Someone reached out with concern about speed of installing an #rstats package with its dependencies, all from source.
I _strongly_ recommend installing R package binaries whenever possible, and it's easy to do so on Mac, Windows or Ubuntu. Some details:
1
15
1
99
{nullranges} on Bioc has 97 packages it depends on. Many of these are likely already installed if you've got tidyverse + Bioc core, but even on a clean Mac OS X 13 (so only w/ R + recommended), installing them all takes about a minute with binaries 馃帀
youtube.com/7sHiRShfUmw
1
3
4
What about Ubuntu? Again, don't install from source if you don't have to. Use the CRAN (+ Bioc) binaries provided by @eddelbuettel's r2u
E.g.:
apt install r-cran-data.table r-bioc-deseq2
eddelbuettel.github.io/r2u/
Not all Bioc pkgs are provided, but many are, especially core ones
2
2
9
Installing all the dependencies needed for {nullranges} on Ubuntu 22.04 with a fresh R installation (only r-recommended) takes a little over a minute.
Note that this is providing binaries for 3.16 (the latest Bioc release from October)
youtube.com/zevcuAgq_6w
1
1
1
Finally installing {nullranges} with BiocManager::install() only takes another minute
We've updated the nullranges README to give pointers on how to take advantage of binaries on Mac/Windows/Ubuntu.
Feel free to ask Qs here / GitHub / support / Slack
youtube.com/6WFlqrGUOLk
1
Finally, I know that people working on HPC don't have access to `apt`. Every institute I've worked at has had great IT folks that keep R and major/popular R and Bioc pkgs up-to-date. It's a twice a year job for someone, so I think it's totally fair to ask for support for this.
1
5
One weird trick... again from @eddelbuettel:
For Linux: you can avoid `apt` altogether if you follow this step from r2u. Then you can get the R package binaries from install.packages()
For me, I like to see what's there via apt-cache, so i prefer apt, but to each their own...
2
2
3
Out of curiosity, when I use rig to manage different #RStats versions, will this still work?
1
1
1
Orthogonal: #r2u offers a complete .libPaths() for the default (current) #Rstats, with full dependency resolution, as binaries, fast. That, Just. Works.
You can combine it with different R interpreters if you set their .libPaths(). They may complain about newer package builds.
Dec 4, 2022 路 1:56 PM UTC
1
2