search

SantanderMetGroup / downscaleR

An R package for climate data bias correction and downscaling (part of the climate4R bundle)
https://github.com/SantanderMetGroup/climate4R
GNU General Public License v3.0
103 stars 59 forks source link

Adding python dependencies #106

Open cofinoa opened 8 months ago

cofinoa commented 8 months ago

@miturbide I have added the python dependencies as a conda environment file

I don't know if there is any dependency with specific versions, please review!!

cofinoa commented 8 months ago

This relates to SantanderMetGroup/climate4R#60 and SantanderMetGroup/climate4R/issues/61

mfansler commented 8 months ago

It would be better to instead create a dedicated environment (not r-reticulate) that provides only the software needed by the Python scripts in this package. Since r-reticulate is the default for the reticulate package, it is very much prone to arbitrary changes by the user.

I'd also suggest to more explicitly constrain versions in the YAML file. What matters most is (presumably) correctness and consistency of execution. Ideally, installing say v3.4.0 of this package should always result in mostly the same Python environment. Without versions in the YAML, someone today might end up with Python 3.11, but a year from now with Python 3.13. And it would be unpredictable whether such version-creep will lead to identical evaluation of the Python scripts. I recommend identifying an environment with more definite versions (e.g., major + minor). Whenever a change to those package versions happens, it should involve updating the version of the R package.