- Automatic detection and listing of performance regressions. (#236)
- Support for Windows. (#282)
- New
setup_cachemethod. (#277) - Exclude/include rules in configuration matrix. (#329)
- Command-line option for selecting environments. (#352)
- Possibility to include packages via pip in conda environments. (#373)
- The
pretty_nameattribute can be used to change the display name of benchmarks. (#425) - Git submodules are supported. (#426)
- The time when benchmarks were run is tracked. (#428)
- Mirrors are no longer created for local repositories. (#314)
- The parent directory of the benchmark suite is no longer inserted
into
sys.path. (#307) - In asv.conf.json matrix,
nullpreviously meant (undocumented) the latest version. Now it means that the package is to not be installed. (#329) - Previously, the
setupandteardownmethods were run only once even when the benchmark method was run multiple times, for example due torepeat > 1being present in timing benchmarks. This is now changed so that also they are run multiple times. (#316) - The default branch for Mercurial is now
default, nottip(#394). - Benchmark results are now by default ordered by commit, not by date. (#429)
- When
asv runand other commands are called without specifying revisions, the default values are taken from the branches inasv.conf.json. (#430)
- Output will display on non-Unicode consoles. (#313, #318, #336)
- Longer default install timeout. (#342)
- Various other bugfixes and minor improvements.
First full release.
Include pip_requirements.txt.
Display version correctly in docs.