[Rd] package post install instructions
Hello, I've looked in tools:::.install_packages for some sort of hook that would let packages developers point to further instructions after a package is installed. For example, some packages need to setup environment variables, ... Is there something I have missed ? Romain -- Romain Francois Independent R Consultant +33(0) 6 28 91 30 30 http://romainfrancois.blog.free.fr |- http://tr.im/qzSl : using ImageJ from R: the RImageJ package |- http://tr.im/qzSJ : with semantics for java objects in rJava `- http://tr.im/qzTs : Better completion popups __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Conditional dependency between packages
Hi Seth,
And thanks for your suggestion! I was not able to do exactly what you
described (I have no earlier experience with using environments), but
you mentioning .onLoad was a good starting point. I have now removed all
references to pkg1 from the NAMESPACE, and wrote the following .onLoad
function:
.onLoad - function(libname, pkgname) {
if (pkg1 %in% rownames(utils:::installed.packages()) ) {
library(pkg1)
info = matrix(c(fun1, fun2, fun3, rep(pkg2, 3), rep(NA,3)),
ncol = 3)
registerS3methods(info, package = pkg1, env = environment(funInPkg2))
}
}
New methods for functions fun1, fun2 and fun3 seem to be available if
pkg1 is installed, while they are ignored if pkg1 is not installed. The
function above loads pkg1 automatically if installed (I would prefer
this to be optional), but at least it will not be necessary to download
pkg1 (with all its dependencies) for users without interest in it.
I have not found any description of registerS3methods (except from an
old .Rd file stating that it is not intended to be called directly), so
there might be better ways of doing this... And I am sure there is a
better way of assigning the right environment...
Thanks again,
Jon
Seth Falcon wrote:
Hi Jon,
* On 2009-06-30 at 15:27 +0200 Jon Olav Skoien wrote:
I work on two packages, pkg1 and pkg2 (in two different projects). pkg1 is
quite generic, pkg2 tries to solve a particular problem within same field
(geostatistics). Therefore, there might be users who want to use pkg2 as an
add-on package to increase the functionality of pkg1. In other words,
functions in pkg1 are based on the S3 class system, and I want pkg2 to
offer methods for pkg2-objects to functions defined in pkg1, for users
having both packages installed. Merging the packages or making pkg2 always
depend pkg1 would be the easiest solution, but it is not preferred as most
users will only be interested in one of the packages.
I'm not sure I understand the above, I think you may have a pkg2 where
you meant pkg1, but I'm not sure it matters.
I think the short version is, pkg2 can be used on its own but will do
more if pkg1 is available. I don't think R's packaging system
currently supports conditional dependencies as you might like.
However, I think you can get the behavior you want by following a
recipe like:
* In pkg2 DESCRIPTION, list Suggests: pkg1.
* In pkg2 code, you might define a package-level environment and
in .onLoad check to see if pkg1 is available.
PKG_INFO - new.env(parent=emptyenv())
.onLoad - function(libname, pkgname) {
if (check if pkg1 is available) {
PKG_INFO[[pkg1]] - TRUE
}
}
* Then your methods can check PKG_INFO[[pkg1]].
+ seth
__
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] V2.9.0 changes [Sec=Unclassified]
Troy Robertson wrote:
Well...
My performance problems were in the pass-by-value semantics of R.
I have just changed my classes to inherit from .environment and then moved
data members from S4 slots to the .xData objects as Martin suggested.
Actually, I had hoped the take-home message would be in the final paragraph:
Of course I haven't seen your code, but a different interpretation of
your performance issues is that, within the rules of S4, you've chosen
to implement functionality in an inefficient way. So it might be
instructive to step back a bit and try to reformulate your data
structures and methods. This is hard to do.
Martin
That meant I could remove all my returns and assignments on all method calls.
This has sped execution time for my model up by more than an order of
magnitude. Eg one test simulation from 1931 secs down to 175 secs.
Not bad seeing as though the class structure, functionality and logic has not
been touched.
