[R-SIG-Mac] Macports installation problem

David Winsemius dwinsemius at comcast.net
Sun Oct 22 19:31:08 CEST 2017


> On Oct 21, 2017, at 5:47 AM, clark richards <clark.richards at gmail.com> wrote:
> 
> Interesting discussion, even if it is technically off-topic (i.e. is about a Macports issue rather than an R issue).
> 
> For myself, I have found instances where I needed to install R through non-standard channels (my choice lately has been homebrew). This was largely related to having the required system tools necessary for a package development environment that permits the compilation of C/fortran code and some system libraries (gdal, etc). As Don and others pointed out many such dependencies should be available through the OSX R develop page (forget the address right now ...), I have some older machines for which those precompiled binaries do not install, including a 9 year old Macbook which will run OSX up to ElCapitan but is not compatible with the newer binaries. In that case, in order to have an R and gfortran that played together I had to use homebrew as I could not find a binary gfortran that would work for my architecture.

I have two older Macs of roughly the same vintage, both running El Cap. Regarding gfortran for package devel, I wondered if you had noticed that advice at the top of the CRAN Tools page to add some locations to a Makevars file:

https://cran.r-project.org/bin/macosx/tools/

I was having difficulties until someone gently reminded me to "read the manual".

GDAL (and associated packages) are generally obtained from Kyngchaos and that has been the case for as long as I remember.

-- 
David.


> 
> I love working on OSX, though I have often lamented the difficulties in getting a suitable package development environment configured -- at least compared to Windows (just install Rtools.exe) and Linux (everything required is easily installed from the distro package manager).
> 
> Clark
> 
> 
> Message: 2
> Date: Fri, 20 Oct 2017 18:02:33 +0000
> From: "MacQueen, Don" <macqueen1 at llnl.gov>
> To: David Winsemius <dwinsemius at comcast.net>, Rainer Krug
>         <rainer_krug at icloud.com>
> Cc: John <miaojpm at gmail.com>, "r-sig-mac at r-project.org"
>         <r-sig-mac at r-project.org>
> Subject: Re: [R-SIG-Mac] Macports installation problem
> Message-ID: <303CD3A3-230A-4D18-A190-E6C2D1160DF3 at llnl.gov>
> Content-Type: text/plain; charset="utf-8"
> 
> The OP's question has inspired an interestingly large number of responses.
> 
> One thing I'm curious about, and I don't think I've seen in the discussion, is why the OP needed to install Cairo graphics with respect to R (otherwise the question was indeed off-topic for R-sig-Mac).
> 
> In summary (and I think this is reasonably consistent with other responses; discussion to support my summary follows):
> 
> 1)  Use CRAN R, install dependencies needed for extended R capability on the OS as needed, from whatever source you prefer, or whichever is easiest for R to use
> 2)  If one chooses to install R using MacPorts or homebrew, be prepared to exercise a greater depth of computer science smarts
> 3)  Feel free to install other useful stuff using MacPorts or homebrew, whichever one prefers; there is unlikely to be any conflict with R
> 
> 
> In close to 20 years of using R on Mac, I have always been able to use whatever capabilities R offers that I wanted to use, using the R provided on CRAN. Nothing has ever motivated me to install R using MacPorts or homebrew in order to obtain some capability that I couldn't have using CRAN's binary download for Mac. This includes things like access to remote Oracle databases, using various spatial packages built around sp and rgdal, reading and writing MS Office files (both Word and Excel), building simple tcl/tk interfaces, writing my own packages that link to Fortran, and probably more that I don't remember. I see no necessity to install R from MacPorts or homebrew. Of course, if one wishes to for whatever reason, great! And I'm glad that people have wanted to make R available from those package managers.
> 
> I do use MacPorts, however, to get useful software such as an X Windows aware version of emacs, ghostscript, and ImageMagick. Indeed, I even have cairo installed using MacPorts, albeit on a 10.11.6 system, not a 10.12.x system like the OP. (It's probably there as a dependency for some other MacPorts package.) From that point of view, the OP's command to install it using MacPorts was correct. I'd need more information to make a guess at why it didn't succeed.
> 
> Some R packages do, of course, depend on external libraries not provided with R. One has to install these from somewhere.  For example, the spatial packages need GDAL. In this case, I use the KyngChaos frameworks, but I'm aware that they're also available from MacPorts (and probably homebrew as well). Could I use them instead? Probably; R is pretty good about letting the user specify where dependencies are found.
> 
> I do not know what R extended capability depends on installing Cairo graphics on the OS. The X11() graphics device, as provided (I believe), has a version that uses cairo. There is the Cairo R package, but it has a binary version, at least for R on 10.11.x, so there's no need to install Cairo graphics to install the Cairo package.
> 
> -Don
> 
> --
> Don MacQueen
> Lawrence Livermore National Laboratory
> 7000 East Ave., L-627
> Livermore, CA 94550
> 925-423-1062
> Lab cell 925-724-7509
> 
> 

David Winsemius
Alameda, CA, USA

'Any technology distinguishable from magic is insufficiently advanced.'   -Gehm's Corollary to Clarke's Third Law



More information about the R-SIG-Mac mailing list