speeding up check_windspeed_drift (test mean 980us -> 482us)

[gabrielecalvo]

Speeding up execution by avoiding copy and keeping only needed data around.

The time has been cut to half from:

------------------------------------------------------------- benchmark: 1 tests ------------------------------------------------------------
Name (time in us)                         Min         Max      Mean    StdDev    Median       IQR  Outliers  OPS (Kops/s)  Rounds  Iterations
---------------------------------------------------------------------------------------------------------------------------------------------
test_calc_rolling_windspeed_diff     830.3000  2,590.3000  980.2053  230.7853  900.2000  133.6000     23;23        1.0202     206           1
---------------------------------------------------------------------------------------------------------------------------------------------

to

------------------------------------------------------------ benchmark: 1 tests ------------------------------------------------------------
Name (time in us)                         Min         Max      Mean    StdDev    Median      IQR  Outliers  OPS (Kops/s)  Rounds  Iterations
--------------------------------------------------------------------------------------------------------------------------------------------
test_calc_rolling_windspeed_diff     369.5000  1,267.7000  425.4475  104.1946  390.4500  30.1000     31;54        2.3505     362           1
--------------------------------------------------------------------------------------------------------------------------------------------

[aclerc]

Thanks @gabrielecalvo! Looks great, I will review asap. Did you happen to benchmark test_pre_post_pp_analysis_with_reversal? That would be relevant to the changes you made to the pp_raw_df function.

[gabrielecalvo]

Thanks @gabrielecalvo! Looks great, I will review asap. Did you happen to benchmark test_pre_post_pp_analysis_with_reversal? That would be relevant to the changes you made to the pp_raw_df function.

yes, the last commit I did exactly that. Used the pytest-benchmark library and iterated over the pp_raw_df to make it faster.
The reason I chose that function is because when running cProfile on smart_ole that seemed one of the hotpaths.

[aclerc]

Thanks @gabrielecalvo! Looks great, I will review asap. Did you happen to benchmark test_pre_post_pp_analysis_with_reversal? That would be relevant to the changes you made to the pp_raw_df function.

yes, the last commit I did exactly that. Used the pytest-benchmark library and iterated over the pp_raw_df to make it faster. The reason I chose that function is because when running cProfile on smart_ole that seemed one of the hotpaths.

thanks, sorry I see that info in your commit message now

[aclerc]

Very nice improvements and thank you for adding the new test test_calc_rolling_windspeed_diff