep='/'))
>> dir(lib)
> [1] "00check.log" "00install.out" "testdir" > sessionInfo()
> R version 2.9.1 (2009-06-26)
> i386-pc-mingw32
>
> locale:
> LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
> States.1252;LC_MONETARY=English_United
> States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
>
> attached base packages:
> [1] stats graphics grDevices utils datasets methods base
>
> Again, this all worked (and still works) correctly in 2.8.1.
>
> Kevin
>
> Tony Plate wrote:
>>
>> This error message looks like it comes from src/library/tools/R/install.R,
>> which contains the following test:
>>
>> if (!.file_test("-d", lib) || file.access(lib, 2L))
>> stop("ERROR: no permission to install to directory ",
>> sQuote(lib), call. = FALSE)
>>
>> The function .file_test() is defined earlier in the same file (it looks at
>> file.info(lib)$isdir) and appears to be intended to be the same as
>> utils:::file_test().
>>
>> You could start trying to figure out what the problem is by trying these
>> calls on the directory in question from an interactive R session.
>> (They seem to work as intended on my Windows system with a NetApp file
>> system mounted on a letter drive.)
>>
>> -- Tony Plate
>>
>> Kevin R. Coombes wrote:
>>>
>>> Hi,
>>>
>>> I have just updated R from version 2.8.1 to version 2.9.1. I am running
>>> Windows XP Professional, Service Pack 3.
>>>
>>> With the update, I decided to update a set of packages that I maintain by
>>> compiling them for the new version. Everything worked fine except for one
>>> package. This package is unique (among the six I was working on) in that is
>>> stored on a UNIX-based file server that is exported to the Windows network
>>> via Samba. The root of that network path is mapped to drive "N:" on the
>>> local machine.
>>>
>>> 'Rcmd check' fails for this package under 2.9.1. The error message in
>>> '00install.out' is:
>>> "Error: ERROR: no permission to install to directory
>>> 'N:/krc/Umpire/R-Package/Umpire.Rcheck'"
>>>
>>> 'Rcmd check' works for this package under 2.8.1.
>>>
>>> 'Rcmd check' works for this package if the directory is copied onto a
>>> local hard drive instead of the network drive.
>>>
>>> 'Rcmd build' and 'Rcmd build --binary' work under both versions.
>>>
>>> It would be nice if someone could figure out what has changed and fix
>>> it
>>>
>>> Best,
>>> Kevin Coombes
>>>
>>> __
>>> 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
>
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Improvement of [dpq]wilcox functions
Ivo Ugrina wrote: > Hi, > > I believe I have significantly improved [dpq]wilcox > functions by implementing Harding's algorithm: > Harding, E.F. (1984): An Efficient, Minimal-storage Procedure > for Calculating the Mann-Whitney U, Generalized U and Similar > Distributions, App. Statist., 33, 1-6 > > Results on my computer show (against R-2.9.1): > >> system.time( dwilcox( 800, 800, 80) ) >user system elapsed > 0.240 0.180 0.443 >> system.time( dwilcox( (1:2800), 80, 80) ) >user system elapsed > 0.290 0.020 0.30 >> system.time( dwilcox( (1:2800), 160, 160) ) >user system elapsed > 0.810 0.060 0.868 >> system.time( dwilcox( (1:28000), 210, 210) ) >user system elapsed > 17.700 0.600 18.313 > RAM: ~ 700MB >> system.time( pwilcox( 21000, 211, 211) ) >user system elapsed > 18.110 0.640 18.762 >> system.time( a <- qwilcox( 0.43, 200, 400) ) > ^C > Timing stopped at: 14.39 1.43 18.794 > RAM: > 1.4GB at interrupt time > >> system.time( dwilcox(800, 800, 80) ) >user system elapsed > 0.140 0.000 0.144 >> system.time( dwilcox( (1:2800), 80, 80) ) >user system elapsed > 0.010 0.000 0.007 >> system.time( dwilcox( (1:2800), 160, 160) ) >user system elapsed > 0.020 0.000 0.016 >> system.time( dwilcox( (1:28000), 210, 210) ) >user system elapsed > 0.050 0.000 0.05 > RAM: < 1MB >> system.time( pwilcox( 21000, 211, 211) ) >user system elapsed > 0.020 0.000 0.025 >> system.time( a <- qwilcox( 0.43, 200, 400) ) >user system elapsed > 0.040 0.000 0.07 > RAM: < 1MB > > There is no more need for > wilcox_free at [dpq]wilcox in src/library/stats/distn.R > (every other call after the first one with the same m,n > will just read the results from the array so it will be > really fast) and for > #define WILCOX_MAX 50 in src/nmath/nmath.h > > p.s. modified files are in the attachment > > have fun, > So is this an improvement or not? -- Ivo Ugrina << http://web.math.hr/~iugrina >> Teaching/Research Assistant at Department of Mathematics University of Zagreb, Croatia __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] matrix algebra in c
