A R Setup, renv, and Docker

R. Heller · Date: 2026-05-09 · Last changed: 2026-11-10

A.1 R installation and current first-use path

Install R from CRAN. The current candidate was locally validated with R 4.6.1. An IDE such as RStudio is optional. Rendering also requires Pandoc; PDF output additionally needs XeLaTeX/TinyTeX and R with Cairo support. Compiled packages may require a C/C++ toolchain and system libraries; the Linux dependencies are listed in .github/workflows/render.yml.

From the repository root, install and check the book’s R dependencies:

RENV_CONFIG_AUTOLOADER_ENABLED=false Rscript --vanilla scripts/install-book-deps.R
RENV_CONFIG_AUTOLOADER_ENABLED=false Rscript --vanilla scripts/install-book-deps.R --check

The installer installs missing packages from current CRAN and checks that their namespaces load. It does not restore historical versions or certify an immutable software environment. --check performs no installation or network access; --list prints the package inventory. Use --library=/path/to/library to select an R library, or --include-optional for packages used only by unevaluated examples and exercises. Optional Python, SQL, OCR and deployment prerequisites must be configured separately.

Internet access is required for installation and uncached provider requests. Configure an activated OpenAlex key locally as OPENALEX_API_KEY, outside Git, for requests that need authenticated paging. SCIENTOMETRICS_OFFLINE=1 requires existing exact-query OpenAlex caches in _freeze/openalex_cache/; it does not make other providers offline and cannot supply missing cache entries. The repository README.md describes the helper functions in R/ and the sample data in data/.

A.2 Package management with renv

renv can record and restore a project’s R package versions. This repository retains a historical renv.lock and automatic activation in .Rprofile, but that lockfile has not been qualified for the current book. The first-use commands above bypass project startup and use the dependency installer. Do not assume that restoring the historical lockfile will reproduce the current validated candidate.

The following is an optional restoration example for a project whose lockfile you have independently validated, rather than a required setup step for this book:

install.packages("renv")
renv::restore()

Restoration attempts to install the recorded versions; success also depends on compatible R, system libraries and available package sources. In a development environment, add a package and deliberately update the lockfile only after validating the resulting environment:

renv::install("newpackage") # Replace with the intended package name.
renv::snapshot()

A software lockfile does not freeze live API results. Exact reproduction also requires the same licensed input snapshots and analysis configuration.

A.3 Core packages

The table below lists the primary packages used throughout the book.

Package Purpose Chapter(s)
openalexR OpenAlex API access 5–37
tidyverse Data wrangling and visualisation All
igraph Network analysis 15–20
tidygraph Tidy network manipulation 15–20
ggraph Network visualisation 15–20
quanteda Text mining and corpus management 21–25
topicmodels LDA topic modelling 22, CS-5
stm Structural topic models 22
gt Publication-quality tables Throughout
targets Pipeline orchestration 34
plotly Interactive charts 35
DT Interactive data tables 35
shiny Web applications 37
word2vec Word embeddings 24
uwot UMAP dimensionality reduction 24

A.4 Docker environment

The repository’s Dockerfile is an optional historical recipe, based on rocker/verse:4.4.1 and renv::restore(). Its restore path has not been qualified for the current book. It also downloads an amd64 Quarto package, so other architectures need a compatible recipe. Docker does not remove the need to validate the lockfile, system dependencies, API access and resulting outputs.

For a deliberate evaluation of that recipe, build it and inspect its R version:

docker build -t scientometrics-in-r .
docker run --rm scientometrics-in-r R --version

The image’s default command calls bookdown::render_book() for all formats; it does not launch RStudio Server. Successful construction and execution of this historical image are not claimed here. Use the current first-use path above for the locally validated candidate.

A.5 Render the book

This repository is configured for bookdown, with _bookdown.yml and _output.yml. From the repository root, render HTML and refresh its download manifest:

SCIENTOMETRICS_REBUILD_CHUNKS=1 RENV_CONFIG_AUTOLOADER_ENABLED=false Rscript --vanilla -e 'bookdown::render_book("index.Rmd", output_format = "bookdown::bs4_book")'
Rscript --vanilla scripts/build-download-manifest.R

SCIENTOMETRICS_REBUILD_CHUNKS=1 re-executes analysis chunks while retaining exact-query provider caches. Missing provider caches still require API access. The configured output directory is docs/. Serve it locally, for example:

python3 -m http.server 8000 --directory docs

Then visit http://localhost:8000. PDF additionally needs XeLaTeX/TinyTeX; output_format = "all" requests HTML, PDF and EPUB. Before distribution, run Rscript --vanilla scripts/repair-epub.R docs/scientometrics-in-r.epub to validate the ebook’s internal links and bundle its citation and licence files, then run Rscript --vanilla scripts/build-download-manifest.R. Local outputs do not update the published site. The README lists the PDF font prerequisites.

Quarto appears in optional examples elsewhere in the book, but quarto render is not the build command for this repository. A Quarto CLI is needed only when working through those separate examples.

This book was built by the bookdown R package.