Re:RStudio is now Posit, our mission continues
by toddm on Friday November 04, 2022 @06:19PM
(#33474699)
I'm a scientist who has been coding for over 30 years now, and this is important for context. I am not a software developer/engineer. Almost all of my python coding is done within Jupyter notebooks.
Recently, I had to complete a take-home project involving characterizing the pharmacological properties of a molecule. Most of code was lifted and modified from a very complete set of tutorials, so there were no functional problems on that end.
However - and mind you, this is being done and redone with
!conda install <whatever> -c conda-forge - nglview isn't compatible with ipywidgets versions later than 7.1.1. So this is going on in your Jupyter environment, so who knows what state it's in.
The fact that the yaml conda install requirements worked on my Mac (M2, Monterey 12.6) and not on my Windoze 10 box is itself a problem - I couldn't get that to work at all, and got dinged by the potential employer for non-portability of code - is itself a bigger problem.
While the mainstream use cases for python are not going to entail headaches with conda/pip (which pip? oh, right), incompatible dependencies, and grabbing libraries from non-standard channels (conda), I can assure you that the workflow I encounter in "data scientist" roles is very much a mess.
Finally, the issue of scientists/researchers writing crappy code and hard-coding things and so on is a behavioral problem endemic to the "data scientists" and is language-independent.
Recently, I had to complete a take-home project involving characterizing the pharmacological properties of a molecule. Most of code was lifted and modified from a very complete set of tutorials, so there were no functional problems on that end.
However - and mind you, this is being done and redone with !conda install <whatever> -c conda-forge - nglview isn't compatible with ipywidgets versions later than 7.1.1. So this is going on in your Jupyter environment, so who knows what state it's in.
The fact that the yaml conda install requirements worked on my Mac (M2, Monterey 12.6) and not on my Windoze 10 box is itself a problem - I couldn't get that to work at all, and got dinged by the potential employer for non-portability of code - is itself a bigger problem.
While the mainstream use cases for python are not going to entail headaches with conda/pip (which pip? oh, right), incompatible dependencies, and grabbing libraries from non-standard channels (conda), I can assure you that the workflow I encounter in "data scientist" roles is very much a mess.
Finally, the issue of scientists/researchers writing crappy code and hard-coding things and so on is a behavioral problem endemic to the "data scientists" and is language-independent.