Ask HN: Is there empirical evidence that static types reduce bugs? | Hacker News Reader

Ask HN: Is there empirical evidence that static types reduce bugs?

4 pointsKinrany6 comments

Hillel will shank the owner of that open source library you're competing with if you can provide access to relevant internal studies of some tech corp: https://twitter.com/hillelogram/status/1291063196317605889

View on HN

I think a better way to phrase the question is: "Is there empirical evidence that static types reduce software defects that make it into production?"

Or alternatively if that's not a concern "Is there evidence that static typing increases developer velocity by reducing debugging time?"

It's pretty unquestionable that some classes of bugs are eliminated at compile time. But whether it actually ends up mattering in the big scheme of things is another question.

I'm guessing the answer to the first question is a qualified yes, while the answer to the second is no.

How could it not? By definition, strong statically typed languages won't let you release code with certain classes of bugs so they must reduce bugs going live unless humans are perfect bug detectors.

A lot of research goes into testing things that seem obvious. Sometimes it turns out the obvious answer is wrong. I think this is why Kinrany asked for empirical evidence.

What specific part of my argument are you disagreeing with though? You don't need empirical evidence for what you can derive from definitions. It would be more complex if the question involved time or cost.

I agree with what you're saying in general and I'd love to see some robust studies too but you have to accept here that it's prohibitively impractical to get many multiples of developers/teams to develop the same app over several months through to the maintenance stage with all other variables isolated.

That's why most studies use inexperienced students or are based on crawling GitHub repos, then end up being really flawed.

It would probably be challenging to show autocomplete, syntax highlighting and undo/redo lead to productivity gains as well - nobody has the resources and/or will to investigate these at a large scale with all other variables isolated.

It's not that there's anything specific to disagree with in your statement. The point is that when the question is "Is there evidence...?" an answer like "This is what makes sense to me..." is almost irrelevant.

I like the way another comment handled the same concerns: https://news.ycombinator.com/item?id=24069315

They offered some different questions, pointed out the issues with the line of inquiry, then restated the question under the question in a different way:

>> "It's pretty unquestionable that some classes of bugs are eliminated at compile time. But whether it actually ends up mattering in the big scheme of things is another question."

And that question is worth studying.

Reply on news.ycombinator.com