tiledbsoma (R)

R implementation of the SOMA API specification.

Installation

From R-universe (recommended)

R-universe serves macOS binaries and source packages for other Unix-like platforms.

install.packages("tiledbsoma", repos = c("https://tiledb-inc.r-universe.dev",
                                         "https://cloud.r-project.org"))

Installing from source on Unix-like platforms requires cmake, git, and a C++ compiler with C++20 support (e.g., g++ 13+).

From Conda

Binary packages for multiple architectures are available from Conda:

conda install -c conda-forge -c tiledb r-tiledbsoma

From GitHub (development)

To install the latest development version from the main branch:

remotes::install_github("https://github.com/single-cell-data/TileDB-SOMA", subdir = "apis/r")

Requirements

Source installation requires:

• tiledb R package, which depends on a local installation of TileDB Core or its provided build artifacts
• cmake and git
• libtiledbsoma, which is built automatically during package installation or can be provided as a system library (see the libtiledbsoma directory for manual installation)

All other R package dependencies are listed in DESCRIPTION and can be installed with:

remotes::install_deps(".", dependencies = TRUE)

Troubleshooting: fmt and spdlog conflicts

If you have system versions of fmt or spdlog installed, they can conflict with the versions vendored by TileDB.

• undefined symbol: _ZTIN3fmt2v912format_errorE: Uninstall your system versions of fmt and spdlog. On Linux, verify removal with dpkg -l | egrep "lib(spdlog|fmt)".
• fatal error: spdlog/spdlog.h: No such file or directory: Remove the leftover cmake config directory at /usr/local/lib/cmake/spdlog, which the system uninstall of spdlog fails to clean up.

Development

Setup

1. Clone the repository:

   git clone https://github.com/single-cell-data/TileDB-SOMA.git
   cd TileDB-SOMA/apis/r
   

2. Install R package dependencies:

   remotes::install_deps(".", dependencies = TRUE)
   

3. Optionally, install additional data packages used by tests:

   • pbmc3k provides single-cell data for SingleCellExperiment and SummarizedExperiment ingestion tests:

     install.packages("pbmc3k", repos = "https://tiledb-inc.r-universe.dev")
     

   • pbmc3k.tiledb provides pre-built SOMA test data:

     install.packages("pbmc3k.tiledb", repos = "https://ghrr.github.io/drat")
     

Build and install

# Clean previous build artifacts (not needed on first build)
./cleanup

# Build the source tarball (includes libtiledbsoma via symlink)
R CMD build .

# Check the package (skipping vignettes and manual, which require texlive)
R CMD check --no-vignettes --no-manual tiledbsoma_*.tar.gz

# Install
R CMD INSTALL tiledbsoma_*.tar.gz

During installation, TileDB Core is downloaded if needed and libtiledbsoma is built if a system installation is not found.

After editing source files

• Changed src/*.cpp function signatures: Run Rscript -e 'Rcpp::compileAttributes()' to update Rcpp-generated glue code.
• Changed C++ function header documentation: Run Rscript -e 'roxygen2::roxygenise()' to update the corresponding R files.

Once installed, the package sources can be edited and re-installed iteratively. Running ./cleanup before rebuilding ensures a full rebuild including libtiledbsoma.