I really do think S4 could benfit from having its slots stored in environment
when the class enherits from .environment. It would be a lot more sensible
if my data members were still declared as S4 slots instead of having to hide
them in .xData
Troy
Troy Robertson
Database and Computing Support Provider
Southern Ocean Ecosystems, ERM/Fish
Australian Antarctic Division
Channel Highway, Kingston 7050
PH: 03 62323571
troy.robert...@aad.gov.au
-Original Message-
From: Martin Morgan [mailto:mtmor...@fhcrc.org]
Sent: Tuesday, 23 June 2009 11:25 PM
To: Troy Robertson
Cc: 'r-devel@R-project.org'
Subject: Re: [Rd] V2.9.0 changes [Sec=Unclassified]
Troy Robertson wrote:
Hi all,
Prefix: I am a frustrated Java coder in R.
ah good, if you're frustrated with Java you'll find R very different ;)
I am coding a medium sized ecosystem modelling program in R. I have
changed to using S4 objects and it has cost me an order of magnitude in
execution speed over the functional model. I cannot afford this penalty
and have found that it is the result of all the passing-by-value of
objects.
I see that you can now safely inherit from environment in V2.9.0.
That got me all excited that I would now be able to pass objects by
reference.
But...
That doesn't seem to be the case.
It only seem that passing an environment which holds the object allows
for pass-by-reference and that passing an object which inherits from
environment doesn't.
Why is this the case, either an object inherits the properties of its
parent or it doesn't.
The object inherits slots from it's parent, and the methods defined on
the parent class. Maybe this example helps?
setClass(Ticker, contains=.environment)
## initialize essential, so each instance gets its own environment
setMethod(initialize, Ticker,
function(.Object, ..., .xData=new.env(parent=emptyenv()))
{
.xData[[count]] - 0
callNextMethod(.Object, ..., .xData=.xData)
})
## tick: increment (private) counter by n
setGeneric(tick, function(reference, n=1L) standardGeneric(tick),
signature=reference)
setMethod(tick, Ticker, function(reference, n=1L) {
reference[[count]] - reference[[count]] + n
})
## tock: report current value of counter
setGeneric(tock, function(reference) standardGeneric(tock))
setMethod(tock, Ticker, function(reference) {
reference[[count]]
})
and then
e - new(Ticker)
tock(e)
[1] 0
tick(e); tick(e, 10); tock(e)
[1] 11
f - e
tock(f); tick(e); tock(f)
[1] 11
[1] 12
The data inside .environment could be structured, too, using S4.
Probably it would be more appropriate to have the environment as a slot,
rather the class that is being extended. And in terms of inherited
'properties', e.g., the [[ function as defined on environments is
available
e[[count]]
Of course I haven't seen your code, but a different interpretation of
your performance issues is that, within the rules of S4, you've chosen
to implement functionality in an inefficient way. So it might be
instructive to step back a bit and try to reformulate your data
structures and methods. This is hard to do.
Martin
Has anyone else had a play with this? Or have I got it all wrong.
I tried the below:
-
setClass('foo', representation=representation(stuff='list',
bar='numeric'),
prototype=list(stuff=list(), bar=0),
contains='.environment')
setGeneric('doit', function(.Object, newfoo='environment')
standardGeneric('doit'))
setMethod('doit', 'foo', function(.Object, newfoo){new...@bar - 10})
z - new('foo')
z...@stuff$x - new('foo')
doit(z,z...@stuff$x)
z...@stuff$x@bar
[1] 0
--
Can anyone help with a better way of doing this.
