R SIG Debian - Apr 2005 - "Debain" way of installing packages

Hello!

I use Debian GNU/Linux at work and I really like apt-get tool. As far as
I understand the whole picture, there are now some Debian packages of R
packages. Build of Debian packages is done automagically by Perl script.

By the way, what is the status of Debianizing R packages? Is it stopped, 
just holded, ... I don't know so I am asking
There is also Bioconductor and I am aware that the list of packages is 
growing quite fast and afterall I understand that there is probably to 
few people to work on Debianization.

The main point of this mail is however installation. I usually first
try to use apt-get install and if there is no Debian package I use
R tools. However some packages that I have already installed with R tools
might  eventually appear as Debian packages, ... or I might use 
update.packages() for Debian packages. By the way is field
'Package-Manager: Debian' already implemented. I would check, but I 
unfortunately don't have Debain at home.

I was thinking on some functions or eventually an R package r-debian,
which would ease this installation dilemma. Actually I wrote something
already in december but put it off. It is given bellow. It is probably
far from optimal but it was just a proof of concept. Many more functions
would be needed to fuzz everything up. Any comments?

debian.install.packages <- function(package, ...) {

    # Description:
    # Installs package on Debian GNU/Linux system. For given R package it
    # tries to get Debian package and install/upgrade it with apt-get if it
    # exists.  Otherwise install.packages() is used.
    # Gregor GORJANC 2004-12-30

    # Arguments:
    # package: R package name as character, only one
    # ...: other parameters for install.packages() or update.packages()

    # Examples:
    # debian.install.packages("Design")

    # Code:
    # apt-get update i.e. refresh list of Debian packages
    system(apt-get update)

    # Transform R package name to Debian package name policy i.e change R
    # package name to lowercase and test if this Debian package exists
    # (test1). This could be done without grep, but then it is not possible
    # to test on exit status, since apt-cache search always returns 0.
    packagelow <- tolower(package)
    tmp1 <- c("system(\"apt-cache search ^r- | grep")
    tmp2 <- c("| awk '{print $1}'\", intern = TRUE\)")
    tmp3 <- as.expression(paste(tmp1, packagelow, tmp2, "\n"))
    deb <- eval(parse(file = "", text = tmp3))

    cat("\nR package:", package, "\n")
    cat("Corresponding Debian package name:", deb, "\n")

    # Package install/update or upgrade in apt meaning
    if (length(deb) != 0) { # Debian package does exist
        cat("Debian package", deb, "exists: yes\n")
        cat("Proceeding with apt-get.\n")

        # apt-get install, which can also be used for package upgrade
        tmp1 <- c("system(\"apt-get install")
        tmp2 <- as.expression(paste(tmp1, deb, "\")",
"\n"))
        eval(parse(file = "", text = tmp2))
    }
    else { # Debian package doesn't exist or something went wrong in test1
        cat("Debian package", deb, "exists: no\n")
#        cat("Proceeding with install/update.packages().\n")
#        install.update.packages(package, ...)
        cat("Proceeding with install.packages().\n")
        install.packages(package, ...)
    }
}

One more thing. Why is r-noncran-lindsey and not r-other-lindsey?

--
Lep pozdrav / With regards,
    Gregor Gorjanc

------------------------------------------------------------------------
University of Ljubljana
Biotechnical Faculty       URI: http://www.bfro.uni-lj.si/MR/ggorjan
Zootechnical Department    email: gregor.gorjanc <at> bfro.uni-lj.si
Groblje 3                  tel: +386 (0)1 72 17 861
SI-1230 Domzale            fax: +386 (0)1 72 17 888
Slovenia

YMMV, but what I do is to install duplicates (i.e. R CMD INSTALL
places things in /usr/local/, and apt-get in /usr/), and resolve the
library I want at load time.

This has a few advantages, and the obvious disadvantage that you can
shoot yourself in the foot REALLY EASILY if you don't pay attention to
what you've got configured.

