diff --git a/python/benchmarks/README.md b/python/benchmarks/README.md index 38431b22bbae2..32b9a3fb83500 100644 --- a/python/benchmarks/README.md +++ b/python/benchmarks/README.md @@ -32,9 +32,13 @@ You can also specify the test class to run: ./python/asv run --python=same --quick -b 'bench_arrow.LongArrowToPandasBenchmark' ``` +If you want to run against your local PySpark changes, make sure to package & install PySpark +in your local environment first. + ### Full run against a commit -Run benchmarks in an isolated virtualenv (builds pyspark from source): +Run benchmarks in an isolated virtualenv. +Note: This builds Spark & PySpark from source and will take a while. ```bash ./python/asv run master^! # Run on latest master commit diff --git a/python/benchmarks/asv.conf.json b/python/benchmarks/asv.conf.json index 28278b6b6b37f..ee91e45da1ae1 100644 --- a/python/benchmarks/asv.conf.json +++ b/python/benchmarks/asv.conf.json @@ -12,10 +12,12 @@ "results_dir": ".asv/results", "html_dir": ".asv/html", "build_command": [ - "python -m pip wheel --no-deps -w {build_cache_dir} {conf_dir}/../packaging/classic" + "{build_dir}/build/sbt -batch -DskipTests clean package", + "python -m pip wheel --no-deps -w {build_cache_dir} {build_dir}/python/packaging/classic" ], + "install_timeout": null, "install_command": [ - "python -m pip install --find-links {build_cache_dir} pyspark" + "in-dir={env_dir} python -m pip install --force-reinstall {wheel_file}" ], "uninstall_command": [ "python -m pip uninstall -y pyspark"