Re: [Rd] setClassUnion with numeric; extending class union
Hi John,
sorry for not posting more info. Strangely I get warnings about
setClassUnion with numeric in a very special case: if I define it in a
clean R session then there are no warnings, however if I load a number
of my packages where there are other classes derived from numeric and
exported then I get the following warnings:
> setClassUnion("numericOrNULL", c("numeric","NULL"))
[1] "numericOrNULL"
Warning messages:
1: In .checkSubclasses(class1, classDef, class2, classDef2, where1, :
Subclass "TimeDateBase" of class "numeric" is not local and cannot be
updated for new inheritance information; consider setClassUnion()
2: In .checkSubclasses(class1, classDef, class2, classDef2, where1, :
Subclass "TimeDate" of class "numeric" is not local and cannot be
updated for new inheritance information; consider setClassUnion()
3: In .checkSubclasses(class1, classDef, class2, classDef2, where1, :
Subclass "Time" of class "numeric" is not local and cannot be updated
for new inheritance information; consider setClassUnion()
The class is operational even with those warnings though. Now, the above
classes are defined as follows:
## - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - -
setClass("TimeDateBase",
representation("numeric", mode="character"),
prototype(mode="posix")
)
## - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - -
setClass("TimeDate",
representation("TimeDateBase", tzone="character"),
prototype(tzone="Europe/London")
)
## - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - -
setClass("Time",
representation("TimeDateBase")
)
Theses classes work perfectly fine on their own and are used throughout
our code for all possible time and date operations extending the
existing functionality of R and available third party packages by an
order of magnitude. I do not see a relation between the above class
definitions and the newly defined class union though apart from the fact
that they are in a package namespace and therefore locked. Sorry I
cannot provide more source code as the code is not yet public.
It would definitely be nice to somehow have a .Data slot in NULL or even
a data.frame, although I do understand that this is quite a substantial
piece of work to make it all robust and backward compatible.
> sessionInfo() ## of a clean session
R version 2.9.0 Under development (unstable) (2009-02-02 r47821)
x86_64-unknown-linux-gnu
locale:
C
attached base packages:
[1] stats graphics utils datasets grDevices methods base
Any thoughts are greatly appreciated.
Kind regards,
Oleg
Dr Oleg Sklyar
Research Technologist
AHL / Man Investments Ltd
+44 (0)20 7144 3107
oskl...@maninvestments.com
> -Original Message-
> From: John Chambers [mailto:j...@r-project.org]
> Sent: 11 February 2009 20:40
> To: Sklyar, Oleg (London)
> Cc: r-devel@r-project.org
> Subject: Re: [Rd] setClassUnion with numeric; extending class union
>
> So, I was intrigued and played around a bit more. Still
> can't get any
> warnings, but the following may be the issue.
>
> One thing NOT currently possible is to have a class that has
> NULL as its
> data part, because type NULL is abnormal and can't have attributes.
>
> So if you want a class that contains a union including NULL,
> you're in
> trouble generating a value from the class that is NULL. It's
> not really
> a consequence of the setUnion() per se.
>
> > setClass("bar", contains = "numericOrNULL")
> [1] "bar"
> > zz = new("bar", NULL)
> Error in validObject(.Object) :
> invalid class "bar" object: invalid object for slot ".Data"
> in class
> "bar": got class "list", should be or extend class "numericOrNULL"
>
> (How one got from the error to the message is a question, but in any
> case this can't currently work.)
>
> As in my example and in your example with a slot called "data", no
> problem in having a slot value that is NULL.
>
> Looking ahead, I'm working on some extensions that would
> allow classes
> to contain "abnormal" data types (externalptr, environment, ...) by
> using a reserved slot name, since one can not make the actual
> data type
> one of those types.
>
> John Chambers wrote:
> > What warnings? Which part of the following is not what
> you're looking
> > for? (The usual information is needed, like version of R,
> reproducible
> > example, etc.)
