How to update all packages for a new R version quickly and easily?

Question

Suppose I have R version 3.x.x installed, and I upgrade to version 4.x.x, is there any quick/easy way to install all the new versions of the libraries I had installed?

Please assume all the packages are on CRAN

Chronometry

Does this answer your question? Painless way to install a new version of R? – Fillip

1) The call to update.packages() will always do that for you. 2) In the special case of an R major version change, you can set option checkBuilt=TRUE which triggers a reinstallation even if the package did not change at CRAN but because you moved R versions. – Rounders

@DirkEddelbuettel thanks Dirk. Should I provide a vector of package names to update.packages() or does it figure that out? – Chronometry

It takes the existing ones as given, see help(update.packages()). – Rounders

Answer 1

Don't know if this is quick and easy, but I think the pacman package can be useful.

1. Under the previous version, use pacman::p_lib() to return a vector of your installed packages, and save them onto disk use saveRDS().

For instance,

mypks <- pacman::p_lib()
saveRDS(mypks, "~/mypks.rds")

2. Update R.

3. Import the vector from step 1 using readRDS() and run install.packages() with the object.

For instance,

mypks <- readRDS("~/mypks.rds")
install.packages(mypks)

Cocklebur

Thanks, this was quick and painless. I ended up using a lazier version of your approach, which I noted below for future reference – Chronometry

Answer 2

Following works for me:

update.packages(ask = FALSE, checkBuilt = TRUE)

Ashkhabad

This is the correct answer: this needs nothing but base R functionality. (As my earlier comment from 2021 already mentioned.) – Rounders

Answer 3

Step 1

Run this in the previous R installation:

# install.packages("pacman")
library(pacman)
dput(pacman::p_lib())

Copy the output to clipboard.

Step 2

Open your new R version, paste the output from the previous step in place of ***paste output here***:

vector_of_packages <- ***paste output here***
install.packages(vector_of_packages)

Notes:

• This is just a lazier way of doing what @Phil suggests.

Chronometry

Answer 4

The latest versions of RStudio have an Update option next to the list of packages installed under the Packages tab.

Burro

That answer is somewhere between 'irrelevant' (as R itself always had update.packages() as a function) and 'misleading'. – Rounders

Answer 5

Luckily,

update.packages(ask = FALSE, checkBuilt = TRUE)

works if you just mv or

cd oldlibdir
tar -cf . -| (cd newdir; tar -xf -) 

from the shell and then run it from the R console in the new version (usually). The problem with having to save a vector of installed R packages from within your previous version to feed to update.packages is that automatic software package upgrades (pkg, apr, rpm) will upgrade R before you get a chance 😄

I always pkg lock R so it only gets upgraded when I explicitly invoke an R software upgrade.

Trillion

Hello, I recently upgraded R 4.3.3 on debian stable via backports. Now, if for instance I open an R session and I do library(ggplot2), I do not get a warning that the library was built under the previous R version (actually I do not get any such warning for any library). Two questions then: 1) do I still need to reinstall all the user-installed packages? 2) any way to do that with the pak library instead of base R and check.Built=T? I find pak nicer and more performant. Many thanks! – Imagism

R packages will often continue to work as long as the R version change isn't breaking ABI compatibility with installed libraries and they are still able to be seen and loaded from a standard library directory location. In your case it sounds like the system upgrade didn't break anything. – Trillion