PERF: EWMGroupby by mroeschke · Pull Request #39664 · pandas-dev/pandas (original) (raw)

Here are the new benchmarks I added for other methods

       before           after         ratio
     [e889b941]       [1f1d2ef9]
     <ref/ewm_groupby~1^2>       <ref/ewm_groupby>
-      22.6±0.7ms       2.29±0.2ms     0.10  rolling.GroupbyEWM.time_groupby_method('var')
-        22.2±2ms       2.21±0.3ms     0.10  rolling.GroupbyEWMEngine.time_groupby_mean('cython')
-      30.9±0.1ms       2.42±0.3ms     0.08  rolling.GroupbyEWM.time_groupby_method('std')
-       234±0.9ms       4.42±0.2ms     0.02  rolling.GroupbyEWM.time_groupby_method('cov')
-        251±20ms       4.54±0.7ms     0.02  rolling.GroupbyEWM.time_groupby_method('corr')

SOME BENCHMARKS HAVE CHANGED SIGNIFICANTLY.
PERFORMANCE INCREASED.

Looks like the CI / Checks is failing for an unrelated reason

Run ci/code_checks.sh docstrings
Validate docstrings (GL03, GL04, GL05, GL06, GL07, GL09, GL10, SS01, SS02, SS04, SS05, PR03, PR04, PR05, PR10, EX04, RT01, RT04, RT05, SA02, SA03)
Traceback (most recent call last):
  File "ci/../scripts/validate_docstrings.py", line 50, in <module>
    from numpydoc.validate import validate, Docstring  # isort:skip
ImportError: cannot import name 'Docstring' from 'numpydoc.validate' (/usr/share/miniconda/envs/pandas-dev/lib/python3.8/site-packages/numpydoc/validate.py)