---

title: 'Installing spatial R packages on Linux'

author: Robin Lovelace

date: '2020-03-30'

slug: ggplot2-inset-maps

categories:

- setup

tags:

- geocompr

- rspatial

- packages

- gdal

- geos

- proj

- linux

draft: false

output: github_document

---

```{r setup, include=FALSE}

knitr::opts_chunk$set(echo = TRUE, message = FALSE, warning = FALSE, eval = TRUE,

fig.width = 8, fig.height = 4)

```

This post explains how to quickly get key R packages for geographic research installed on Linux machines.

<!--  -->

A recent thread on the [r-spatial GitHub organization](https://github.com/r-spatial/discuss/issues/35) alludes to many considerations when choosing a Linux set-up for geocomputational work, ranging from whether to get close to the bleeding edge to the choice of Linux distribution (distro) and whether to use binary or compiled versions of packages and upstream libraries (binaries are faster to install).

Now is an excellent time to be thinking about the topic because changes are in the pipeline and getting set-up (or preparing to get set-up) now could save hours in the future.

These imminent changes include:

- The next major release of R ([4.0.0](https://stat.ethz.ch/R-manual/R-devel/doc/html/NEWS.html)), scheduled for the 24^th^ April ([2020-04-24](https://developer.r-project.org/))

- The next major release of Ubuntu ([20.04](https://www.omgubuntu.co.uk/2019/10/ubuntu-20-04-release-features)), a Long Term Support ([LTS](https://wiki.ubuntu.com/LTS)) release that will be used by millions of servers and research computers worldwide for years to come. Coincidentally, Ubuntu 20.04 will be released a day earlier than R 4.0.0, on 23^rd^ April ([2020-04-23](https://itsfoss.com/ubuntu-20-04-release-features/)).

- Ongoing changes to the OSGeo stack on which key geographic R packages, leading to a huge amount of heroic [r-spatial](https://github.com/r-spatial/) development activity to support the new features, including more precise [coordinate reference system definitions](https://www.r-spatial.org/r/2020/03/17/wkt.html).

This post touches on some of these things but, its main purpose is to **provide advice on getting R's key spatial packages up-and-running on a future-proof Linux operating system** (Ubuntu).^[

I plan to keep this post updated with new developments such as the release of Ubuntu 20.04 (due to be released [2020-04-23](https://itsfoss.com/ubuntu-20-04-release-features/)) and R 4.0.0 (due to be released the day after, on [2020-04-24](https://developer.r-project.org/)).

]

The advice in the next sections comes with a caveat: there is no universally 'right' way of doing things (a benefit of Linux operating systems is that they offer choice and prevent 'lock-in').

There are many other R packages for working with spatial data in R.

However, I think the concise guide below will be useful, especially beginners and people planning to switch to Linux as the basis of their geographic work (see caveats and details in subsequent sections).

By 'key packages' I mean the following, which are the basis of much of [Geocomputation with R](https://geocompr.robinlovelace.net/).

These three packages provide huge power and flexibility, enabling you to do perhaps 80% of day-to-day geographic data processing and visualization tasks:

- [**sf**](https://github.com/r-spatial/sf#installing) for reading, writing and working with a range geographic vector file formats and geometry types

- [**raster**](https://github.com/rspatial/raster/), a mature package for working with geographic raster data (see the [**terra**](https://github.com/rspatial/terra/) for an in-development replacement for **raster**)

- [**tmap**](https://github.com/mtennekes/tmap), a flexible package for making static and interactive maps

The focus is on Ubuntu because that's what I've got most experience with and it is well supported by the community.

Links for installing geographic R packages on other distros are provided in a subsequent.

The pre-requisite for this next section is a computer with a recent Ubuntu or Ubuntu-based (such as Mint) operating system.

You can buy a computer with Ubuntu and other Linux distros installed from [hardware company that supports open-source software](https://itsfoss.com/get-linux-laptops/) such as [System 76](https://system76.com/) or [Entroware](https://www.entroware.com/) (that supplied to computer on which this post was written, shown below).

The lower cost and more environmentally friendly option is to install Ubuntu to give a new lease of life to a pre-existing computer by following this [online guide](https://ubuntu.com/tutorials/tutorial-install-ubuntu-desktop#1-overview).


# 1. Installing spatial R packages on Ubuntu

<!-- Of course, it depends on what Linux distribution you're running, and we'll cover installation on some of the most popular [distros](https://distrowatch.com/). -->

<!-- ## Ubuntu -->

R's spatial packages can be installed from source on the latest version of this popular operating system, once the appropriate repository has been set-up, meaning faster install times (only a few minutes including the installation of upstream dependencies).

