rstudio-img

Customized Docker images for RStudio Server with scientific computing packages, supporting multiple R versions.

Docker Images

Available on:

• Docker Hub: zatzmanm/rstudio
• GitHub Container Registry: ghcr.io/mjz1/rstudio-img

Supported R Versions

• zatzmanm/rstudio:4.3 or ghcr.io/mjz1/rstudio-img:4.3 - Latest R 4.3.X
• zatzmanm/rstudio:4.4 or ghcr.io/mjz1/rstudio-img:4.4 - Latest R 4.4.X
• zatzmanm/rstudio:4.5 or ghcr.io/mjz1/rstudio-img:4.5 - Latest R 4.5.X
• zatzmanm/rstudio:4.6 or ghcr.io/mjz1/rstudio-img:4.6 - Latest R 4.6.X
• zatzmanm/rstudio:latest or ghcr.io/mjz1/rstudio-img:latest - Points to R 4.6 (highest version)

All images ship the latest stable RStudio Server release regardless of R version — the rocker base images pin the RStudio version that was current when each R version was last built, and this repo upgrades it at build time.

Versioned Releases

For each GitHub release (e.g., v1.0.0), the following additional tags are created on both registries:

• v1.0.0 - Release snapshot
• v1.0.0-r4.3 - Release + R 4.3
• v1.0.0-r4.4 - Release + R 4.4
• v1.0.0-r4.5 - Release + R 4.5
• v1.0.0-r4.6 - Release + R 4.6

Prerequisites

• Docker Desktop or Docker Engine
• Docker Compose (for local development)
• 8GB+ RAM recommended
• 10GB+ disk space

Quick Start

Run RStudio Server locally (using either Docker Hub or GHCR):

# From Docker Hub
docker run -d -p 8787:8787 -e PASSWORD=yourpassword zatzmanm/rstudio:latest

# Or from GitHub Container Registry
docker run -d -p 8787:8787 -e PASSWORD=yourpassword ghcr.io/mjz1/rstudio-img:latest

Access at http://localhost:8787

• Username: rstudio
• Password: yourpassword

Note: Images are built for linux/amd64 (AMD64/x86_64). Docker automatically handles emulation on linux/arm64 (ARM64) hosts (e.g., Apple Silicon).

Local Development

Using Docker Compose (Recommended)

1. Clone the repository:

   git clone https://github.com/zatzmanm/rstudio-img.git
   cd rstudio-img

2. Copy the example environment file:

   cp .env.example .env

3. Edit .env to configure your setup:

   # R version (4.3, 4.4, 4.5, or 4.6)
   R_VERSION=4.6
   
   # RStudio credentials
   RSTUDIO_USER=rstudio
   RSTUDIO_PASSWORD=yourpassword

4. Start the container:

   docker-compose up -d

5. Access RStudio at http://localhost:8787

The ./work and ./data directories are automatically mounted for persistent storage.

Using Makefile

For convenience, use the provided Makefile:

# View all available commands
make help

# Start RStudio Server
make up

# View logs
make logs

# Build specific R version
make build-4.6

# Lint code
make lint

# Stop and cleanup
make down

Manual Build

Build for a specific R version:

docker build --build-arg R_VERSION=4.6 -t rstudio-local:4.6 .

Run the built image:

docker run -d -p 8787:8787 -e PASSWORD=rstudio rstudio-local:4.6

By default the build installs the latest stable RStudio Server release. To pin an exact version (or use the preview/daily channel):

docker build --build-arg R_VERSION=4.6 --build-arg RSTUDIO_VERSION=2026.06.0+242 -t rstudio-local:4.6 .

Included Packages

Scientific Computing

• Statistical packages (base R, recommended packages)
• Data manipulation (tidyverse ecosystem ready)
• Machine learning libraries support
• Bioinformatics tools (Bowtie2, etc.)

Development Tools

• Git + Git LFS
• CMake, Make, Automake
• Compilers (GCC, Rust, Cargo)
• Python 3

Document Publishing

• Quarto (latest version)
• TinyTeX (LaTeX distribution), installed to /opt/TinyTeX and on PATH for all users
• Pandoc
• Headless Google Chrome, for HTML→image/PDF rendering via chromote (webshot2, gt::gtsave("*.png"), pagedown). It runs rootless: CHROMOTE_CHROME points at a --no-sandbox wrapper, since Chrome's sandbox cannot initialise under Singularity.

AI Assistants

Both are enabled in /etc/rstudio/rsession.conf and require the user to sign in before anything is sent anywhere.

• GitHub Copilot (copilot-enabled=1)
• Posit Assistant (posit-assistant-enabled=1, allow-posit-assistant=1), requires RStudio Server ≥ 2026.04. On first use RStudio downloads the assistant agent (~3.7 MB) from cdn.posit.co into ~/.posit/assistant, so the machine running the session needs outbound HTTPS to that host.

To disable Posit Assistant for all users of a derived image, set allow-posit-assistant=0 — that is the option which actually gates the feature, not posit-assistant-enabled, which already defaults to 1.