> >
> >
> > > setClassUnion("numericOrNULL", c("numeric","NULL"))
> > [1] "numericOrNULL"
> > > setClass("foo", representation(x="numericOrNULL"))
> > [1] "foo"
> > > ff = new("foo", x= 1:10)
> > > fg = new("foo", x = NULL)
> > >
> > > ff
> > An object of class "foo"
> > Slot "x":
> > [1] 1 2 3 4 5 6 7 8 9 10
> >
> > > fg
> > An object of class "foo"
> > Slot "x":
> > NULL
> > > fk = new("foo")
> > > fk
> > An object of class "foo"
> > Slot "x":
> > NULL
> >
> > John
> >
> > Sklyar, Oleg (London) wrote:
> >> Dear list:
> >>
> >> I am looking for a good way t
[Rd] Patch for src/main/character.c, systematizing recent fix to do_grep
The attached patch provides a modification to the recent fix/improvement to do_grep already included in the most recent development version. The original fix added new functionality to the grep function by adding a new parameter, 'invert'. In the source code for the underlying do_grep, the value of the parameter is used to invert the logical match-no match flag vector ind. The modification is distributed across several lines of code. The patch systematizes the solution by inverting the logical match flag vector in place, once for each element in the character vector passed to grep as the argument 'x'. In the patched version, the invertion appears just once in the code. The patch does not modify the functionality of grep in any way. If the respective documentation was updated to cover the new functionality introduced by the original modification, it still applies to the patched version. The patch does not solve any immediate problem. However, due to replacing the redundantly distributed original modification with a one-line modofication, the patch is intended to make it easier to understand, maintain, and further modify the source code. The patch also renames the variable 'invert' introduced in the original modification to 'invert_opt', for consistency with how (almost) all other logical flag parameters in do_grep are named. This modification is again functionally transparent and requires no modifications to the documentation. The patch was prepared as follows: svn co https://svn.R-project.org/R/trunk/ cd trunk tools/rsync-recommended # modifications made to src/main/character.c svn diff > do_grep.diff The patched sources were successfully compiled and tested as follows: svn revert -R . patch -p0 < do_grep.diff ./configure make make check Assuming that appropriate tests were prepared for the extended version of grep as of the original modification, the patched version was successfully tested. The patched grep was also tested as follows: bin/R --no-save -q
[Rd] Why is srcref of length 6 and not 4 ?
Hello,
Consider this file (/tmp/test.R) :
f <- function( x, y = 2 ){
z <- x + y
print( z )
}
I get this in R 2.7.2 :
> p <- parse( "/tmp/test.R" )
> str( attr( p, "srcref" ) )
List of 1
$ :Class 'srcref' atomic [1:4] 1 1 4 1
.. ..- attr(*, "srcfile")=Class 'srcfile' length 4
and this in R-devel :
> p <- parse( "/tmp/test.R" )
> str( attr(p, "srcref") )
List of 1
$ :Class 'srcref' atomic [1:6] 1 1 4 1 1 1
.. ..- attr(*, "srcfile")=Class 'srcfile'
What are the two last numbers ?
Romain
--
Romain Francois
Independent R Consultant
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
--
Romain Francois
Independent R Consultant
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] Spearman's rank correlation test
Hi All: help(cor.test) claims For Spearman's test, p-values are computed using algorithm AS 89. Algorithm AS 89 was introduced by the paper D. J. Best & D. E. Roberts (1975), Algorithm AS 89: The Upper Tail Probabilities of Spearman's rho. Applied Statistics, Vol. 24, No. 3, 377-379. Table 1(a) in this paper presents maximum absolute error |\Delta_m|, of the approximation for all possible values of the statistic S for samples sizes n = 7, 9, 11, 13. The presented errors are n |\Delta_m| 7 0.0046 9 0.0011 11 0.0006 13 0.0005 Due to the problem explained in detail including a patch at https://stat.ethz.ch/pipermail/r-devel/2009-January/051936.html the error of R implementation of Spearman's rank correlation test is larger than the above bounds for the sample size n = 11 and some of the values of S, which correspond to positive correlation. For example, for n = 11 and S = 90, we have x <- 1:11 y <- c(6:1, 7, 11:8) out <- cor.test(x, y, method="spearman", alternative="greater") out$statistic # 90 out$p.value # 0.02921104 while the correct p-value is 0.03044548, so the absolute difference is 0.00123444. This is larger than the absolute error 0.0006 guaranteed for AS 89. In my opinion, this means that the claim from help(cor.test) cited above is not correct. To see the error of AS 89 in the example above, one can use cor.test(x, -y, method="spearman", alternative="less")$p.value # 0.03036413 since on the side of negative correlation, R calls AS 89 correctly. So, for the x, y above, correctly called AS 89 has absolute error 0.8135. There is a package pspearman currently included to CRAN, which provides a correction of the problem without the need to modify R base. Petr. __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Why is srcref of length 6 and not 4 ?
