Being Left Behind Enjoyer@thomasfuchs@hachyderm.io
It’s always “AI is great for generating boilerplate code” and never “why do we even need boilerplate code, maybe programming is broken”
Being Left Behind Enjoyer@thomasfuchs@hachyderm.io
David Chisnall (*Now with 50% more sarcasm!*)@david_chisnall@infosec.exchange
@thomasfuchs@hachyderm.io
It's also worth noting that boilerplate is technical debt. Any code that is copied between large numbers of consumers of an API mostly unmodified imposes a bunch of undocumented constraints on how that API can evolve.
If you factor out the boilerplate, then you have to handle changes to the APIs in one place. If you duplicate it, you have to fix it everywhere.
A tool that allows you to write boilerplate more quickly is removing the incentive to avoid boilerplate and so is encouraging code that is hard to maintain.
Zhaknafein :StarWars: :debian:@zhaknafein@mastodon.uno
@thomasfuchs@hachyderm.io Why not toasterplate code? Or fridgeplate code...
Brian@bbaugh@infosec.exchange
@thomasfuchs@hachyderm.io boiler plate code lies in the dead zone between core functionality provided by the language and finished product. Such code needs to meet the needs of the finished product but those needs can differ product-to-product. Every product will solve many of the same problem but slightly differently. We have tried (and failed at) at least one alternative solution: packaged modules.
David Chisnall (*Now with 50% more sarcasm!*)@david_chisnall@infosec.exchange
@thomasfuchs@hachyderm.io
The one good use I’ve seen for LLMs in programming is to spot missing abstractions. If an LLM can generate code for a particular task, it means that this task has been repeated enough times that it should be factored out.
Allan Chow@grumpasaurus@infosec.exchange
@thomasfuchs@hachyderm.io it's great for boiler plate code when you enjoy reviewing and correcting boiler plate code