> 1. use F77_CALL() to call matrix multiplication functions from blas. > This should be ok as long as I remember that c and FORTRAN store > matrices differently. Concerning 1. you might want to look at some C++ wrapper for BLAS and Lapack like http://flens.sourceforge.net, this makes things easier. -- View this message in context: http://www.nabble.com/matrix-algebra-in-c-tp24484891p24604315.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] Rcmd check fails on Windows Samba network path in R 2.9.1
Hi,
Thanks; that confirms that the problem originates where I thought it
did, and provides a fix from within R.
However, the original problem was that running "Rcmd check" from a
Windows/DOS command prompt on a package on a mounted Samba network drive
fails. The failure occurs because of the file.access call in
"install.R", which appears to get incorrect information from the
operating system. Is it possible to get install.R changed so it works
in this situation?
(Under the philosophy that "it is easier to beg forgiveness than to ask
permission", perhaps install R should just try to create/write what it
wants and only fail iby checking the result rather than 'asking
permission")
Best,
Kevin
Henrik Bengtsson wrote:
See r-devel thread '[Rd] file.access() on network (mounted) drive on
Windows Vista' on Nov 26, 2008:
http://tolstoy.newcastle.edu.au/R/e5/devel/08/11/0806.html
where it was concluded that file.access() is "not 100%", e.g.
file.access() and file.info() can give different answers.
I added fileAccess(..., safe=TRUE) to R.utils, which imitates
file.access() but relies also on file.info() and file.exists() to
infer the rights. It is still not 100% but (hopefully) closer than
file.access().
/H
On Tue, Jul 21, 2009 at 1:55 PM, Kevin R.
Coombes wrote:
Hi,
The problem almost certainly has something to do with Samba. We also have a
NetApp file system, and copying the package source to that drive and running
Rcmd check from Windows works just fine.
When running the commands from an interactive R session, file_test returns
TRUE and file.access to test write permission indicates a failure, even
though that information is incorrect.Here is a session transcript:
-
getwd()
[1] "n:/krc/Umpire/R-Package"
lib <- "Umpire.Rcheck"
file_test("-d", lib)
[1] TRUE
file.access(lib, 2)
Umpire.Rcheck
-1
dir.create(paste(lib, "testdir", sep='/'))
dir(lib)
[1] "00check.log" "00install.out" "testdir" > sessionInfo()
R version 2.9.1 (2009-06-26)
i386-pc-mingw32
locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
States.1252;LC_MONETARY=English_United
States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
Again, this all worked (and still works) correctly in 2.8.1.
Kevin
Tony Plate wrote:
This error message looks like it comes from src/library/tools/R/install.R,
which contains the following test:
if (!.file_test("-d", lib) || file.access(lib, 2L))
stop("ERROR: no permission to install to directory ",
sQuote(lib), call. = FALSE)
The function .file_test() is defined earlier in the same file (it looks at
file.info(lib)$isdir) and appears to be intended to be the same as
utils:::file_test().
You could start trying to figure out what the problem is by trying these
calls on the directory in question from an interactive R session.
(They seem to work as intended on my Windows system with a NetApp file
system mounted on a letter drive.)
-- Tony Plate
Kevin R. Coombes wrote:
Hi,
I have just updated R from version 2.8.1 to version 2.9.1. I am running
Windows XP Professional, Service Pack 3.