The following bash commands should install key geographic R packages on Ubuntu 19.10:

```bash

# bash

sudo add-apt-repository ppa:marutter/rrutter3.5

sudo apt update

sudo apt install libudunits2-dev libgdal-dev libgeos-dev libproj-dev

sudo apt install r-base-dev r-cran-sf r-cran-raster r-cran-rjava

```

If you are using an older version of Ubuntu and don't want to upgrade to 19.10, which will upgrade to (20.04) by the end of April 2020, see instructions at [github.com/r-spatial/sf](https://github.com/r-spatial/sf#installing).

To test your installation of R has worked, try running R in an IDE such as RStudio or in the terminal by entering `R`.

You should be able to run the following commands without problem:

```r

# R

library(sf)

#> Linking to GEOS 3.7.2, GDAL 2.4.2, PROJ 5.2.0

install.packages("tmap")

```

<!-- ## Fedora -->

<!-- The following command should install all the dependencies required: -->

<!-- ``` -->

<!-- sudo dnf install gdal-devel proj-devel proj-epsg proj-nad geos-devel udunits2-devel -->

<!-- ``` -->

<!-- ## Arch -->

<!-- ## Other Linux distros -->

<!-- # 3. Which operating system to use? -->

# 2. Why Linux?

Linux operating systems are free and open-source, like R itself, and provide a number of [advantages](https://itsfoss.com/linux-better-than-windows/) over Windows, including:

- Continuous upgrades: while Microsoft pushes [updates](https://www.catalog.update.microsoft.com/Search.aspx?q=windows%20security%20update%202020) every month or so, updates on Linux distributions tend to be more regular, ensuring you can access more up-to-date, secure, stable and performant software sooner

- Ideal for learning to program: while Windows (and to some extent Mac) tries to reduce the need for computing skills such, Linux encourages you to learn to program via access to a powerful command-line interface (which will use below)

- Freedom: Most Linux distributions are free to download, modify and pass-on, so you can install up-to-date operating systems on all your computers without worrying about licensing issues

Like R, you cannot learn to use Linux overnight but distributions designed with user-friendliness in mind, such as Ubuntu and Mint, mean that Linux is now something you can recommend to your [grandmother](https://www.computerworld.com/article/2468094/linux-for-grandma---grandpa.html).

# 3. Installing geographic R packages on other Linux operating systems

If you are in the fortunate position of switching to Linux and being able to choose the distribution that best fits your needs, it's worth thinking about which distribution will be both user-friendly (more on that soon), performant and future-proof.

Ubuntu is a solid choice, with a large user community and repositories such as 'ubuntugis' providing more up-to-date versions of upstream geographic libraries such as GDAL.

QGIS is also well-supported on Ubuntu.

However, you can install R and key geographic packages on other operating systems, although it may take longer.

Useful links on installing R and geographic libraries are provided below for reference:

- Installing R on **Debian** is covered on the [CRAN website](https://cran.r-project.org/bin/linux/debian/). Upstream dependencies such as GDAL can be installed on recent versions of Debian, such as [buster](https://www.debian.org/releases/), with commands such as `apt-get install libgdal-dev` as per instructions on the [rocker/geospatial](https://github.com/rocker-org/geospatial/blob/eaf5e92f90737ce9771753cab48f3a2f1d779216/Dockerfile).

- Installing R on **Fedora/Red Hat** is straightforward, as outlined on [CRAN](https://cran.r-project.org/bin/linux/redhat/README). GDAL and other spatial libraries can be installed from Fedora's `dnf` package manager, e.g. as documented [here](https://github.com/r-spatial/sf#fedora) for **sf**.

- **Arch Linux** has a growing R community. Information on installing and setting-up R can be found on the [ArchLinux wiki](https://wiki.archlinux.org/index.php/R). Installing upstream dependencies such as [GDAL on Arch](https://www.archlinux.org/packages/community/x86_64/gdal/) is also relatively straightforward. There is also a detailed guide for installing R plus geographic packages by [Patrick Schratz](https://pat-s.me/post/arch-install-guide-for-r/).

# 4. How close to the bleeding edge should you get?

One of the greatest things about the freedom provided by using Linux is also one of the scariest:

you can go as deep into the operating system and as close to the 'cutting edge' of computing as you want, with few restrictions.