On 12/02/2009 7:01 AM, Romain Francois wrote:
Hello,
Consider this file (/tmp/test.R) :
f <- function( x, y = 2 ){
z <- x + y
print( z )
}
I get this in R 2.7.2 :
> p <- parse( "/tmp/test.R" )
> str( attr( p, "srcref" ) )
List of 1
$ :Class 'srcref' atomic [1:4] 1 1 4 1
.. ..- attr(*, "srcfile")=Class 'srcfile' length 4
and this in R-devel :
> p <- parse( "/tmp/test.R" )
> str( attr(p, "srcref") )
List of 1
$ :Class 'srcref' atomic [1:6] 1 1 4 1 1 1
.. ..- attr(*, "srcfile")=Class 'srcfile'
What are the two last numbers ?
The original design for srcref gave 4 entries: start line, start byte,
stop line, stop byte. However, in multibyte strings, bytes don't
correspond to columns, so error messages could often report the wrong
location according to what a user sees in an editor. To support the
more useful error messages in R-devel, I added two more values: start
column and stop column. With pure ASCII text these will be the same as
start byte and stop byte; with UTF-8 text and non-ASCII characters they
will be be different. Other multibyte encodings are only supported if
the platform can convert them to UTF-8 (and are not well tested; error
reports would be welcome, if there's a way to improve the performance.)
If you are using these for error reports, I recommend using the two new
values. If you are trying to retrieve the text from the source file,
use the originals.
Duncan Murdoch
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Why is srcref of length 6 and not 4 ?
Duncan Murdoch wrote:
On 12/02/2009 7:01 AM, Romain Francois wrote:
Hello,
Consider this file (/tmp/test.R) :
f <- function( x, y = 2 ){
z <- x + y
print( z )
}
I get this in R 2.7.2 :
> p <- parse( "/tmp/test.R" )
> str( attr( p, "srcref" ) )
List of 1
$ :Class 'srcref' atomic [1:4] 1 1 4 1
.. ..- attr(*, "srcfile")=Class 'srcfile' length 4
and this in R-devel :
> p <- parse( "/tmp/test.R" )
> str( attr(p, "srcref") )
List of 1
$ :Class 'srcref' atomic [1:6] 1 1 4 1 1 1
.. ..- attr(*, "srcfile")=Class 'srcfile'
What are the two last numbers ?
The original design for srcref gave 4 entries: start line, start byte,
stop line, stop byte. However, in multibyte strings, bytes don't
correspond to columns, so error messages could often report the wrong
location according to what a user sees in an editor. To support the
more useful error messages in R-devel, I added two more values: start
column and stop column. With pure ASCII text these will be the same
as start byte and stop byte; with UTF-8 text and non-ASCII characters
they will be be different. Other multibyte encodings are only
supported if the platform can convert them to UTF-8 (and are not well
tested; error reports would be welcome, if there's a way to improve
the performance.)
If you are using these for error reports, I recommend using the two
new values. If you are trying to retrieve the text from the source
file, use the originals.
Duncan Murdoch
Thank you Duncan,
I am using this to massage the output of "parse" into a data frame to
represent it as a tree
(see http://addictedtor.free.fr/misc/sidekick.png)
> cat( readLines( "/tmp/test.R" ), sep = "\n" )
f <- function( x, y = 2 ){
z <- x + y
g <- function( x ){
print( x )
xx <- x + 1
}
g( x )
}
>
> sidekick( "/tmp/test.R", encoding = "utf-8" )
id parent mode srcref1 srcref2 srcref3 srcref4
description
1 1 0 function 1 1 8 1 f <- function(x, y
= 2) {
2 2 1 name 1 26 1
26 {
3 3 1 call 2 2 2 11z <-
x + y
4 4 1 function 3 2 6 2g <-
function(x) {
5 5 1 call 7 2 7
7 g(x)
6 6 4 name 3 20 3
20 {
7 7 4 call 4 4 4 13
print(x)
8 8 4 call 5 4 5 14 xx <-
x + 1
--
Romain Francois
Independent R Consultant
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Why is srcref of length 6 and not 4 ?