With the update, I decided to update a set of packages that I maintain by
compiling them for the new version. Everything worked fine except for one
package. This package is unique (among the six I was working on) in that is
stored on a UNIX-based file server that is exported to the Windows network
via Samba. The root of that network path is mapped to drive "N:" on the
local machine.
'Rcmd check' fails for this package under 2.9.1. The error message in
'00install.out' is:
"Error: ERROR: no permission to install to directory
'N:/krc/Umpire/R-Package/Umpire.Rcheck'"
'Rcmd check' works for this package under 2.8.1.
'Rcmd check' works for this package if the directory is copied onto a
local hard drive instead of the network drive.
'Rcmd build' and 'Rcmd build --binary' work under both versions.
It would be nice if someone could figure out what has changed and fix
it
Best,
Kevin Coombes
__
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
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] Link to documentation in another package
Dear all,
One of the functions that I wrote (ggsave.latex) extents the
functionality of a function (ggsave) in another package (ggplot2).
Instead of copying all the information I would like to create a link in
the helpfile of ggsave.latex to the helpfile of ggsave. I tried
\code{\link{ggsave}} and \code{\link{ggplot2::ggsave}}, but neither
worked. Both cases gave a 'missing link' warning.
Any suggestions?
Thierry
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
thierry.onkel...@inbo.be
www.inbo.be
To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data.
~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer
en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is
door een geldig ondertekend document. The views expressed in this message
and any annex are purely those of the writer and may not be regarded as stating
an official position of INBO, as long as the message is not confirmed by a duly
signed document.
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Link to documentation in another package
Try this: \code{\link[ggplot2]{ggsave}}
Nicholas L. Crookston, Operations Research Analyst
Rocky Mountain Research Station
USDA Forest Service
1221 South Main, Moscow, ID 83843
Office: (208) 883-2317, FAX: (208) 883-2318
EMail: ncrooks...@fs.fed.us
"ONKELINX, Thierry"
Sent by: r-devel-boun...@r-project.org
07/22/2009 06:55 AM
To
cc
Subject
[Rd] Link to documentation in another package
Dear all,
One of the functions that I wrote (ggsave.latex) extents the
functionality of a function (ggsave) in another package (ggplot2).
Instead of copying all the information I would like to create a link in
the helpfile of ggsave.latex to the helpfile of ggsave. I tried
\code{\link{ggsave}} and \code{\link{ggplot2::ggsave}}, but neither
worked. Both cases gave a 'missing link' warning.
Any suggestions?
Thierry
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
thierry.onkel...@inbo.be
www.inbo.be
To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data.
~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver
weer
en binden het INBO onder geen enkel beding, zolang dit bericht niet
bevestigd is
door een geldig ondertekend document. The views expressed in this message
and any annex are purely those of the writer and may not be regarded as
stating
an official position of INBO, as long as the message is not confirmed by a
duly
signed document.
__
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] maximum supported lag is 350
I have found a strange error:
> fit = Arima(flow, c(1,0,1), list(order=c(0,1,1), period=96*7))
Error in makeARIMA(trarma[[1L]], trarma[[2L]], Delta, kappa) :
maximum supported lag is 350
Is in fact quite common to have a lag > 350 using a weekly period in
15min steps = 672 (standard in traffic flow prediction litterature for
example).
Is there any reasons against doing what is suggested in the comment and
avoid the hardcoded limit?
Thanks,
Matteo Bertini
# https://svn.r-project.org/R/trunk/src/library/stats/src/arima.c
SEXP getQ0(SEXP sPhi, SEXP sTheta)
{
[8<]
/* This is the limit using an int index. We could use
size_t and get more on a 64-bit system,
but there seems no practical need. */
if(r > 350) error(_("maximum supported lag is 350"));
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Rcmd check fails on Windows Samba network path in R 2.9.1
On 7/22/2009 9:10 AM, Kevin R. Coombes wrote:
Hi,
Thanks; that confirms that the problem originates where I thought it
did, and provides a fix from within R.
However, the original problem was that running "Rcmd check" from a
Windows/DOS command prompt on a package on a mounted Samba network drive
fails. The failure occurs because of the file.access call in
"install.R", which appears to get incorrect information from the
operating system. Is it possible to get install.R changed so it works
in this situation?
