Installing DSS with R, Mac OS X 10.15, Homebrew? Or other way?

[tgb417]

I'm working on a fresh install of DSS on a computer with Mac OS X 10.15 Catalina.

As many folks know Mac OS does not really have a package manager. 😖

I had a less than wonderful experience with Anaconda Navigator on Mac OS V10.13 (High Sierra) particularly when it came to R, but Python was also causing me some challenges in that configuration as well. (And the older Mac OS was not helping either.)

So I thought I'd try Home Brew both formulas and Casks on Mac OS X 10.15.

I was able to install DSS 8.0.4 via a HomeBrew cask.  (Sort of Cool)

I'm installing other applications via Homebrew casks as well.  For example, Zoom.us which updates almost every week is being maintained by Homebrew.  Cool. 

However, when I've gotten to R, I've hit a roadblock.  The DSS documentation here says:

DSS requires R version 3.4 to 3.6. R version 4.x is not supported.

Hmmm.  The current version of R supported by HomeBrew is 4.0.3.  And there does not seem to be a straightforward way to install an older version via a cask or formula?

This leaves me with a number of questions:

• What package manager are data scientists using on their Macintosh computers?  (I know real data scientists use Linux. However, I see a lot of Macintosh computers in the community. So, I'm hoping folks have figured this out.)
• When will Dataiku Support R Version 4.x?
• How are folks successfully installing R on Macintosh for use with DSS?  
• Are you able to use a package manager to install R and keep it up to date?
• I see other notes about potential problems with HomeBrew installs of R.  (What am I getting myself into if I use HomeBrew to install DSS and R?)
• Should I do an "Advanced Mac OS" installation?  What are the advantages/disadvantages?
• There appears to be a DSS shell script way to assessing or installing R.  How do you find this on a Mac?  Is this only available in the "Advanced Max OS" install? 

  install-deps.sh -check -without-java -without-python -with-r

• Why does the Installation of R with DSS have to be complicated?

Looking for folks' thoughts on installing DSS on Macintosh, and getting the R features working in as reproducible, and self-maintaining, a method as possible. 

P.S. I have not installed anaconda yet onto the macintosh either.  Again thoughts about best practices on a Macintosh?

[sergeyd]

Hi @tgb417 

For OSX the easiest way to install R is from the pkg file located here: https://cran.r-project.org/bin/macosx/R-3.6.3.nn.pkg. It should install the R app, the terminal access and any dependencies. 

As R app doesn’t make R available globally and available only within R app, the only thing you will need to add is the path to the R binary to the env-site.sh file via PATH variable :

export PATH=/Library/Frameworks/R.framework/Versions/3.6/Resources/bin:$PATH

Or make a softlink for /usr/local/bin/R -> /Library/Frameworks/R.framework/Versions/3.6/Resources/bin/R so it will be available globally. 

[tgb417]

@sergeyd ,

Thanks for the quick reply.

So, I'm hearing the suggestion for a hand install from a pkg file on CRAN, and a manual path update. Thus removing the advantages of a package manager from a maintenance point of view.  (Given that we need to use the older version maybe that makes some sense.)

This still leaves me with a bunch of other questions above for you and other community members.  

What about the other dependencies that some seem to think it takes to run R successfully and install packages.  

• gcc
• gettext
• jpeg
• libpng
• openblas
• pcre2
• deadline
• tcl-tk
• xz

Are folks using a package manager like HomeBrew (or something else please explain) to keep those components up to date.  

If the manual approach is used for DSS install of the older R. What about the installation of things like R-Studio.  The DSS R Studio Integration.   Compilation of more complicated R packages like the Stan based Facebook Profit?

All, Thoughts?