This can be exciting but it can also be wasteful if you end up spending hours fiddling with your set-up.

It's worth considering the stability-bleeding continuum before diving into a particular set-up if the previous section hasn't already made-up your mind.

Even if you have decided on a particular OS there are many options that can take you closer to the bleeding edge that you may or may not want to take.

For me, that means the latest version of Ubuntu plus CRAN versions of most R packages.

For most people I recommend installing the release version as follows:

```{r, eval=FALSE}

# R

install.packages("tmap")

```

It is good practice to keep your packages up-to-date.

To do so you run the following command regularly:

```{r, eval=FALSE}

# R

update.packages()

```

If you want to test development versions of various packages, you can always do so from inside R with commands such as the following (which installs the development versions of **tmap** and **tmaptools**):

```{r, eval=FALSE}

# R

remotes::install_github("mtennekes/tmap")

remotes::install_github("mtennekes/tmaptools")

```

You can get more recent versions on OSGeo packages on Ubuntu by adding the `ubuntugis` repository as follows:

```bash

# bash

sudo add-apt-repository ppa:ubuntugis/ubuntugis-unstable

sudo apt-get update

```

# 5. Geographic R packages on Docker

As with cars, ease of use is important for the popularity of computer technology.^[

The history of cars can provide insight into the importance of ease of use of technologies today.

Cars, have arguably transformed our settlements and lifestyles more than any other technology, were initially hard to use.

Before they became a consumer product in the 1950s (by the end of which 1/6^th^ of jobs in the USA were in the [car industry](https://en.wikipedia.org/wiki/1950s_American_automobile_culture)) relied on a [hand cranks](https://www.youtube.com/watch?v=iFd8uo7ogpM) to start them until the proliferation of electric starter motors following U.S. Patent [1,150,523](https://patents.google.com/patent/US1150523), which was subsequently used by Cadillac in [1912](https://www.hemmings.com/blog/2012/02/27/the-accident-that-started-it-all/) and onwards.

Like cars, people tend to go for computer technologies that are easy to use, that are 'plug and play', so it's important for the future of open-source software that the solutions I recommend are easy to set-up and use.

]

The Ubuntu installation instructions outlined above provide such an easy and future-proof set-up.

But if you want an even easier way to get the power of key geographic packages running on Linux, and have plenty of RAM and HD space, running R on the '[Docker Engine](https://docs.docker.com/install/)' may be an attractive option.

Advantages of using Docker include **reproducibility** (code will always run the same on any given image, and images can be saved), **portability** (Docker can run on Linux, Windows and Mac) and **scalability** (Docker provides a platform for scaling-up computations across multiple nodes).

For an introduction to using R/RStudio in Docker, see the [Rocker project](https://www.rocker-project.org/).

Using that approach, I recommend the following Docker images for using R as a basis for geographic research:

- [`rocker/geospatial`](https://hub.docker.com/r/rocker/geospatial) which contains key geographic packages, including those listed above

- [`robinlovelace/geocompr`](https://hub.docker.com/r/robinlovelace/geocompr/) which contains all the packages needed to reproduce the contents of the [book](https://geocompr.robinlovelace.net/), and which you can run with the following command in a shell in which Docker is installed:

```bash

docker run -e PASSWORD=yourpassword --rm -p 8787:8787 robinlovelace/geocompr

```

To test-out the Ubuntu 19.10 set-up recommended above I created a [Dockerfile](https://github.com/Robinlovelace/geocompr/blob/master/docker/ubuntu-eoan/Dockerfile) and associated image on Dockerhub that you can test-out as follows:

```bash

# bash

docker run -it robinlovelace/geocompr:ubuntu-eoan

```

```{r, eval=FALSE}

# R

library(sf)

#> Linking to GEOS 3.7.2, GDAL 2.4.2, PROJ 5.2.0

library(raster)

library(tmap)

```

The previous commands should take you to a terminal inside the docker container where you try out the Linux command line and R.

# 6. Fin

In summary, R is an open-source language heavily inspired by Unix/Linux so it should come as no surprise that it runs well on a variety of Linux distributions, Ubuntu (covered in this post) in particular.

Building on [OSGeo](https://www.osgeo.org/) libraries, key geographic R packages are also easy to set-up, run and develop on Linux.

I hope that this tutorial provides some useful pointers and encourages more people to switch from proprietary software to open source solutions as the basis of their geographic and computational work.


Be the [FOSS4G](https://wiki.osgeo.org/wiki/FOSS4G) change you want to see in the world!