Of course that's possible, but I'd put it at a low priority. The right
solution here is to fix Samba. A simple workaround is to copy the
package to a non-Samba drive. Since this is not our bug, and it has a
simple workaround, I don't want to put an effort into fixing it.
Duncan Murdoch
(Under the philosophy that "it is easier to beg forgiveness than to ask
permission", perhaps install R should just try to create/write what it
wants and only fail iby checking the result rather than 'asking
permission")
Best,
Kevin
Henrik Bengtsson wrote:
See r-devel thread '[Rd] file.access() on network (mounted) drive on
Windows Vista' on Nov 26, 2008:
http://tolstoy.newcastle.edu.au/R/e5/devel/08/11/0806.html
where it was concluded that file.access() is "not 100%", e.g.
file.access() and file.info() can give different answers.
I added fileAccess(..., safe=TRUE) to R.utils, which imitates
file.access() but relies also on file.info() and file.exists() to
infer the rights. It is still not 100% but (hopefully) closer than
file.access().
/H
On Tue, Jul 21, 2009 at 1:55 PM, Kevin R.
Coombes wrote:
Hi,
The problem almost certainly has something to do with Samba. We also have a
NetApp file system, and copying the package source to that drive and running
Rcmd check from Windows works just fine.
When running the commands from an interactive R session, file_test returns
TRUE and file.access to test write permission indicates a failure, even
though that information is incorrect.Here is a session transcript:
-
getwd()
[1] "n:/krc/Umpire/R-Package"
lib <- "Umpire.Rcheck"
file_test("-d", lib)
[1] TRUE
file.access(lib, 2)
Umpire.Rcheck
-1
dir.create(paste(lib, "testdir", sep='/'))
dir(lib)
[1] "00check.log" "00install.out" "testdir" > sessionInfo()
R version 2.9.1 (2009-06-26)
i386-pc-mingw32
locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
States.1252;LC_MONETARY=English_United
States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
Again, this all worked (and still works) correctly in 2.8.1.
Kevin
Tony Plate wrote:
This error message looks like it comes from src/library/tools/R/install.R,
which contains the following test:
if (!.file_test("-d", lib) || file.access(lib, 2L))
stop("ERROR: no permission to install to directory ",
sQuote(lib), call. = FALSE)
The function .file_test() is defined earlier in the same file (it looks at
file.info(lib)$isdir) and appears to be intended to be the same as
utils:::file_test().
You could start trying to figure out what the problem is by trying these
calls on the directory in question from an interactive R session.
(They seem to work as intended on my Windows system with a NetApp file
system mounted on a letter drive.)
-- Tony Plate
Kevin R. Coombes wrote:
Hi,
I have just updated R from version 2.8.1 to version 2.9.1. I am running
Windows XP Professional, Service Pack 3.
With the update, I decided to update a set of packages that I maintain by
compiling them for the new version. Everything worked fine except for one
package. This package is unique (among the six I was working on) in that is
stored on a UNIX-based file server that is exported to the Windows network
via Samba. The root of that network path is mapped to drive "N:" on the
local machine.
'Rcmd check' fails for this package under 2.9.1. The error message in
'00install.out' is:
"Error: ERROR: no permission to install to directory
'N:/krc/Umpire/R-Package/Umpire.Rcheck'"
'Rcmd check' works for this package under 2.8.1.
'Rcmd check' works for this package if the directory is copied onto a
local hard drive instead of the network drive.
'Rcmd build' and 'Rcmd build --binary' work under both versions.
It would be nice if someone could figure out what has changed and fix
it
Best,
Kevin Coombes
__
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
__
R-devel@r-project.org mailing list
https://stat.ethz.
Re: [Rd] Link to documentation in another package
Thanks. That did solve my problem.
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
thierry.onkel...@inbo.be
www.inbo.be
To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data.
