16
\$\begingroup\$

What I'd wish for is that I wouldn't have to be a human lint tool.
— Pimgd

There are questions on CR where running the code through lint or a similar tool could give useful tips to the poster, especially for beginners who don't know about static analysis tools.

Posting an output from an existing tool wouldn't make for a good answer, but if it was tailored specifically for CR (justifying the review, including helpful links, using friendly language, etc.) maybe it could.

What do you think? Could a bot that posted automated review answers make sense? Or maybe a comment with a link to a report?

\$\endgroup\$
6
  • 1
    \$\begingroup\$ I think we should at least have this discussion, regardless of the outcome. It's important to know what exactly the place of validators and automated reviews is on codereview. \$\endgroup\$
    – Mast Mod
    May 1, 2015 at 14:28
  • \$\begingroup\$ Why not post this as a CW answer--not with a bot \$\endgroup\$
    – nhgrif
    May 1, 2015 at 14:29
  • 3
    \$\begingroup\$ I for one have run our vba static analysis tool on posted code and used the output as part of my answer. The keyword being "part". I don't think posting such results on their own makes for a very good answer. \$\endgroup\$
    – RubberDuck
    May 1, 2015 at 14:30
  • \$\begingroup\$ @RubberDuck Why not? Aren't partial reviews encouraged? \$\endgroup\$
    – svick
    May 1, 2015 at 14:31
  • \$\begingroup\$ Yeah. They are, and maybe I'd feel different if it wasn't output from my project. \$\endgroup\$
    – RubberDuck
    May 1, 2015 at 14:32
  • 2
    \$\begingroup\$ Would a list of static analysis tools help jumpstart this? \$\endgroup\$
    – h.j.k.
    May 1, 2015 at 14:36

2 Answers 2

Reset to default
15
\$\begingroup\$

I'm not sure that an automated answering bot or whatever would be appropriate, for a couple of reasons. The primary one is this:

It is not a person.

Why does this matter? Well, to me, it matters a lot. Sure, a "lint" type tool can show you errors and potential problems, but it cannot teach. I feel one of the great things in having a real programmer review your code, is they not only can help improve the code, they also help to teach you how to learn better. This, a lint tool cannot do.

There are a few other concerns, mostly logistical in nature:

  • Bot would probably be a PITA to design;

  • Bot would likely need to be a legit SE user (like @Duga) to have privileges to post on the site;

  • Bot would take away opportunities for human users to gain reputation/badges/etc.

  • Bot might confuse new users into thinking it's a real person;

  • How/who manages the bots reputation, badges, etc.?

Instead...

Someone suggested in comments:

Would a list of static analysis tools help jumpstart this? – h.j.k.

I think having a list of static tools in itself would be useful enough for reference.

Update!

Such a list of static tools is being developed here:

What static analysis tools exist that can help pick up points for answering CR questions?

\$\endgroup\$
6
  • 4
    \$\begingroup\$ "but it cannot teach" Yes. This. ++ \$\endgroup\$
    – RubberDuck
    May 1, 2015 at 14:54
  • 2
    \$\begingroup\$ Doesn't teaching just mean explaining why something is wrong? Current lint tools usually don't do that, but a CR-specific one could. \$\endgroup\$
    – svick
    May 1, 2015 at 15:00
  • 2
    \$\begingroup\$ A lint tool would not pick up on things like code repetition, design patterns, inefficient (albeit not "wrong") code, how to improve algorithm, etc. etc. \$\endgroup\$
    – Phrancis
    May 1, 2015 at 15:12
  • \$\begingroup\$ @sᴉɔuɐɹɥԀ Yes, I'm not expecting automated reviews to be a replacement for human reviews. I expect it to only catch things that can be detected automatically and that are tedious for human reviewers. \$\endgroup\$
    – svick
    May 1, 2015 at 16:50
  • \$\begingroup\$ +1 for not human. If static analysis was perfect, there wouldn't be options in them to ignore warnings. \$\endgroup\$
    – Comintern
    May 3, 2015 at 14:35
  • \$\begingroup\$ This answer seems to present a false dilemma, such a linting tool/pre-commit does not have to be a complete answer, nor does it have to replace the human answers. Instead, it could be a(n optional) support tool to provide help, not to replace humans. (This suggestion, as an optional assist comment to the report is also included in the question). I do appreciate the link to the tools that are being developed, that is constructive and helpful, thank you! \$\endgroup\$
    – a.t.
    Jul 8, 2022 at 15:32
7
\$\begingroup\$

I often use standard tools (e.g. lint, valgrind) or my own tools to help me compose a review. Recently, I suggested creating such tools as a possible Code Challenge.

There's no substitute for an experienced human looking at one's code, but promoting automated tools seems to me to be a good adjunct to Code Review's mission.

I should clarify that I would not be in favor of having bot-posted reviews, but I am in favor of helping programmers help themselves, such as with a list of tools and resources.

One possibility might be to add to the Help Center and describe how to turn on compiler warnings for various different compilers, assemblers, etc. and to give the suggestion that compiling cleanly without errors or warnings before posting code for review is a way to keep the signal-to-noise ratio on Code Review high.

\$\endgroup\$
3
  • 2
    \$\begingroup\$ Perhaps we should have a community wiki here on meta with a list of such tools? Another option is to add them to the tag wikis of specific languages each tool targets. \$\endgroup\$
    – RubberDuck
    May 1, 2015 at 15:20
  • \$\begingroup\$ @RubberDuck I like the tag wiki idea a lot! \$\endgroup\$
    – Phrancis
    May 1, 2015 at 15:36
  • \$\begingroup\$ Looks like there's a list of tools growing here on meta as we speak. \$\endgroup\$
    – RubberDuck
    May 1, 2015 at 15:38

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .