Using INLA on Windows

INLA is an R package for Bayesian inference for Latent Gaussian Models. There are a few challenges in getting started:-

  • It is not on CRAN, because the building process is non-standard…
  • …because at present building from source on windows does not work,
  • INLA do provide working binaries, but no retro-building. When there is a new INLA release, one binary is made for the current latest minor version of R (4.4 at present). No binaries of new INLA for old R, or old INLA for new R.
  • If you follow the install instructions on an older version of R such as 4.3, R knows there is a newer INLA source version and tries to install it from source. Which does not go well.

For reference, the error you get attempting to install INLA from source (whether you intended to install from source or not), looks something like this, which I got for R 4.3 this morning:-

> install.packages("INLA",
    repos = c(getOption("repos"),
      INLA = "https://inla.r-inla-download.org/R/stable"), 
    dep = TRUE) 

There is a binary version available but the source version is later:
         binary   source needs_compilation
INLA 24.05.01-1 24.05.10             FALSE

installing the source package ‘INLA’

trying URL 'https://inla.r-inla-download.org/R/stable/src/contrib/INLA_24.05.10.tar.gz'
Content type 'application/x-gzip' length 61448956 bytes (58.6 MB)
downloaded 58.6 MB

* installing *binary* package 'INLA' ...
C:\WINDOWS\cp.exe: invalid option -- )
Try `C:\WINDOWS\cp.exe --help' for more information.
ERROR: installing binary package failed
* removing 'E:/Rlib/INLA'
Warning in install.packages :
  installation of package ‘INLA’ had non-zero exit status

The latest binary release turns out to be 24.05.01-1 - but it knows there is a more recent source available and tries to build it. The easiest fix is to force type = "binary" in install.packages, and settle for the 24.05.01-1 binary.

See this issue for the discussion, leading to the options below. Begin by asking yourself three questions:-

  1. Do I want to install locally, or provision with Hipercow
  2. Do I want to use the latest version of INLA, or an older one.
  3. And you do know what your R version is? Should you upgrade?

Local installations

Latest available version of INLA for your R version

install.packages("INLA", 
  repos = c(getOption("repos"), 
          INLA = "https://inla.r-inla-download.org/R/stable"), 
      dep = TRUE, type = "binary")

This works. Right now, if you do this from within R 4.4, you will get version 24.05.10, whereas for R 4.3, you will get 24.05.01-1. The exact version you get will be the most recent version you find in here for R 4.4 or here for R 4.3, or other folders for even older R.

If you are not on the latest version of R, you likely will not be able to get the latest version of INLA, because the INLA team only make new releases for the latest R version. Therefore, if you really need the latest INLA, you must be on the latest version of R.

Specific version of INLA, where available.

So install.packages does not allow us to choose a version, and remotes::install_version only works when type="source", which you may recall me mentioning once or twice, cannot succeed for reasons.

If you want a specific version of INLA, you are limited to the binaries the INLA team have made for your R version. So look in here for R 4.4 or here for R 4.3 for example, and find your desired version. For R 4.3, INLA_23.09.09.zip looks nice.

Download the zip, and use R CMD INSTALL INLA_23.09.09.zip - or to do this all in R, you could try:-

tmpdir <- tempdir()
version <- "23.09.09"
file <- sprintf("INLA_%s.zip", version)
url <- sprintf("https://inla.r-inla-download.org/R/stable/bin/windows/contrib/4.3/%s", file)
tmpfile <- path.file(tmpdir, file)
curl::curl_download(url, tmpfile)
install.packages(tmpfile)
unlink(tmpfile)

Hipercow provisioning

Latest version of INLA for latest R version.

The current latest R version is 4.4. If you are using that version, then you can either use the script or the pkgdepends method.

For pkgdepends, in your hipercow root you can write in pkgdepends.txt :-

repo::https://inla.r-inla-download.org/R/stable
INLA

For the script method, you instead write provision.R and paste this code to install the package, which is the same as the local install:-

install.packages("INLA", 
  repos = c(getOption("repos"), 
          INLA = "https://inla.r-inla-download.org/R/stable"), 
      dep = TRUE, type = "binary")

Then hipercow_provision() will work in the usual way. See vignette("packages") for more details.

Latest version of INLA for older R version

Here, the script method above will work as before, but if you prefer pkgdepends, you will have to be specific about the version you want to install, and it must be the most recent version for your R version. So for R 4.3, your pkgdepends.txt would have to say:-

repo::https://inla.r-inla-download.org/R/stable
[email protected]

because that is the latest version we see here.

Specific version of INLA, where available.

See the instructions above for a local installation, put the code in provision.R to use the script method with hipercow_provision().