~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey
Van: Nicholas L Crookston [mailto:ncrooks...@fs.fed.us]
Verzonden: woensdag 22 juli 2009 16:10
Aan: ONKELINX, Thierry
CC: r-devel@r-project.org; r-devel-boun...@r-project.org
Onderwerp: Re: [Rd] Link to documentation in another package
Try this: \code{\link[ggplot2]{ggsave}}
Nicholas L. Crookston, Operations Research Analyst
Rocky Mountain Research Station
USDA Forest Service
1221 South Main, Moscow, ID 83843
Office: (208) 883-2317, FAX: (208) 883-2318
EMail: ncrooks...@fs.fed.us
"ONKELINX, Thierry"
Sent by: r-devel-boun...@r-project.org
07/22/2009 06:55 AM
To
cc
Subject
[Rd] Link to documentation in another package
Dear all,
One of the functions that I wrote (ggsave.latex) extents the
functionality of a function (ggsave) in another package (ggplot2).
Instead of copying all the information I would like to create a link in
the helpfile of ggsave.latex to the helpfile of ggsave. I tried
\code{\link{ggsave}} and \code{\link{ggplot2::ggsave}}, but neither
worked. Both cases gave a 'missing link' warning.
Any suggestions?
Thierry
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium
tel. + 32 54/436 185
thierry.onkel...@inbo.be
www.inbo.be
To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data.
~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver
weer
en binden het INBO onder geen enkel beding, zolang dit bericht niet
bevestigd is
door een geldig ondertekend document. The views expressed in this
message
and any annex are purely those of the writer and may not be regarded as
stating
an official position of INBO, as long as the message is not confirmed by
a duly
signed document.
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer
en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is
door een geldig ondertekend document. The views expressed in this message
and any annex are purely those of the writer and may not be regarded as stating
an official position of INBO, as long as the message is not confirmed by a duly
signed document.
[[alternative HTML version deleted]]
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Rcmd check fails on Windows Samba network path in R 2.9.1
On Jul 22, 2009, at 10:28 , Duncan Murdoch wrote:
On 7/22/2009 9:10 AM, Kevin R. Coombes wrote:
Hi,
Thanks; that confirms that the problem originates where I thought
it did, and provides a fix from within R.
However, the original problem was that running "Rcmd check" from a
Windows/DOS command prompt on a package on a mounted Samba network
drive fails. The failure occurs because of the file.access call in
"install.R", which appears to get incorrect information from the
operating system. Is it possible to get install.R changed so it
works in this situation?
Of course that's possible, but I'd put it at a low priority. The
right solution here is to fix Samba. A simple workaround is to copy
the package to a non-Samba drive. Since this is not our bug, and it
has a simple workaround, I don't want to put an effort into fixing it.
It *is* our bug - this happens on other drives, too, not just samba
(see my "R CMD check failing due to unreliable file.access() on
Windows" post from 7/2 on R-core about this) - hence your advice is
quite far off. Also note that other tools have no issue finding the
correct flags. I thought this is being worked on (according to that
thread)... I'm not saying that file.access() can be fixed easily
(essentially you have to deal with GetFileSecurityW returning
ERROR_NOT_SUPPORTED and use the GetFileAttributesW result or something
similar instead), but check surely can use the same old reality-based
check that it used before 2.9.0.
Cheers,
Simon
Duncan Murdoch
(Under the philosophy that "it is easier to beg forgiveness than to
ask permission", perhaps install R should just try to create/write
what it wants and only fail iby checking the result rather than
'asking permission")
Best,
Kevin
Henrik Bengtsson wrote:
See r-devel thread '[Rd] file.access() on network (mounted) drive on
Windows Vista' on Nov 26, 2008:
http://tolstoy.newcastle.edu.au/R/e5/devel/08/11/0806.html
where it was concluded that file.access() is "not 100%", e.g.
file.access() and file.info() can give different answers.
I added fileAccess(..., safe=TRUE) to R.utils, which imitates
file.access() but relies also on file.info() and file.exists() to
infer the rights. It is still not 100% but (hopefully) closer than
file.access().
/H
On Tue, Jul 21, 2009 at 1:55 PM, Kevin R.
Coombes wrote:
Hi,
The problem almost certainly has something to do with Samba. We
also have a
NetApp file system, and copying the package source to that drive
and running
Rcmd check from Windows works just fine.