best,
-tony



On 4/24/05, Gorjanc Gregor <Gregor.Gorjanc@bfro.uni-lj.si>
wrote:
> Hello!
> 
> I use Debian GNU/Linux at work and I really like apt-get tool. As far as
> I understand the whole picture, there are now some Debian packages of R
> packages. Build of Debian packages is done automagically by Perl script.
> 
> By the way, what is the status of Debianizing R packages? Is it stopped,
> just holded, ... I don't know so I am asking
> There is also Bioconductor and I am aware that the list of packages is
> growing quite fast and afterall I understand that there is probably to
> few people to work on Debianization.
> 
> The main point of this mail is however installation. I usually first
> try to use apt-get install and if there is no Debian package I use
> R tools. However some packages that I have already installed with R tools
> might  eventually appear as Debian packages, ... or I might use
> update.packages() for Debian packages. By the way is field
> 'Package-Manager: Debian' already implemented. I would check, but I
> unfortunately don't have Debain at home.
> 
> I was thinking on some functions or eventually an R package r-debian,
> which would ease this installation dilemma. Actually I wrote something
> already in december but put it off. It is given bellow. It is probably
> far from optimal but it was just a proof of concept. Many more functions
> would be needed to fuzz everything up. Any comments?
> 
> debian.install.packages <- function(package, ...) {
> 
>     # Description:
>     # Installs package on Debian GNU/Linux system. For given R package it
>     # tries to get Debian package and install/upgrade it with apt-get if it
>     # exists.  Otherwise install.packages() is used.
>     # Gregor GORJANC 2004-12-30
> 
>     # Arguments:
>     # package: R package name as character, only one
>     # ...: other parameters for install.packages() or update.packages()
> 
>     # Examples:
>     # debian.install.packages("Design")
> 
>     # Code:
>     # apt-get update i.e. refresh list of Debian packages
>     system(apt-get update)
> 
>     # Transform R package name to Debian package name policy i.e change R
>     # package name to lowercase and test if this Debian package exists
>     # (test1). This could be done without grep, but then it is not possible
>     # to test on exit status, since apt-cache search always returns 0.
>     packagelow <- tolower(package)
>     tmp1 <- c("system(\"apt-cache search ^r- | grep")
>     tmp2 <- c("| awk '{print $1}'\", intern =
TRUE\)")
>     tmp3 <- as.expression(paste(tmp1, packagelow, tmp2, "\n"))
>     deb <- eval(parse(file = "", text = tmp3))
> 
>     cat("\nR package:", package, "\n")
>     cat("Corresponding Debian package name:", deb,
"\n")
> 
>     # Package install/update or upgrade in apt meaning
>     if (length(deb) != 0) { # Debian package does exist
>         cat("Debian package", deb, "exists: yes\n")
>         cat("Proceeding with apt-get.\n")
> 
>         # apt-get install, which can also be used for package upgrade
>         tmp1 <- c("system(\"apt-get install")
>         tmp2 <- as.expression(paste(tmp1, deb, "\")",
"\n"))
>         eval(parse(file = "", text = tmp2))
>     }
>     else { # Debian package doesn't exist or something went wrong in
test1
>         cat("Debian package", deb, "exists: no\n")
> #        cat("Proceeding with install/update.packages().\n")
> #        install.update.packages(package, ...)
>         cat("Proceeding with install.packages().\n")
>         install.packages(package, ...)
>     }
> }
> 
> One more thing. Why is r-noncran-lindsey and not r-other-lindsey?
> 
> --
> Lep pozdrav / With regards,
>     Gregor Gorjanc
> 
> ------------------------------------------------------------------------
> University of Ljubljana
> Biotechnical Faculty       URI: http://www.bfro.uni-lj.si/MR/ggorjan
> Zootechnical Department    email: gregor.gorjanc <at> bfro.uni-lj.si
> Groblje 3                  tel: +386 (0)1 72 17 861
> SI-1230 Domzale            fax: +386 (0)1 72 17 888
> Slovenia
> 
> _______________________________________________
> R-SIG-Debian mailing list
> R-SIG-Debian@r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-debian
> 

-- 
best,
-tony

"Commit early,commit often, and commit in a repository from which we can
easily
roll-back your mistakes" (AJR, 4Jan05).

A.J. Rossini
blindglobe@gmail.com

Hi Gregor,

Your enthusiasm in this matter is greatly appreciated. I am CC'ing this to
the existing list on the alioth.debian.org hosting system: pkg-bioc-devel.
This was initially set up my Matt Hope to foster debianising BioC, I then
threw my exisiting Perl infrastructure (used previously for Quantian) at it
and Egon Willighagen carried it from there with further improvements,
incl. large scale builds of amd64 packages.

With a little push from all of us (Hi Tony, Hi Matt, Hi Egon, Hi
anybody-else-willing-to-chip-in :) we can make this even better and hopefully
have something proper by the summer, maybe for DSC 20003.

