Wow, the examples of Trelent are so incredibly useless. If you're going to generate comments like that, just don't. It is a waste of time to go through it as it is obvious from the function signature. And anything that could probably be written to be useful in the comment can't be grasped by LLM. LLMs just add padding to data, they add no content.
Programming
Welcome to the main community in programming.dev! Feel free to post anything relating to programming here!
Cross posting is strongly encouraged in the instance. If you feel your post or another person's post makes sense in another community cross post into it.
Hope you enjoy the instance!
Rules
Rules
- Follow the programming.dev instance rules
- Keep content related to programming in some way
- If you're posting long videos try to add in some form of tldr for those who don't want to watch videos
Wormhole
Follow the wormhole through a path of communities [email protected]
yeah, any comment that can be auto-generated doesn't need to be there in the first place. The real issue might be overly-dogmatic comment/docstring requirements
Even worse. Auto generated comments just copy the how and not the why. If the code has a bug so does the comment.. and if you fix the bug but not the comment somebody else might change the code back to match the comment.
I think the typical approach is in the opposite direction.. Write a comment about what the code should do then have the AI write the code, adjusting for any mistakes it makes.
Hm... just looking at Trelent briefly, I wouldn't automatically assume that it's privacy-protecting -- it uploads the source to their server so they can do ML on it. That seems like a bigger issue than the registration or not to be honest.
Maybe just run a local Vicuna instance and copy-paste code with requests for Vicuna to comment it? I do that with GPT and it seems to work quite well. Less convenient than what you're saying but maybe serviceable?
I will give it a try. Thanks for sharing
Not saying a tool like this isn't helpful for you but personally I don't think I'd ever trust a service to analyze my code and provide accurate comments. There's so many little quirks and ultimately nuance in what is written vs what and why it's doing it that I'd never expect anyone but the original author of the code to get it right.
RE your edit: I also support that conclusion and I’m glad you’ll give it a shot yourself. A mindset that helps me is this: commenting is part of the iterative code writing process. When I’m struggling to put a concise and understandable comment above some code, it almost universally means that there’s something about the code itself I should arrange more clearly. This is your chance to do some rubber ducking, it’s valuable to both you and the next person to read your code!
This is gold advice. Thank you.