I'm trying to avoid all the
Re: [Rd] V2.9.0 changes [Sec=Unclassified]
On Thu, Jul 2, 2009 at 1:37 AM, Troy Robertsontroy.robert...@aad.gov.au wrote: Well... My performance problems were in the pass-by-value semantics of R. I have just changed my classes to inherit from .environment and then moved data members from S4 slots to the .xData objects as Martin suggested. Note that the R.oo and proto packages already use environments for their storage. e.g. library(proto) p - proto(a = 1, incr = function(.) .$a - .$a + 1) class(p) # c(proto, environment) p$a # 1 p$incr() p$a # 2 p$ls() # c(a, incr) ls(p) # same __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
[Rd] OOP performance, was: V2.9.0 changes
Hi Troy, first of all a question, what kind of ecosystem models are you developing in R? Differential equations or individual-based? Your write that you are a frustrated Java developer in R. I have a similar experience, however I still like JAVA, and I'm now more happy with R as it is much more efficient (i.e. sum(programming + runtime)) for the things I usually do: ecological data analysis and modelling. After using functional R quite a time and Java in parallel I had the same idea, to make R more JAVA like and to model ecosystems in an object oriented manner. At that time I took a look into R.oo (thanks Henrik Bengtssson) and was one of the Co-authors of proto. I still think that R.oo is very good and that proto is a cool idea, but finally I switched to the recommended S4 for my ecological simulation package. Note also, that my solution was *not* to model the ecosystems as objects (habitat - populations- individuals), but instead to model ecological models (equations, inputs, parameters, time steps, outputs, ...). This works quite well with S4. A speed test (see useR!2006 poster on http://simecol.r-forge.r-project.org/) showed that all OOP flavours had quite comparable performance. The only thing I have to have in mind are a few rules: - avoid unnecessary copying of large objects. Sometimes it helps to prefer matrices over data frames. - use vectorization. This means for an individual-based model that one has to re-think how to model an individual: not many [S4] objects like in JAVA, but R structures (arrays, lists, data frames) where vectorized functions (e.g. arithmetics or subset) can work with. - avoid interpolation (i.e. approx) and if unavoidable, minimize the tables. If all these things do not help, I write core functions in C (others use Fortran). This can be done in a mixed style and even a full C to C communication is possible (see the deSolve documentation how to do this with differential equation models). Thomas P. -- Thomas Petzoldt Technische Universitaet Dresden Institut fuer Hydrobiologiethomas.petzo...@tu-dresden.de 01062 Dresden http://tu-dresden.de/hydrobiologie/ GERMANY __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Conditional dependency between packages
if (pkg1 %in% rownames(utils:::installed.packages()) ) {
library(pkg1)
...
}
can be replaced by:
if (require(pkg1)) {
...
}
/Henrik
On Thu, Jul 2, 2009 at 5:29 AM, Jon Olav Skoienj.sko...@geo.uu.nl wrote:
Hi Seth,
And thanks for your suggestion! I was not able to do exactly what you
described (I have no earlier experience with using environments), but you
mentioning .onLoad was a good starting point. I have now removed all
references to pkg1 from the NAMESPACE, and wrote the following .onLoad
function:
.onLoad - function(libname, pkgname) {
if (pkg1 %in% rownames(utils:::installed.packages()) ) {
library(pkg1)
info = matrix(c(fun1, fun2, fun3, rep(pkg2, 3), rep(NA,3)), ncol =
3)
registerS3methods(info, package = pkg1, env = environment(funInPkg2))
}
}
New methods for functions fun1, fun2 and fun3 seem to be available if pkg1
is installed, while they are ignored if pkg1 is not installed. The function
above loads pkg1 automatically if installed (I would prefer this to be
optional), but at least it will not be necessary to download pkg1 (with all
its dependencies) for users without interest in it.
I have not found any description of registerS3methods (except from an old
.Rd file stating that it is not intended to be called directly), so there
might be better ways of doing this... And I am sure there is a better way of
assigning the right environment...
Thanks again,
Jon
Seth Falcon wrote:
Hi Jon,
* On 2009-06-30 at 15:27 +0200 Jon Olav Skoien wrote:
I work on two packages, pkg1 and pkg2 (in two different projects). pkg1
is quite generic, pkg2 tries to solve a particular problem within same field
(geostatistics). Therefore, there might be users who want to use pkg2 as an
add-on package to increase the functionality of pkg1. In other words,
functions in pkg1 are based on the S3 class system, and I want pkg2 to offer
methods for pkg2-objects to functions defined in pkg1, for users having both
packages installed. Merging the packages or making pkg2 always depend pkg1
would be the easiest solution, but it is not preferred as most users will
only be interested in one of the packages.