When running the commands from an interactive R session,
file_test returns
TRUE and file.access to test write permission indicates a
failure, even
though that information is incorrect.Here is a session transcript:
-
getwd()
[1] "n:/krc/Umpire/R-Package"
lib <- "Umpire.Rcheck"
file_test("-d", lib)
[1] TRUE
file.access(lib, 2)
Umpire.Rcheck
-1
dir.create(paste(lib, "testdir", sep='/'))
dir(lib)
[1] "00check.log" "00install.out" "testdir" > sessionInfo()
R version 2.9.1 (2009-06-26)
i386-pc-mingw32
locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
States.1252;LC_MONETARY=English_United
States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods
base
Again, this all worked (and still works) correctly in 2.8.1.
Kevin
Tony Plate wrote:
This error message looks like it comes from src/library/tools/R/
install.R,
which contains the following test:
if (!.file_test("-d", lib) || file.access(lib, 2L))
stop("ERROR: no permission to install to directory ",
sQuote(lib), call. = FALSE)
The function .file_test() is defined earlier in the same file
(it looks at
file.info(lib)$isdir) and appears to be intended to be the same as
utils:::file_test().
You could start trying to figure out what the problem is by
trying these
calls on the directory in question from an interactive R session.
(They seem to work as intended on my Windows system with a
NetApp file
system mounted on a letter drive.)
-- Tony Plate
Kevin R. Coombes wrote:
Hi,
I have just updated R from version 2.8.1 to version 2.9.1. I
am running
Windows XP Professional, Service Pack 3.
With the update, I decided to update a set of packages that I
maintain by
compiling them for the new version. Everything worked fine
except for one
package. This package is unique (among the six I was working
on) in that is
stored on a UNIX-based file server that is exported to the
Windows network
via Samba. The root of that network path is mapped to drive
"N:" on the
local machine.
'Rcmd check' fails for this package under 2.9.1. The error
message in
'00install.out' is:
"Error: ERROR: no permission to install to directory
'N:/krc/Umpire/R-Package/Umpire
Re: [Rd] Improvement of [dpq]wilcox functions
Hi Ivo, Excerpts from Ivo Ugrina's message of Thu Jul 09 17:05:27 +0200 2009: > I believe I have significantly improved [dpq]wilcox > functions by implementing Harding's algorithm: > Harding, E.F. (1984): An Efficient, Minimal-storage Procedure > for Calculating the Mann-Whitney U, Generalized U and Similar > Distributions, App. Statist., 33, 1-6 I've looked at your code and it is indeed quite a bit faster. Sadly in some cases it produces inaccurate results. See for example: R 2.9.1: > sum(dwilcox(1:(500*100), 500, 100)) [1] 1 R 2.9.1 + your patch: > sum(dwilcox(1:(500*100), 500, 100)) [1] 1.001377 > sum(dwilcox(1:(500*200), 500, 200)) [1] -132443.2 > sum(dwilcox(1:(1000*200), 1000, 200)) [1] 3.412391e+13 The last two examples run out of memory on my machine in an unpatched R. I think if you can sort out the numerical issuses your patch would be interesting since it is indeed quite a bit faster than the current implementation. Cheers, Olaf Mersmann __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] inclusion of a new bandwidth selector for kernel density estimation in R stats
I understand that Prof. Ripley is largely responsible for the many kernel density estimation functions in R, especially the bw.SJ functions for the popular Sheather and Jones bandwidth selector. We have recently developed a new bandwidth selector that significantly improves SJ selector for difficult densities that deviates severely from normality but retains the superior performance of SJ selector for smoother ones. The paper (to appear in Journal of nonparametric statistics) is available at http://www.geocities.com/jg_liao/papers/kernel.pdf and the R function bw.liao is at http://www.geocities.com/jg_liao/software/bw_liao.txt The algorithm and R function appears to be very stable as it has processed more than 2 datasets without problem. We would love to work with Prof. Ripley or other R core member to make the new method part of R stats. Thanks. Jason Jason Liao, http://www.geocities.com/jg_liao [[alternative HTML version deleted]] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] ks.test - The two-sample two-sided Kolmogorov-Smirnov test with ties (PR#13848)
Full_Name: Thomas Waterhouse
Version: 2.9.1
OS: OS X 10.5.7
Submission from: (NULL) (216.239.45.4)
ks.test uses a biased approximation to the p-value in the case of the two-sample
test with ties in the pooled data. This has been justified in R in the past by
the argument that the KS test assumes a continuous distribution. But the
two-sample test can be extended to arbitrary distributions by a combinatorial
argument (below), which I would like to see implemented in R.
