search

NREL / flasc

A rich floris-driven suite for SCADA analysis
https://nrel.github.io/flasc/
BSD 3-Clause "New" or "Revised" License
32 stars 18 forks source link

Update requirements #15

Closed paulf81 closed 2 years ago

paulf81 commented 2 years ago

I think that the current requirements in the setup file and separate requirements doc were set by @Bartdoekemeijer to specify a not-too recent package for a few modules to avoid an issue in pandas. I think with the most recent version of pandas we can revert to specifying a minimum version, rather than a precise version or max version. @misi9170 would you take a look?

Bartdoekemeijer commented 2 years ago

Hi @paulf81, you are correct! I fixed FLASC to a specific pandas version range,

pandas>=1.3.0,<=1.4.0

To avoid some issues with pandas time resampling functionalities (I think it was in pd.downsample or pd.rolling) that started appearing after version 1.4.0. Perhaps that is solved with a more recent release of Pandas. It's worth churning through the example scripts with the latest Pandas version. If nothing comes up, I think you should be safe to move forward with the latest Pandas version! You'll still want to restrict the requirements such that at the least it skip the versions (I'm guessing 1.4.1 and possibly onward) that yield errors with the time resampling functions.

paulf81 commented 2 years ago

Thanks for the clarification @Bartdoekemeijer, I think @misi9170 has a plan and I do believe that if we use the latest versions of all these modules will avoid the issue. Agree that specifying a version past this issue does make sense in case of installing into an existing environment

misi9170 commented 2 years ago

Thanks @Bartdoekemeijer. I haven't been able to reproduce the original bug. I do see mentions in recent pandas releases of bugs/regressions with rolling and resample; but I don't receive an obvious error when running through the raw_data_processing examples using pandas 1.4.1, 1.4.2, or 1.4.3 (latest).

To be clear, I'm mostly focused on a_06a... as it calls df_downsample, which in turn calls df.resample. I actually don't see any example code that results in a call to df.rolling, as df_movingaverage does not seem to be called anywhere other than its own unit test. What am I missing here?

Bartdoekemeijer commented 2 years ago

Hi @misi9170. Those were the examples I was thinking of too. Unfortunately I never really logged this bug properly, so it's hard to reproduce. I wish I had done this, in hindsight! I would say, if it's not giving you any issues in your work nor the example scripts, feel free to go ahead and set pandas to its most recent version. It's probably fine. If it's not, we'll see soon enough!

misi9170 commented 2 years ago

Will do, thanks!