.. -- rst --

===================== PyWavelets benchmarks

Benchmarking PyWavelets with Airspeed Velocity.

Usage

Airspeed Velocity manages building and Python virtualenvs (or conda environments) by itself, unless told otherwise. To run the benchmarks, you do not need to install a development version of PyWavelets to your current Python environment.

First navigate to the benchmarks subfolder of the repository.

cd benchmarks

To run all benchmarks against the current build of PyWavelets::

asv run --python=same --quick

The following notation (tag followed by ^!) can be used to run only on a specific tag or commit. (In this case, a python version for the virtualenv must be provided)

asv run --python=3.5 --quick v0.4.0^!

To record the results use:

asv publish

And to see the results via a web browser, run:

asv preview

More on how to use asv can be found in ASV documentation_ Command-line help is available as usual via asv --help and asv run --help.

.. _ASV documentation: https://asv.readthedocs.io/

Writing benchmarks

See ASV documentation_ for basics on how to write benchmarks.

Some things to consider:

• The benchmark files need to be importable when benchmarking old versions of PyWavelets. So if anything from PyWavelets is imported at the top level, it should be done as:

  try: from pywt import cwt except ImportError: pass

The benchmarks themselves don't need any guarding against missing features --- only the top-level imports.

To allow tests of newer functions to be marked as "n/a" (not available) rather than "failed" for older versions, the setup method itself can raise a NotImplemented error. See the following example for the CWT:

  try:
      from pywt import cwt
  except ImportError:
      raise NotImplementedError("cwt not available")

• Try to keep the runtime of the benchmark reasonable.

• Use ASV's time_ methods for benchmarking times rather than cooking up time measurements via time.clock, even if it requires some juggling when writing the benchmark.

• Preparing arrays etc. should generally be put in the setup method rather than the time_ methods, to avoid counting preparation time together with the time of the benchmarked operation.