Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I've been writing unit tests for over 15 years (actually longer, but before that they were just throwaway run-once testing stubs). I wouldn't consider what you got ChatGPT to produce to be an adequate rewrite of a function to make it unit testable (and several others in this thread have expressed the same view). Even the "hack" using monkey patching makes for a more actually-useful test.


I've been writing them for over 20. Doesn't matter. What matters is factual correctness.

See here what a unit test is versus integration test: https://www.testim.io/blog/unit-test-vs-integration-test/

And also see my reply to the other guy where I definitively show that not only is what chatgpt produced correct, but better:

https://news.ycombinator.com/item?id=36064097


I'm perfectly aware of what the difference is, thank you. The function you gave to ChatGPT explicitly requests a stream to format integers as decimals, separated by newlines. The version it gave as being 'unit-testable' did not, and hence wasn't a 'factually correct' answer. In some cases that may be perfectly fine, but in others it most definitely isn't.


This is the function I gave chatGPT:

   ME:
   def (x: int):
      for i in range(x):
         print(i)
It takes an int. Are you trolling now or did you get mixed up with something else?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: