Tuesday, June 13, 2017

Another "Problems With Python" List

This: https://darkf.github.io/posts/problems-i-have-with-python.html

Summary.
  1. Slow.
  2. Threads.
  3. Legacy Python 2 Code.
  4. Some Inconsistencies.
  5. Functional Programming.
  6. Class Definitions Don't Have Enough Features.
  7. Switch Statement.

Responses.
  1. Find the 20% that needs a speedup and write that in C. Most of the time that's already available in numpy. The rest of the time you may have found something useful.
  2. Generally, most folks make a hash of threaded programming. A focus on process-level parallelism is simpler and essentially guarantees success by avoiding shared data structures.
  3. Maybe stop using the legacy projects?
  4. Yup.
  5. Proper functional programming would requires an optimizer, something that doesn't fit well with easy-to-debug Python. It would also require adding some features to cope the optimization of functional code (e.g., monads.) It seems to be a net loss. And http://coconut-lang.org.
  6. Consider a metaclass that provides the missing features?
  7. I can't figure out why 'elif' is considered hard to use. The more complex matching rules are pretty easy to implement, but I guess this falls into the "awful hacks" category.

What causes me to write this is the lack of concrete "do this instead" for most of the points. It sounds too much like empty complaining.

I hope for some follow-on from this on Twitter:
But I'm not optimistic. It's too easy to complain and too hard to solve.