> I am using this to massage the output of "parse" into a data frame to > represent it as a tree > (see http://addictedtor.free.fr/misc/sidekick.png) You might also want to take a look at http://github.com/hadley/eval.with.details/blob/master/R/parse.r where I'm trying to do something similar for a different purpose. Hadley -- http://had.co.nz/ __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] proposed simulate.glm method
I have found the "simulate" method (incorporated
in some packages) very handy. As far as I can tell the
only class for which simulate is actually implemented
in base R is lm ... this is actually a little dangerous
for a naive user who might be tempted to try
simulate(X) where X is a glm fit instead, because
it defaults to simulate.lm (since glm inherits from
the lm class), and the answers make no sense ...
Here is my simulate.glm(), which is modeled on
simulate.lm . It implements simulation for poisson
and binomial (binary or non-binary) models, should
be easy to implement others if that seems necessary.
I hereby request comments and suggest that it wouldn't
hurt to incorporate it into base R ... (I will write
docs for it if necessary, perhaps by modifying ?simulate --
there is no specific documentation for simulate.lm)
cheers
Ben Bolker
simulate.glm <- function (object, nsim = 1, seed = NULL, ...)
{
## RNG stuff copied from simulate.lm
if (!exists(".Random.seed", envir = .GlobalEnv, inherits = FALSE))
runif(1)
if (is.null(seed))
RNGstate <- get(".Random.seed", envir = .GlobalEnv)
else {
R.seed <- get(".Random.seed", envir = .GlobalEnv)
set.seed(seed)
RNGstate <- structure(seed, kind = as.list(RNGkind()))
on.exit(assign(".Random.seed", R.seed, envir = .GlobalEnv))
}
## get probabilities/intensities
pred <- matrix(rep(predict(object,type="response"),nsim),ncol=nsim)
ntot <- length(pred)
if (object$family$family=="binomial") {
resp <- object$model[[1]]
size <- if (is.matrix(resp)) rowSums(resp) else 1
}
val <- switch(object$family$family,
poisson=rpois(ntot,pred),
binomial=rbinom(ntot,prob=pred,size=size),
stop("family ",object$family$family," not implemented"))
ans <- as.data.frame(matrix(val,ncol=nsim))
attr(ans, "seed") <- RNGstate
ans
}
if (FALSE) {
## examples: modified from ?simulate
x <- 1:10
n <- 10
y <- rbinom(length(x),prob=plogis((x-5)/2),size=n)
y2 <- c("a","b")[1+rbinom(length(x),prob=plogis((x-5)/2),size=1)]
mod1 <- glm(cbind(y,n-y) ~ x,family=binomial)
mod2 <- glm(factor(y2) ~ x,family=binomial)
S1 <- simulate(mod1, nsim = 4)
S1B <- simulate(mod2, nsim = 4)
## repeat the simulation:
.Random.seed <- attr(S1, "seed")
identical(S1, simulate(mod1, nsim = 4))
S2 <- simulate(mod1, nsim = 200, seed = 101)
rowMeans(S2)/10 # after correcting for binomial sample size, should be
about
fitted(mod1)
plot(rowMeans(S2)/10)
lines(fitted(mod1))
## repeat identically:
(sseed <- attr(S2, "seed")) # seed; RNGkind as attribute
stopifnot(identical(S2, simulate(mod1, nsim = 200, seed = sseed)))
}
--
Ben Bolker
Associate professor, Biology Dep't, Univ. of Florida
bol...@ufl.edu / www.zoology.ufl.edu/bolker
GPG key: www.zoology.ufl.edu/bolker/benbolker-publickey.asc
signature.asc
Description: OpenPGP digital signature
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] proposed simulate.glm method
There is functionality similar to this included in the Zelig package with it's "sim" method. The "sim" method goes a step further and replicates the fitted model's analysis on the generated datasets as well. I would suggest taking a look -- Zelig supports most (if not all) glm models and a wide range of others. The Zelig maintainers' site can be found at: http://gking.harvard.edu/zelig/. Full disclosure: I am an employee of the Institute for Quantitative Social Science, which performs most of the development and support for the Zelig package. Best, Alex D'Amour Statistical Programmer Harvard Institute for Quantitative Social Science 2009/2/12 Ben Bolker : > > I have found the "simulate" method (incorporated > in some packages) very handy. As far as I can tell the > only class for which simulate is actually implemented > in base R is lm ... this is actually a little dangerous > for a naive user who might be tempted to try > simulate(X) where X is a glm fit instead, because > it defaults to simulate.lm (since glm inherits from > the lm class), and the answers make no sense ... > > Here is my simulate.glm(), which is modeled on > simulate.lm . It implements simulation for poisson > and binomial (binary or non-binary) models, should > be easy to implement others if that seems necessary. > > I hereby request comments and suggest that it wouldn't > hurt to incorporate it into base R ... (I will write > docs for it if necessary, perhaps by modifying ?simulate -- > there is no specific documentation for simulate.lm) > > cheers >Ben Bolker > -- > Ben Bolker > Associate professor, Biology Dep't, Univ. of Florida > bol...@ufl.edu / www.zoology.ufl.edu/bolker > GPG key: www.zoology.ufl.edu/bolker/benbolker-publickey.asc > > > __ > 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] proposed simulate.glm method
Elsewhere (at least in lme4), refit(sim(model)) does the same thing [and so one would need something like apply(sim(model,1000),2,refit)]. sim() is quite interesting, as is Zelig, but I'm not sure I am ready to leap to it yet -- this was basically a suggestion that simulate.glm could be included in "vanilla" R ... Also (for better or worse), it looks like sim() also does parametric bootstrapping on the parameter values, whereas simulate.[g]lm() just uses "plug-in" estimates. cheers Ben Bolker Alex D'Amour wrote: > There is functionality similar to this included in the Zelig package > with it's "sim" method. The "sim" method goes a step further and > replicates the fitted model's analysis on the generated datasets as > well. I would suggest taking a look -- Zelig supports most (if not > all) glm models and a wide range of others. > > The Zelig maintainers' site can be found at: http://gking.harvard.edu/zelig/. > > Full disclosure: I am an employee of the Institute for Quantitative > Social Science, which performs most of the development and support for > the Zelig package. > > Best, > Alex D'Amour > Statistical Programmer > Harvard Institute for Quantitative Social Science > > > 2009/2/12 Ben Bolker : >> I have found the "simulate" method (incorporated >> in some packages) very handy. As far as I can tell the >> only class for which simulate is actually implemented >> in base R is lm ... this is actually a little dangerous >> for a naive user who might be tempted to try >> simulate(X) where X is a glm fit instead, because >> it defaults to simulate.lm (since glm inherits from >> the lm class), and the answers make no sense ... >> >> Here is my simulate.glm(), which is modeled on >> simulate.lm . It implements simulation for poisson >> and binomial (binary or non-binary) models, should >> be easy to implement others if that seems necessary. >> >> I hereby request comments and suggest that it wouldn't >> hurt to incorporate it into base R ... (I will write >> docs for it if necessary, perhaps by modifying ?simulate -- >> there is no specific documentation for simulate.lm) >> >> cheers >>Ben Bolker >> -- >> Ben Bolker >> Associate professor, Biology Dep't, Univ. of Florida >> bol...@ufl.edu / www.zoology.ufl.edu/bolker >> GPG key: www.zoology.ufl.edu/bolker/benbolker-publickey.asc >> >> >> __ >> R-devel@r-project.org mailing list >> https://stat.ethz.ch/mailman/listinfo/r-devel >> >> -- Ben Bolker Associate professor, Biology Dep't, Univ. of Florida bol...@ufl.edu / www.zoology.ufl.edu/bolker GPG key: www.zoology.ufl.edu/bolker/benbolker-publickey.asc __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] typo in example(dbEscapeStrings) (PR#13521)
Full_Name: Mayeul Kauffmann
Version: 2.8.1
OS: x86_64-pc-linux-gnu (kubuntu)
Submission from: (NULL) (86.200.212.40)
The file /library/RMySQL/html/dbEscapeStrings.html documents dbEscapeStrings()
In the example, an 's' is missing in line 3:
## Not run:
tmp <- sprintf("select * from emp where lname = %s", "O'Reilly")
sql <- dbEscapeString(con, tmp)
dbGetQuery(con, sql)
## End(Not run)
sql <- dbEscapeString(con, tmp)
should be:
sql <- dbEscapeStrings(con, tmp)
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