Hardware Acceleration

• OpenCL support
• GPU: one image serves both CPU and GPU. The host driver is exposed by the runtime (singularity exec --nv, docker --gpus); the CUDA toolkit is not in the image — torch downloads a CUDA-enabled backend into the package library and bundles its own cuBLAS/cuDNN.
  • The image does ship system libcudart (both CUDA majors, ~5 MB). R torch's liblantern.so links the plain soname libcudart.so.12 and loads it from ldconfig; PyTorch's bundled copy is hash-mangled and invisible to it, so without this cuda_is_available() is FALSE on a GPU node. This is the minimum that makes R torch GPU work, and it is how rocker's retired CUDA images worked (system runtime via ldconfig; see rocker-org/ml).
  • WITH_CUDA=1 adds the fuller runtime (cuBLAS/cuFFT/… via CUDA_RUNTIME_PACKAGES) for packages that dlopen system versions beyond cudart. Off by default.
  • Not yet covered by a single image: Python TensorFlow/keras (needs a Python GPU env), GPU-accelerated BLAS (NVBLAS), RAPIDS/cuML. GPU roadmap: #14. nvidia-cuda-dev (4.5 GB of headers) was removed in v1.2.0.

GIS and Spatial Analysis

• GDAL, PROJ, GEOS
• Spatial data libraries

Databases

• PostgreSQL, MySQL clients
• SQLite support

And Many More...

See the Dockerfile for the complete list of installed packages.

Architecture Support

Images are built for linux/amd64 (Intel/AMD 64-bit).

Running on ARM64 (Apple Silicon)

Docker automatically uses emulation (Rosetta) to run AMD64 images on ARM64 hosts:

• Performance is excellent for most R/RStudio workloads
• Docker Desktop handles emulation transparently
• The docker-compose.yml configuration automatically specifies the correct platform
• No manual configuration needed for Apple Silicon Macs

Why AMD64 Only?

These images are designed primarily for server deployments where consistency and identical behavior across all environments is critical. Building for a single architecture ensures:

• Identical package versions and behavior everywhere
• No architecture-specific bugs or differences
• Simplified maintenance and testing
• Faster CI/CD build times (~2-3x improvement)

For users who need native ARM64 performance for local development, we may provide optimized ARM64 builds in the future. The current AMD64 images work well via emulation for most use cases.

Versioning Strategy

• R Version Tags: rstudio:4.3, rstudio:4.4, rstudio:4.5, rstudio:4.6 - Always point to the latest build for each R major version
• Latest Tag: rstudio:latest - Always points to the highest R version (currently 4.6)
• Release Tags: For each GitHub release vX.Y.Z, creates versioned snapshots like rstudio:v1.0.0-r4.6
• Uses latest available rocker/rstudio tags for each major R version
• RStudio Server: Always upgraded to the latest stable Posit release at build time (rocker bases pin old RStudio versions for older R versions)
• Release tags preserve build history for reproducible deployments

Troubleshooting

Running on Apple Silicon (M1/M2/M3)

Images are AMD64 and run via emulation on Apple Silicon:

• Docker Desktop handles emulation automatically via Rosetta
• Performance is excellent for R/RStudio workloads (typically <10% overhead)
• Use docker-compose up - platform is pre-configured in docker-compose.yml
• First launch may take slightly longer as emulation initializes
• If you experience issues, ensure Rosetta emulation is enabled in Docker Desktop settings (Preferences → Features in Development → Use Rosetta for x86/amd64 emulation)

Port Already in Use

If port 8787 is already in use:

# Use a different port
docker run -d -p 8888:8787 -e PASSWORD=rstudio zatzmanm/rstudio:latest
# Access at http://localhost:8888

Permission Issues with Mounted Volumes

Ensure the work and data directories exist and have proper permissions:

mkdir -p work data
chmod 755 work data

Container Won't Start

Check logs:

docker-compose logs
# or
docker logs rstudio-dev

Contributing

Development Workflow

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run linters: make lint
5. Build locally: make build
6. Submit a pull request

Code Standards

• Follow Dockerfile best practices
• Use shellcheck for shell scripts
• Test changes with all R versions (4.3, 4.4, 4.5, 4.6)
• Update CHANGELOG.md for notable changes

Adding New R Versions

To add support for a new R version:

1. Update the matrix in .github/workflows/build_push.yaml
2. Update the matrix in .github/workflows/pr-validation.yaml
3. Update README.md documentation
4. Update the "latest" tag logic if it's the new highest version

CI/CD

Automated Builds

Images are automatically built and pushed to Docker Hub and GitHub Container Registry:

• On GitHub releases (all R versions)
• On manual workflow dispatch
• On a monthly schedule, so the rolling tags (4.3–4.6, latest) pick up new RStudio Server, Quarto, and R patch releases automatically

The CI resolves the current stable RStudio Server version and passes it as a build argument, so a new Posit release invalidates the build cache and gets picked up on the next build.

Dependabot keeps the GitHub Actions versions up to date.

PR Validation

Pull requests trigger:

• Dockerfile linting (hadolint)
• Shell script linting (shellcheck)
• Test builds for all R versions (4.3, 4.4, 4.5, 4.6)

License

See repository license file.

Acknowledgments

Built on top of the excellent Rocker Project images.

Support

For issues and questions:

• GitHub Issues: https://github.com/mjz1/rstudio-img/issues
• Docker Hub: https://hub.docker.com/r/zatzmanm/rstudio
• GitHub Container Registry: https://github.com/mjz1/rstudio-img/pkgs/container/rstudio-img