I'm not sure I understand the above, I think you may have a pkg2 where
you meant pkg1, but I'm not sure it matters.
I think the short version is, pkg2 can be used on its own but will do
more if pkg1 is available. I don't think R's packaging system
currently supports conditional dependencies as you might like.
However, I think you can get the behavior you want by following a
recipe like:
* In pkg2 DESCRIPTION, list Suggests: pkg1.
* In pkg2 code, you might define a package-level environment and in
.onLoad check to see if pkg1 is available.
PKG_INFO - new.env(parent=emptyenv())
.onLoad - function(libname, pkgname) {
if (check if pkg1 is available) {
PKG_INFO[[pkg1]] - TRUE
}
}
* Then your methods can check PKG_INFO[[pkg1]].
+ seth
__
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] V2.9.0 changes [Sec=Unclassified]
-Original Message-
From: Martin Morgan [mailto:mtmor...@fhcrc.org]
Sent: Thursday, 2 July 2009 10:58 PM
To: Troy Robertson
Cc: 'r-devel@R-project.org'
Subject: Re: [Rd] V2.9.0 changes [Sec=Unclassified]
Troy Robertson wrote:
Well...
My performance problems were in the pass-by-value semantics of R.
I have just changed my classes to inherit from .environment and then
moved data members from S4 slots to the .xData objects as Martin
suggested.
Actually, I had hoped the take-home message would be in the final
paragraph:
Of course I haven't seen your code, but a different interpretation of
your performance issues is that, within the rules of S4, you've chosen
to implement functionality in an inefficient way. So it might be
instructive to step back a bit and try to reformulate your data
structures and methods. This is hard to do.
Yes, it just takes a little time and playing around to work out the rules of R
(and S4) though, and how to use them to your advantage rather than be limited
by them. I know I am still probably not using this functional language in the
best way, but there you go.
That meant I could remove all my returns and assignments on all method
calls.
This has sped execution time for my model up by more than an order of
magnitude. Eg one test simulation from 1931 secs down to 175 secs.
Not bad seeing as though the class structure, functionality and logic
has not been touched.
I really do think S4 could benfit from having its slots stored in
environment when the class enherits from .environment. It would be a lot
more sensible if my data members were still declared as S4 slots instead
of having to hide them in .xData
Troy
Troy Robertson
Database and Computing Support Provider
Southern Ocean Ecosystems, ERM/Fish
Australian Antarctic Division
Channel Highway, Kingston 7050
PH: 03 62323571
troy.robert...@aad.gov.au
-Original Message-
From: Martin Morgan [mailto:mtmor...@fhcrc.org]
Sent: Tuesday, 23 June 2009 11:25 PM
To: Troy Robertson
Cc: 'r-devel@R-project.org'
Subject: Re: [Rd] V2.9.0 changes [Sec=Unclassified]
Troy Robertson wrote:
Hi all,
Prefix: I am a frustrated Java coder in R.
ah good, if you're frustrated with Java you'll find R very different ;)
I am coding a medium sized ecosystem modelling program in R. I have
changed to using S4 objects and it has cost me an order of magnitude in
execution speed over the functional model. I cannot afford this
penalty
and have found that it is the result of all the passing-by-value of
objects.
I see that you can now safely inherit from environment in V2.9.0.
That got me all excited that I would now be able to pass objects by
reference.
But...
That doesn't seem to be the case.
It only seem that passing an environment which holds the object allows
for pass-by-reference and that passing an object which inherits from
environment doesn't.
Why is this the case, either an object inherits the properties of its
parent or it doesn't.