On 24 April 2005 at 16:04, Gorjanc Gregor wrote:
| Hello!
| 
| I use Debian GNU/Linux at work and I really like apt-get tool. As far as
| I understand the whole picture, there are now some Debian packages of R
| packages. Build of Debian packages is done automagically by Perl script.
| 
| By the way, what is the status of Debianizing R packages? Is it stopped, 
| just holded, ... I don't know so I am asking
| There is also Bioconductor and I am aware that the list of packages is 
| growing quite fast and afterall I understand that there is probably to 
| few people to work on Debianization.
| 
| The main point of this mail is however installation. I usually first
| try to use apt-get install and if there is no Debian package I use
| R tools. However some packages that I have already installed with R tools
| might  eventually appear as Debian packages, ... or I might use 
| update.packages() for Debian packages. By the way is field
| 'Package-Manager: Debian' already implemented. I would check, but I 
| unfortunately don't have Debain at home.
| 
| I was thinking on some functions or eventually an R package r-debian,
| which would ease this installation dilemma. Actually I wrote something
| already in december but put it off. It is given bellow. It is probably
| far from optimal but it was just a proof of concept. Many more functions
| would be needed to fuzz everything up. Any comments?
| 
| debian.install.packages <- function(package, ...) {
| 
|     # Description:
|     # Installs package on Debian GNU/Linux system. For given R package it
|     # tries to get Debian package and install/upgrade it with apt-get if it
|     # exists.  Otherwise install.packages() is used.
|     # Gregor GORJANC 2004-12-30
| 
|     # Arguments:
|     # package: R package name as character, only one
|     # ...: other parameters for install.packages() or update.packages()
| 
|     # Examples:
|     # debian.install.packages("Design")
| 
|     # Code:
|     # apt-get update i.e. refresh list of Debian packages
|     system(apt-get update)
| 
|     # Transform R package name to Debian package name policy i.e change R
|     # package name to lowercase and test if this Debian package exists
|     # (test1). This could be done without grep, but then it is not possible
|     # to test on exit status, since apt-cache search always returns 0.
|     packagelow <- tolower(package)
|     tmp1 <- c("system(\"apt-cache search ^r- | grep")
|     tmp2 <- c("| awk '{print $1}'\", intern =
TRUE\)")
|     tmp3 <- as.expression(paste(tmp1, packagelow, tmp2, "\n"))
|     deb <- eval(parse(file = "", text = tmp3))
| 
|     cat("\nR package:", package, "\n")
|     cat("Corresponding Debian package name:", deb, "\n")
| 
|     # Package install/update or upgrade in apt meaning
|     if (length(deb) != 0) { # Debian package does exist
|         cat("Debian package", deb, "exists: yes\n")
|         cat("Proceeding with apt-get.\n")
| 
|         # apt-get install, which can also be used for package upgrade
|         tmp1 <- c("system(\"apt-get install")
|         tmp2 <- as.expression(paste(tmp1, deb, "\")",
"\n"))
|         eval(parse(file = "", text = tmp2))
|     }
|     else { # Debian package doesn't exist or something went wrong in test1
|         cat("Debian package", deb, "exists: no\n")
| #        cat("Proceeding with install/update.packages().\n")
| #        install.update.packages(package, ...)
|         cat("Proceeding with install.packages().\n")
|         install.packages(package, ...)
|     }
| }

As Tony said in his follow-up, the mixing and matching is hard to do
properly, which is why I think the high-end solution is to provide a robust
script, maybe with a db backend, that debianises what is on CRAN -- similar
to what exists on Windows. We'd simply use apt-get instead of
install.packages() wherever possible.
 
| One more thing. Why is r-noncran-lindsey and not r-other-lindsey?

Because nobody pressed Chris hard enough to implement the change. I am sure
he would take a patch from you (hint, hint).

Cheers, Dirk

PS Adorable type in you Subject line. I have done gazillion times as well :)

| 
| --
| Lep pozdrav / With regards,
|     Gregor Gorjanc
| 
| ------------------------------------------------------------------------
| University of Ljubljana
| Biotechnical Faculty       URI: http://www.bfro.uni-lj.si/MR/ggorjan
| Zootechnical Department    email: gregor.gorjanc <at> bfro.uni-lj.si
| Groblje 3                  tel: +386 (0)1 72 17 861
| SI-1230 Domzale            fax: +386 (0)1 72 17 888
| Slovenia
| 
| _______________________________________________
| R-SIG-Debian mailing list
| R-SIG-Debian@r-project.org
| https://stat.ethz.ch/mailman/listinfo/r-sig-debian

-- 
Better to have an approximate answer to the right question than a precise 
answer to the wrong question.  --  John Tukey as quoted by John Chambers

On 4/24/05, Dirk Eddelbuettel <edd@debian.org>
wrote:
> Because nobody pressed Chris hard enough to implement the change. I am sure
> he would take a patch from you (hint, hint).
-lindsey is sufficiently annoying for this and other reasons that I
plan to either orphan it or ask for its removal.