The p-value of the test is simply the probability that a random assignment of
the pooled data to two sets of the sizes of the input data sets has a KS test
statistic at least as great as that of the input data. The function psmirnov2x
in ks.c calculates this probability by enumerating all such assignments as
lattice walks, except that it doesn't know how to handle tied data points.
The correct procedure can be deduced by considering that steps in such lattice
walks represent steps along the x-axis in computing the empirical CDFs of the
two data sets. The trick is to consider all occurrences of a repeated data
point together, when determining whether a path exceeds the input statistic or
not. A FORTRAN implementation of this algorithm was published by Nikiforov
(http://www.ams.sunysb.edu/~nkfr/5PUBL.HTM), but it's not for the faint of
heart.
It would be straightforward to change psmirnov2x to include this logic, but I
found it easier simply to rewrite it (with comments this time!) as follows:
void
psmirnov2x(double *x, Sint *m, Sint *n, double *pool)
{
/* Inputs: *x is the value of the KS test statistic (from 0.0 to 1.0)
*m and *n are the lengths of each data set
*pool is an array of length *m + *n formed by combining the two
data sets and sorting in increasing order
Output: *x is unity minus the p-value of the test
Notes: The p-value of the two-sample KS test is the proportion of all
assignments of the pooled data to sets of sizes *m and *n which
have test statistic exceeding the test statistic of the actual
data.
To calculate this, we interpret each assignment as a lattice walk
from the origin to (*m, *n), where the i-th step represents the
i-th
data point in the pooled data set. A step to the right represents
assigning a data point to the *m set, and a step up represents
assigning a data point to the *n set. The difference between the
empirical CDFs of each set can be computed at each step, and the
KS test statistic is simply the maximum of this value along a given
walk.
The p-value is obtained by counting the number of walks that do
not exceed the input value (since this is easier than counting the
number that do) and dividing by the total number of walks; this is
what we return to the caller.
If the same value occurs multiple times in the pooled data, we
need to count all multiplicites at the same time when computing the
empirical CDFs. In terms of the lattice walk, this means we look
only the final occurence of each value when determining whether or
not a walk exceeds the input statistic.
*/
double md, nd, q, *u, w;
Sint i, j;
if (*x == 0.0) {
return 1.0;
}
md = (double) (*m);
nd = (double) (*n);
q = floor(*x * md * nd - 1e-7) / (md * nd); /* just *x minus an epsilon */
u = (double *) R_alloc(*n + 1, sizeof(double));
for (i = 0; i <= *m; i++) {
for (j = 0; j <= *n; j++) {
/* At each step, rescale by w := j/(j+*m); this is equivalent to
dividing by C(*m+*n, *m) at the end, but without the risk of
overflow */
w = (double)(j) / (double)(j + *m);
if (i + j == 0) {
/* Start with 1 walk from the origin to the origin */
u[j] = 1.0;
} else if (i + j < *m + *n &&
pool[i + j - 1] != pool[i + j] &&
fabs(i / md - j / nd) > q) {
/* This walk meets or exceeds *x, so don't count it */
u[j] = 0.0;
} else if (i == 0) {
/* The *n-edge: The number of walks to (0, j) is the same as
the number of walks to (0, j-1) */
u[j] = u[j - 1] * w;
} else if (j == 0) {
/* The *m-edge: The number of walks to (i, 0) is the same as
the number of walks to (i-1, 0) */
/* no-op */
} else {
/* The general case: The number of walks to (i, j) is the
number of walks to (i-1, j) plus the number of walks to
(i, j-1) */
u[j] = u[j] + u[j - 1] * w;
}
}
}
/* Return the proportion of walks that do not exceed the
