Code linters should check code style, not code correctness. I don't remember a single instance of a lint rule for correctness that accurately detected an issue, but I have seen countless instances of a linter recommendation introducing a bug or otherwise creating friction.
-
Show this thread
-
The programmer probably understands what their code is doing better than a trivial heuristic rule that was based on a single example taken from a completely different codebase. Trust the programmer over the heuristic.
3 replies 3 retweets 33 likesShow this thread -
Replying to @fchollet
I'm curious about your opinion on the current research applying Deep Learning to Source Code Do you think in the future linters and other developer tooling will be empowered by ML? Disclaimer: I definitely believe so, that's why I left
@google and joined@sourcedtech +#MLonCode2 replies 0 retweets 2 likes -
It's possible to learn programs that can read and write code, & assist software developers (with linting and beyond). But DL is most definitely not capable of doing this well, since it requires understanding (simulating) code flow, rather than just pattern-matching code snippets
1 reply 0 retweets 5 likes -
Thanks! That's a bit discouraging to hear, but I hope you're maybe wrong for once
I've been reading papers like https://arxiv.org/abs/1409.5718 which are able to detect bad patterns like bubble sort quite successfully (+94% acc)
Would you recommend any other paper to change my mind?1 reply 0 retweets 2 likes
There are useful things you can do with mere pattern recognition, even in software development. You just need to be aware of the limits of your tools. DL does one thing and does it well (within its constraints), but it is not a magic wand you can wave at any problem.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.