Chris
-- 
Chris Lawrence - http://blog.lordsutch.com/

Hello!

Sorry for late response. I just had to check all this allioth stuf etc. 
> -----Original Message-----
> From: Dirk Eddelbuettel [mailto:edd@debian.org]
> Sent: ned 2005-04-24 19:45
> To: Gorjanc Gregor
> Cc: r-sig-debian@r-project.org; pkg-bioc-devel@lists.alioth.debian.org;
lawrencc@debian.org
> Subject: Re: [R-sig-Debian] "Debain" way of installing packages
>  
> Hi Gregor,
> 
> Your enthusiasm in this matter is greatly appreciated. I am CC'ing this
to
> the existing list on the alioth.debian.org hosting system: pkg-bioc-devel.
> This was initially set up my Matt Hope to foster debianising BioC, I then
> threw my exisiting Perl infrastructure (used previously for Quantian) at it
> and Egon Willighagen carried it from there with further improvements,
> incl. large scale builds of amd64 packages.
> 
> With a little push from all of us (Hi Tony, Hi Matt, Hi Egon, Hi
> anybody-else-willing-to-chip-in :) we can make this even better and
hopefully
> have something proper by the summer, maybe for DSC 20003.
[... snip ...]
> As Tony said in his follow-up, the mixing and matching is hard to do
Yes, mixing is hard and confusing.
> properly, which is why I think the high-end solution is to provide a robust
> script, maybe with a db backend, that debianises what is on CRAN -- similar
What do you mean by a db (I suppose databse) backend, that debianises ...?

If I understand you propose to debianize all R packages and then simply use
apt-get. This would probably be the easisest way, however only if this is really
achivable. My help on debianization can be very small since I have never tried
to do that, but I am willing to learn and help.

If previous is true, you think that some sort script I proposed is not OK for
mixing?
> to what exists on Windows. We'd simply use apt-get instead of
What is the case with Windows? Do you mean build of binary packages for Windows
or what?
> | One more thing. Why is r-noncran-lindsey and not r-other-lindsey?
> Because nobody pressed Chris hard enough to implement the change. I am sure
> he would take a patch from you (hint, hint).
>From words of Chris this doesn't seems to be needed, however I don't
know what is
the problem with that package from Lindsey.

Gregor

Hello!
> ----Original Message-----
> From: Dirk Eddelbuettel [mailto:edd@debian.org]
> Sent: sre 2005-04-27 02:35
> To: Gorjanc Gregor
> Cc: r-sig-debian@r-project.org; pkg-bioc-devel@lists.alioth.debian.org;
Chris Lawrence
> Subject: RE: [R-sig-Debian] "Debain" way of installing packages
>  
> On 26 April 2005 at 12:11, Gorjanc Gregor wrote:
> | > properly, which is why I think the high-end solution is to provide a
robust
> | > script, maybe with a db backend, that debianises what is on CRAN --
similar
> | What do you mean by a db (I suppose databse) backend, that debianises
...?
> 
> Yes, to keep track of versions, of what has been packaged, of meta data as
> e.g. mapping of package names between Debian and CRAN, of exceptions (CRAN
is
> somwhat more lenient with licenses) etc pp.  "Debianising" is
stateful, so we
> need memory. Flat files may well do for the time being before we switch to
an
> actual db (which may just be sqlite, ie a binary file as backend).
OK, I see. It's clearer to me now. Thanks!
> | If I understand you propose to debianize all R packages and then simply
use
> | apt-get. This would probably be the easisest way, however only if this is
really
> | achivable. My help on debianization can be very small since I have never
tried
> | to do that, but I am willing to learn and help.
> | 
> | If previous is true, you think that some sort script I proposed is not OK
for
> | mixing?
> 
> We actually have scripts that "work", they simply need testing,
extending and
> polishing. If you register yourself on alioth.d.o, you can check it out of
> the archive and play with it.
I will try that and play with scripts. However my idea was to provide apt-get 
within R i.e. through "debian.install.packages" or something like
that. What
do you think about that feature? It would be nice to have since we can have
all packages debianized, but if someone uses install.packages then ... However,
it is truth that Debian packages will be loaded first, since its lib is by
default before R packages installed by install.packages. Am I right?
> | What is the case with Windows? Do you mean build of binary packages for
Windows
> | or what?
> 
> That was just an example as CRAN provides binaries for Windows. My medium
to
> longer-term vision is binaries for Debian and its derivative distros.
How does Uwe Ligges perform all those binary builds for Windows?

Regards, Gregor