Skip to content

Latest commit

 

History

History
23 lines (18 loc) · 2.77 KB

Python-Love-Hate.md

File metadata and controls

23 lines (18 loc) · 2.77 KB

Python Love/Hate

Three things I love about python:*

  1. Unit testing. Python has it, R doesn't--simple as that. I thought my photometry workflow might stay simple enough to do without this. Ha. And now I can have real confidence in the 'simple' stuff on which my increasingly complex stuff is built.
  2. Object-orientation / real classes: Classes greatly help focus attention. They allow complex code structures without losing your way (read: mind). This is also sufficient reason, by itself, to port.
  3. Execution speed: Nice to have. R is blindingly fast for the vector stuff it's meant for, but pretty slow for general programming.
  4. PyCharm pro IDE: Luscious-looking. Sometimes clumsy with package management (site vs virtualenv), and nags you when it gets typing wrong. But wonderful unit-testing support, and did I mention that it Looks. So. Good.

OK, four--four things I love about python. And now for...

Three things I dearly hate about python:

  1. No private variables, methods, functions, etc: Oh, Jesus how could they have been so ****ing stupid? Stupid as the day is long. No more to say--just stupid. Except that I will say more (of course). Why not just build your house in the middle of a freeway and let everyone else drive through it all day and night? The only justification for prohibiting python coders from having a tool as essential as private objects is: "Don't you trust your fellow coders?" To which I answer: "No, of course I don't--whyever would you?--and furthermore what business is it of yours if I don't?" And did I mention: it's stupid.
  2. Platform schizophrenia / coder abuse: The open-source model is probably at fault here: everyone tries to inflict pet prejudices (read: errors) on the rest of the world. Example: pandas df1.extend(df2) returns a copy of the extended dataframe. Naturally. But python's own list1.extend(list2) returns...nothing. So: df1.extend(df2).extend(df3) does exactly what you'd expect. Whereas the natural syntax list1.extend(list2).extend(list3) stops your program cold. Well, excuse me for being reasonable. And don't even mention statsmodels trying to cram 'exog' and 'endog' variables down their users' throats, when at 2 standard naming systems have been accepted worldwide for a century. And statsmodels mixed-model predict() forgets to include random effects in their predictions at all, even as an option--and their docs brag about this failure. Python is littered with this crap, everywhere, often costing hours to figure out and code around. They could learn a lot from R and even (horrors) from .Net.
  3. 'Self' is plastered everywhere. This is so self-absorbed. Me, me, me. Self, self, self. So perfect for the age of Trump, so perfect for a 30-year-old language still living in its parent's basement.