The object inherits slots from it's parent, and the methods defined on
the parent class. Maybe this example helps?
setClass(Ticker, contains=.environment)
## initialize essential, so each instance gets its own environment
setMethod(initialize, Ticker,
function(.Object, ..., .xData=new.env(parent=emptyenv()))
{
.xData[[count]] - 0
callNextMethod(.Object, ..., .xData=.xData)
})
## tick: increment (private) counter by n
setGeneric(tick, function(reference, n=1L) standardGeneric(tick),
signature=reference)
setMethod(tick, Ticker, function(reference, n=1L) {
reference[[count]] - reference[[count]] + n
})
## tock: report current value of counter
setGeneric(tock, function(reference) standardGeneric(tock))
setMethod(tock, Ticker, function(reference) {
reference[[count]]
})
and then
e - new(Ticker)
tock(e)
[1] 0
tick(e); tick(e, 10); tock(e)
[1] 11
f - e
tock(f); tick(e); tock(f)
[1] 11
[1] 12
The data inside .environment could be structured, too, using S4.
Probably it would be more appropriate to have the environment as a
slot,
rather the class that is being extended. And in terms of inherited
'properties', e.g., the [[ function as defined on environments is
available
e[[count]]
Of course I haven't seen your code, but a different interpretation of
your performance issues is that, within the rules of S4, you've chosen
to implement functionality in an inefficient way. So it might be
instructive to step back a bit and try to reformulate your data
structures and methods. This is hard to do.
Martin
Has anyone else had a play with this? Or have I got it all wrong.
I tried the below:
--
--
-
setClass('foo',
Re: [Rd] OOP performance, was: V2.9.0 changes [SEC=Unclassified]
Hi Thomas, It is a population-based model, but I didn't develop the work. I am just the programmer who has been given the job of coding it. The goal is to allow for a plug and play type approach by users to construction of the model (of both elements and functionality). Hence my focus on OO. You are right about avoiding the copying of large objects. That is what was killing things. I am now working on vectorizing more of the number crunching and removing some of the nested for loops. That should step things up a little too. I do also need to investigate how to move some of the more expensive code to C. Had a quick look at simecol which looks interesting. Will point it out to my boss to check out too. Thanks Troy -Original Message- From: Thomas Petzoldt [mailto:thomas.petzo...@tu-dresden.de] Sent: Friday, 3 July 2009 1:31 AM To: Troy Robertson Cc: 'r-devel@R-project.org' Subject: OOP performance, was: [Rd] V2.9.0 changes [SEC=Unclassified] Hi Troy, first of all a question, what kind of ecosystem models are you developing in R? Differential equations or individual-based? Your write that you are a frustrated Java developer in R. I have a similar experience, however I still like JAVA, and I'm now more happy with R as it is much more efficient (i.e. sum(programming + runtime)) for the things I usually do: ecological data analysis and modelling. After using functional R quite a time and Java in parallel I had the same idea, to make R more JAVA like and to model ecosystems in an object oriented manner. At that time I took a look into R.oo (thanks Henrik Bengtssson) and was one of the Co-authors of proto. I still think that R.oo is very good and that proto is a cool idea, but finally I switched to the recommended S4 for my ecological simulation package. Note also, that my solution was *not* to model the ecosystems as objects (habitat - populations- individuals), but instead to model ecological models (equations, inputs, parameters, time steps, outputs, ...). This works quite well with S4. A speed test (see useR!2006 poster on http://simecol.r-forge.r-project.org/) showed that all OOP flavours had quite comparable performance. The only thing I have to have in mind are a few rules: - avoid unnecessary copying of large objects. Sometimes it helps to prefer matrices over data frames. - use vectorization. This means for an individual-based model that one has to re-think how to model an individual: not many [S4] objects like in JAVA, but R structures (arrays, lists, data frames) where vectorized functions (e.g. arithmetics or subset) can work with. - avoid interpolation (i.e. approx) and if unavoidable, minimize the tables. If all these things do not help, I write core functions in C (others use Fortran). This can be done in a mixed style and even a full C to C communication is possible (see the deSolve documentation how to do this with differential equation models). Thomas P. -- Thomas Petzoldt Technische Universitaet Dresden Institut fuer Hydrobiologiethomas.petzo...@tu-dresden.de 01062 Dresden http://tu-dresden.de/hydrobiologie/ GERMANY ___ Australian Antarctic Division - Commonwealth of Australia IMPORTANT: This transmission is intended for the addressee only. If you are not the intended recipient, you are notified that use or dissemination of this communication is strictly prohibited by Commonwealth law. If you have received this transmission in error, please notify the sender immediately by e-mail or by telephoning +61 3 6232 3209 and DELETE the message. Visit our web site at http://www.antarctica.gov.au/ ___ __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] [R] ggplot2 x axis question
