Benchmarking nanoarrow

This subdirectory contains benchmarks and tools to run them. This is currently only implemented for the C library and Python bindings but may expand to include the R bindings as well.

C Library

The structure of the C benchmarks is as follows:

• Benchmarks are documented inline using Doxygen.
• Configurations are CMake build presets, and CMake handles pulling a previous or local nanoarrow using FetchContent. Benchmarks are run using ctest.
• There is a bare-bones report written as a Quarto document that renders the C library results to markdown.

You can run benchmarks for a single configuration (e.g., local) with:

mkdir build && cd build
cmake .. --preset local
cmake --build .
ctest

The provided benchmark-run-all.sh creates (or reuses, if they are already present) build directories in the form build/<preset> for each preset and runs ctest.

You can build a full report by running:

python generate-fixtures.py # requires pyarrow
./benchmark-run-all.sh
cd apidoc && doxygen && cd ..
quarto render benchmark-report.qmd

Python bindings

The Python benchmarks are a standard asv project. You can run the benchmarks with:

# pip install asv
python generate-fixtures.py # requires pyarrow
asv run