4 ms·

The part on “Make code more concise“ struck me as an anti-pattern that should instead be a simple function composition. Or the functional programmer in me notes

by tbenst 7y ago

The part on “Make code more concise“ struck me as an anti-pattern that should instead be a simple function composition. Or the functional programmer in me notes that the author re-invented the Maybe monad. I do like the point in Learn the user’s language, as friendlier error messages is something we should all strive for, although I’ve never had an issue with particular problem in Python’s stacktraces, and actually having types like Julia or at least annotations via mypy seems a better solution. CRAN is a great point, and pythons packaging is in a sorry state with a crazy number of approaches and undeclared dependencies. R does a great job here. Functional programming section is ironic given the lack of functional patterns in the post. R has even fewer higher order functions than the python standard library. It’s hard for me to see how R is better for production than python, and the argument against pandas seems a bit strawman considering that numpy/scipy are quite stable and more central to the ecosystem than DataFrame. R is fantastic for data science and highly productive, until you need to do data mugging or anything else that involves a general purpose language.

cameronh90 7y ago
"CRAN is a great point, and pythons packaging is in a sorry state with a crazy number of approaches and undeclared dependencies. R does a great job here" But for production usage, again it's a huge pain. It's difficult to keep version stability with developer machines since there's no standard lock file, and the CRAN servers often delete or silently update old versions of packages. Packrat and Microsoft's MRAN really helps, but another curious issue is that it and other CRAN servers seem to have terrible stability - often going down for hours at a time (or worse). Python's packaging is ~impossible to understand for new developers (and really absolutely needs improving), but in an organisation you just pick an approach suited for your use case.
- tbenst 7y ago
  Nix really shines here, has ~17000 R packages pinned and reproducibly built: https://hydra.nixos.org/eval/1560470#tabs-still-succeed https://hydra.nixos.org/eval/1560470#tabs-still-succeed
zamalek 7y ago
The problem with Python conciseness is that it requires cognition to write. That cognition needs to multiplied ten fold for understanding. I think I'm a good developer, I can't understand idiomatic Python. Python could use some verbosity for the sake of everyone else. If R slows you IQ 9000 people down, please, make it standard.

by tbenst 7y ago

cameronh90 7y ago

"CRAN is a great point, and pythons packaging is in a sorry state with a crazy number of approaches and undeclared dependencies. R does a great job here" But for production usage, again it's a huge pain. It's difficult to keep version stability with developer machines since there's no standard lock file, and the CRAN servers often delete or silently update old versions of packages. Packrat and Microsoft's MRAN really helps, but another curious issue is that it and other CRAN servers seem to have terrible stability - often going down for hours at a time (or worse). Python's packaging is ~impossible to understand for new developers (and really absolutely needs improving), but in an organisation you just pick an approach suited for your use case.

tbenst 7y ago
Nix really shines here, has ~17000 R packages pinned and reproducibly built: https://hydra.nixos.org/eval/1560470#tabs-still-succeed https://hydra.nixos.org/eval/1560470#tabs-still-succeed

zamalek 7y ago

The problem with Python conciseness is that it requires cognition to write. That cognition needs to multiplied ten fold for understanding. I think I'm a good developer, I can't understand idiomatic Python. Python could use some verbosity for the sake of everyone else. If R slows you IQ 9000 people down, please, make it standard.