On Mon, Jun 29, 2009 at 9:05 AM, hadley wickhamh.wick...@gmail.com wrote: In that case, try: qplot(reorder(factor(model),delta),delta,data=growthm.bic) Deepayan: do you think there should also be a numeric method for reorder? r-devel now has a reorder.default (replacing reorder.factor and reorder.character), so reorder() should also work for numeric data. -Deepayan Hadley On Mon, Jun 29, 2009 at 10:39 AM, Christopher Desjardinscddesjard...@gmail.com wrote: Hi Hadley, Thanks for the reply and the great graphing package. That code is giving me the following error: qplot(reorder(model,delta),delta,data=growthm.bic) Error in UseMethod(reorder) : no applicable method for reorder Cheers, Chris [...] __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] OOP performance, was: V2.9.0 changes
In terms of performance if you want the fastest performance in R go with S3 and if you want even faster performance rewrite your inner loops in C. All the other approaches will usually be slower. Also S3 is simple, elegant and will result in less code and take you much less time to design, program and debug. For 100% R code, particularly for simulations, proto can sometimes be even faster than pure R code based S3 as proto supports hand optimizations that cannot readily be done in other systems. (For unoptimized code it would be slower.) The key trick is based on its ability to separate dispatching from calling so that if method f and object p are unchanged in a loop for(...) p$f(...) then the loop can be rewritten f - p$f; for(...) f(...) Note that this still retains dynamic dispatch but just factors it out of the loop. With S3 the best you could get would be for(...) f.p(...) where f is a method of class p but this is really tantamount to not using OO at all since no dispatch is done at all. On Thu, Jul 2, 2009 at 11:31 AM, Thomas Petzoldtthomas.petzo...@tu-dresden.de wrote: Hi Troy, first of all a question, what kind of ecosystem models are you developing in R? Differential equations or individual-based? Your write that you are a frustrated Java developer in R. I have a similar experience, however I still like JAVA, and I'm now more happy with R as it is much more efficient (i.e. sum(programming + runtime)) for the things I usually do: ecological data analysis and modelling. After using functional R quite a time and Java in parallel I had the same idea, to make R more JAVA like and to model ecosystems in an object oriented manner. At that time I took a look into R.oo (thanks Henrik Bengtssson) and was one of the Co-authors of proto. I still think that R.oo is very good and that proto is a cool idea, but finally I switched to the recommended S4 for my ecological simulation package. Note also, that my solution was *not* to model the ecosystems as objects (habitat - populations- individuals), but instead to model ecological models (equations, inputs, parameters, time steps, outputs, ...). This works quite well with S4. A speed test (see useR!2006 poster on http://simecol.r-forge.r-project.org/) showed that all OOP flavours had quite comparable performance. The only thing I have to have in mind are a few rules: - avoid unnecessary copying of large objects. Sometimes it helps to prefer matrices over data frames. - use vectorization. This means for an individual-based model that one has to re-think how to model an individual: not many [S4] objects like in JAVA, but R structures (arrays, lists, data frames) where vectorized functions (e.g. arithmetics or subset) can work with. - avoid interpolation (i.e. approx) and if unavoidable, minimize the tables. If all these things do not help, I write core functions in C (others use Fortran). This can be done in a mixed style and even a full C to C communication is possible (see the deSolve documentation how to do this with differential equation models). Thomas P. -- Thomas Petzoldt Technische Universitaet Dresden Institut fuer Hydrobiologie thomas.petzo...@tu-dresden.de 01062 Dresden http://tu-dresden.de/hydrobiologie/ GERMANY __ 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] OOP performance, was: V2.9.0 changes [SEC=Unclassified]
