[R-SIG-Mac] upgradeable R layout

Fri Nov 14 04:53:13 CET 2008

> -----Original Message-----
> From: Alexy Khrabrov [mailto:deliverable at gmail.com]
> Sent: Thursday, November 13, 2008 7:13 PM
> To: Steven McKinney
> Cc: r-sig-mac at stat.math.ethz.ch
> Subject: Re: [R-SIG-Mac] upgradeable R layout
> 
> Steven -- thanks a lot!  Now it's clearer what to do with the 64 bits,
> but still it doesn't answer the question of upgradeability -- meaning,
> you went ahead and installed a bunch of packages, then a new version
> of R comes along, and what do you do?

With R 2.8.0, as with any major release change, structure 
can change substantially, including with associated software 
such as the Bioconductor set of packages, so I have not 
found a useful way to just put everything back that was in 
a prior release.

> 
> One way I can see handling it it with a local install, outside of the
> framework, and reinstalling with the command from the previous list --
> and you still have to do that.  Going over a list of 60+ packages in
> the installer is not really fun...

No, sometimes I'd rather be poking a stick in my eye,
but other things can make it fun, like remembering how
few packages were available with other commercial
statistical software programs, or how hard it was to
get technical support or answers about internal code
from their tech support, or how much money I had to 
spend ;-) while I install the R packages.  (Though
once I donated to the R project via PayPal while
nursing a new installation!)

I've learned that some R / Bioconductor packages require 
many others, and thanks to improved dependency checking 
and installation I can install e.g. snapCGH in the Bioconductor 
suite and get dozens of other packages installed as a side 
effect.  I can select several packages in the R.app and
click the beautiful "Install Selected" button, and mostly
it all works in one go, so I'm no longer typing in 60+ 
install.packages() at the command prompt.

Smarter people than I may have more useful techniques
for you here.  Sorry.

