---
title: "z - Advanced topic: Understanding the rPackages set release cycle and using bleeding edge packages"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{z-advanced-topic-understanding-the-rpackages-set-release-cycle-and-using-bleeding-edge-packages}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
```

```{r, include=FALSE}
library(rix)
```

## Introduction

It is important to understand the release cycle of the rPackages set and what
steps you should take if you need bleeding edge packages. R packages on
`nixpkgs` tend to get updated alongside a new release of R, and the reason is to
ensure a certain level of quality. The vast majority of CRAN (and BioConductor)
packages are made available through `nixpkgs` in a fully automated way. But some
packages do require some manual intervention to work on Nix, and we only know
this if we try to build these packages, but building packages requires quite a
lot of resources. We can’t build CRAN packages every single day to see if
everything works well on Nix, so we only rebuild the whole tree whenever there’s
a new release of R. Packages get built on a CI infrastructure called *Hydra*,
and then these packages get cached on
[cache.nixos.org](https://cache.nixos.org/) so whenever someone wants to install
a package, a pre-built binary gets download from the cache. This avoids having
to build software from source locally. For packages that don’t need compiling
this is not that big of a time save, but for packages that do need to get
compiled it is huge. Depending on which packages you want to install, if you had
to build everything from source, it could potentially take hours, but if you can
install pre-built binaries it’s just a matter of how quick your Internet
connection is.

## R packages available through Nix

As explained in the introduction, the *rPackages* set on `nixpkgs` gets updated
shortly after a new release of R. The process involves first updating the
package definitions found
[here](https://github.com/NixOS/nixpkgs/tree/nixos-unstable/pkgs/development/r-modules),
and then building the whole tree on a CI platform called *Hydra*. Build failures
then get fixed by volunteers (to learn how you can contribute, read the
`vignette("z-contributing_to_nixpkgs")`). After the most important packages have
been fixed, the whole rPackages set gets updated and made available through
`nixpkgs` master branch.

Essentially this means that if you start a project with `{rix}` using `"latest"`
as the `r_ver` just after the rPackages set got updated, this project will use
very fresh packages. But if instead you start a project just before an R
release, then the environment will be using older packages. In practice this
rarely matters, unless you absolutely need a very recent version of a specific
package because you need a specific feature, or if you need an environment with
bleeding edge packages for development. For cases like this, we provide the
`r_ver = "bleeding_edge"` and `r_ver = "frozen_edge"` options that makes it
possible to use the most recent packages for your environment, but at a cost.
You must be aware of this cost which we detail in the next section.

## Using bleeding and frozen edge package for your environments

CRAN is continuously getting new or updated packages. When you use R outside of
Nix, running `install.packages(pkg)` will install the latest version of the
`{pkg}` package available from CRAN (unless you changed the default repository
url). With Nix, packages do not get downloaded from CRAN but for the `nixpkgs`
repository and they may be outdated as explained above. If you require bleeding
edge packages, use the `"bleeding_edge"` option to `r_ver`. This will **NOT**
download packages from the official `nixpkgs` repository anymore, but from a
fork that we maintain that you can find
[here](https://github.com/rstats-on-nix/nixpkgs/tree/r-daily). This fork gets
updated every 12 hours with the latest commits from both the `nixpkgs`
repository and CRAN. This means that environments generated using this fork will
contain bleeding packages for both the CRAN (and BioConductor) packages as well
as system-level dependencies. Also, be aware that this will generate an
expression that is **NOT** reproducible! Each time you will rebuild the
environment generated using the `"bleeding_edge"` option, you will get exactly
that: an environment with bleeding edge packages at build time. If you want
instead to have bleeding edge packages at the time you generate the expression,
but then want reproducibility (in other words, building should always result in
the same environment), then use the `"frozen_edge"` option instead.

In any case, whatever option you choose, using them comes at a cost of which you
must be aware.

First, because these packages are bleeding edge, they have not had the chance to
be built by *Hydra* yet. *Hydra* periodically builds packages and these get
cached. So if you’re using Nix, pre-compiled binaries get used instead of being
built from source. This is not the case if you use our fork, **unless** you also
use the cache that we provide, courtesy of
[cachix.org](https://app.cachix.org/cache/rstats-on-nix#pull). Cachix provides a
cache for your own projects and works seamlessly with the official Nix cache.
However, we are limited in space, and cannot possible cache all the CRAN
packages. So only the most popular packages get built and cached, and hopefully
the packages you need are part of this limited set. To use the cache, run the
following commands on your computer. First, install the Cachix client:

```
nix-env -iA cachix -f https://cachix.org/api/v1/install
```

then use the cache:

```
cachix use rstats-on-nix
```

that’s it! Packages now not only will get pulled from the official cache, but
also from the dedicated `rstats-on-nix` cache. Our cache also contains the
latest version of R, which sometimes can also lag behind on the official
`nixpkgs` repository.

## Making your own cache

If you need packages that are not included in the `rstats-on-nix` cache, you can
very easily roll out your own binary cache. Simply build the environment once on
one machine, and then push the binaries in a cache for re-use. Refer to the
vignette `vignette("z-binary_cache")` to learn how to set up your own cache.