Hi Gabor, Look, I agree with you about S3 and have at times wished I had chosen that path rather than S4. It seems to do the things I struggle to find answers for with S4. But..., knowing little about R before engaging with this project, I decided to go with the latest OO framework, S4. I do now find that I am undoing some of it, such as the use of data member slots, in order to implement pass-by-ref via environments and improve performance. But its all a learning experience. Troy -Original Message- From: Gabor Grothendieck [mailto:ggrothendi...@gmail.com] Sent: Friday, 3 July 2009 10:29 AM To: Troy Robertson; r-devel@R-project.org Subject: Re: [Rd] OOP performance, was: V2.9.0 changes [SEC=Unclassified] In terms of performance if you want the fastest performance in R go with S3 and if you want even faster performance rewrite your inner loops in C. All the other approaches will usually be slower. Also S3 is simple, elegant and will result in less code and take you much less time to design, program and debug. For 100% R code, particularly for simulations, proto can sometimes be even faster than pure R code based S3 as proto supports hand optimizations that cannot readily be done in other systems. (For unoptimized code it would be slower.) The key trick is based on its ability to separate dispatching from calling so that if method f and object p are unchanged in a loop for(...) p$f(...) then the loop can be rewritten f - p$f; for(...) f(...) Note that this still retains dynamic dispatch but just factors it out of the loop. With S3 the best you could get would be for(...) f.p(...) where f is a method of class p but this is really tantamount to not using OO at all since no dispatch is done at all. On Thu, Jul 2, 2009 at 11:31 AM, Thomas Petzoldtthomas.petzo...@tu-dresden.de wrote: Hi Troy, first of all a question, what kind of ecosystem models are you developing in R? Differential equations or individual-based? Your write that you are a frustrated Java developer in R. I have a similar experience, however I still like JAVA, and I'm now more happy with R as it is much more efficient (i.e. sum(programming + runtime)) for the things I usually do: ecological data analysis and modelling. After using functional R quite a time and Java in parallel I had the same idea, to make R more JAVA like and to model ecosystems in an object oriented manner. At that time I took a look into R.oo (thanks Henrik Bengtssson) and was one of the Co-authors of proto. I still think that R.oo is very good and that proto is a cool idea, but finally I switched to the recommended S4 for my ecological simulation package. Note also, that my solution was *not* to model the ecosystems as objects (habitat - populations- individuals), but instead to model ecological models (equations, inputs, parameters, time steps, outputs, ...). This works quite well with S4. A speed test (see useR!2006 poster on http://simecol.r-forge.r-project.org/) showed that all OOP flavours had quite comparable performance. The only thing I have to have in mind are a few rules: - avoid unnecessary copying of large objects. Sometimes it helps to prefer matrices over data frames. - use vectorization. This means for an individual-based model that one has to re-think how to model an individual: not many [S4] objects like in JAVA, but R structures (arrays, lists, data frames) where vectorized functions (e.g. arithmetics or subset) can work with. - avoid interpolation (i.e. approx) and if unavoidable, minimize the tables. If all these things do not help, I write core functions in C (others use Fortran). This can be done in a mixed style and even a full C to C communication is possible (see the deSolve documentation how to do this with differential equation models). Thomas P. -- Thomas Petzoldt Technische Universitaet Dresden Institut fuer Hydrobiologiethomas.petzo...@tu-dresden.de 01062 Dresden http://tu-dresden.de/hydrobiologie/ GERMANY __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel ___ Australian Antarctic Division - Commonwealth of Australia IMPORTANT: This transmission is intended for the addressee only. If you are not the intended recipient, you are notified that use or dissemination of this communication is strictly prohibited by Commonwealth law. If you have received this transmission in error, please notify the sender immediately by e-mail or by telephoning +61 3 6232 3209 and DELETE the message. Visit our web site at http://www.antarctica.gov.au/ ___ __ R-devel@r-project.org mailing