Best

SteveM

> 
> Cheers,
> Alexy
> 
> On Nov 13, 2008, at 9:03 PM, Steven McKinney wrote:
> 
> >
> >
> > One last bit - I set up a shell script in
> > /usr/bin
> >
> > called R64
> >
> > containing
> >
> > #!/bin/sh
> > R --arch=ppc64
> >
> > to allow easy command line startup of
> > 64 bit R and to allow ESS to find
> > and start up 64 bit R sessions in Emacs.
> >
> >
> >> -----Original Message-----
> >> From: r-sig-mac-bounces at stat.math.ethz.ch [mailto:r-sig-mac-
> >> bounces at stat.math.ethz.ch] On Behalf Of Steven McKinney
> >> Sent: Thursday, November 13, 2008 5:49 PM
> >> To: Alexy Khrabrov; r-sig-mac at stat.math.ethz.ch
> >> Subject: Re: [R-SIG-Mac] upgradeable R layout
> >>
> >> Greetings Alexy,
> >>
> >>
> >>> -----Original Message-----
> >>> From: r-sig-mac-bounces at stat.math.ethz.ch [mailto:r-sig-mac-
> >>> bounces at stat.math.ethz.ch] On Behalf Of Alexy Khrabrov
> >>> Sent: Thursday, November 13, 2008 4:29 PM
> >>> To: r-sig-mac at stat.math.ethz.ch
> >>> Subject: [R-SIG-Mac] upgradeable R layout
> >>>
> >>> Greetings -- I'd like to keep my R on Mac in an upgradeable way. I
> >>> noticed that currently all packages I installed from R.app are in
> > the
> >>> R.framework -- are they going to be clobbered when I replace R
2.7.2
> >>> by 2.8.0?
> >>
> >> There are quite a few changes between R 2.7.2 and R 2.8.0
> >>
> >> I abandoned my R 2.7.2 and have no regrets so far (but you
> >> don't have to - you can have different versions and Simon Urbanek
> >> provides a small RSwitch gui app that lets you switch back and
> >> forth).
> >>
> >>>
> >>> At the same time with the upgrade, I'd also like to switch to 64
> > bit.
> >>> One good page I found about it is
> >>>
> >>> http://www.matthewckeller.com/html/64_bit_r_on_mac.html
> >>>
> >>> It suggests that packages are installed into a separate directory
by
> >>> doing it from command line install.packages with a lib= parameter.
> >>
> >> This website is out of date because of the changes
> >> in place for R 2.8.0
> >>
> >> For R 2.8.0 if you look into one of a package's library
> >> directories, e.g.
> >>
> > /MacintoshHD/Library/Frameworks/R.framework/Versions/2.8/Resources/
> > libra
> >> ry/Matrix
> >>
> >> you will see a subdirectory
> >> /libs
> >> which will contain further subdirectories (once you've installed
for
> > the
> >> arch involved), e.g.
> >> /ppc
> >> /pp64
> >> which hold the architecture-specific files.  So you don't have to
> >> keep
> > a
> >> 64 bit version somewhere separate from your 32 bit version.  Nice.
> >>
> >> I installed R from Simon Urbanek's website - always up to date -
> >> http://r.research.att.com/
> >> using the method
> >>
> >> Leopard builds can be installed as follows - paste in Terminal (for
R
> >> 2.8.x):
> >> curl -s
> > http://r.research.att.com/R-2.8-branch-leopard-universal.tar.gz
> >> | sudo tar fvxz - -C /
> >>
> >> after setting up the various compilers and other software discussed
> >> in the "Tools for R Development" and "Other binaries and tools"
> >> sections.
> >>
> >> Now you have R installed and it has all architectures ready.
> >> (I've wrestled through 64 bit configure / make / install for
> >> prior R versions, this is much better!  Thank you Simon Urbanek
> >> and others!)
> >>
> >> I then installed the 32 bit R-Gui and renamed it to R32.app.
> >> I installed the 64 bit R-Gui and renamed it to R64.app,
> >> so I can fire up either 32 bit or 64 bit (or both!) and
> >> use the excellent package manager and package installer.
> >> The package installer installs the package using the
> >> architecture of the R-Gui it is invoked from, so architecture
> >> is handled automagically.
> >>
> >>>
> >>> Now I'd like to install everything into a lib anyways, but
> > preferably
> >>> without having to specify lib= all the time -- is there a way to
> > make
> >>> my local location the default for install.packages?
> >>
> >> With the R-Gui apps in place, you can install packages easily, they
> >> are installed in their standard place and I've had no problems with
> >> this.
> >>
> >>>
> >>> Also -- can the R.app be tweaked to do all this, if I want to use
it
> >>> with the 64 bit and local install for packages?
> >>
> >> Yes - one way is described above.
> >>
> >>>
> >>> The idea is to make upgrading R smooth -- ideally, it would do
> >>> update.packages(checkBuilt =
> >>> TRUE) by itself as a part of the upgrade.
> >>
> >> The only hiccup I encountered was a TCL/TK issue (now resolved)
> >> but otherwise getting 32 bit and 64 bit R up and running
> >> has been very smooth.  (Now the issue is that ppc64 is going
> >> extinct in the not too distant future - time to put in a
> >> request for an Intel Mac!)
> >>
> >>>
> >>> Are there other ways to simplify R upgrade on Mac and/or am I
> > missing
> >>> something?
> >>
> >> Once you get set up as I have, you won't be missing anything!
> >>
> >> Good luck
> >>
> >>>
> >>> Cheers,
> >>> Alexy
> >>>
> >>> _______________________________________________
> >>> R-SIG-Mac mailing list
> >>> R-SIG-Mac at stat.math.ethz.ch
> >>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
> >>
> >>
> >> Steven McKinney
> >>
> >> Statistician
> >> Molecular Oncology and Breast Cancer Program
> >> British Columbia Cancer Research Centre
> >>
> >> email: smckinney at bccrc.ca
> >> tel: 604-675-8000 x7561
> >>
> >> BCCRC
> >> Molecular Oncology
> >> 675 West 10th Ave, Floor 4
> >> Vancouver B.C.
> >> V5Z 1L3
> >>
> >> Canada
> >>
> >> _______________________________________________
> >> R-SIG-Mac mailing list
> >> R-SIG-Mac at stat.math.ethz.ch
> >> https://stat.ethz.ch/mailman/listinfo/r-sig-mac