3 ms·

> There is a large amount of evidence that shows that homeopathy isn't effective. That's not quite how evidence for the existence of effects works. Evidence fo

by pron 8y ago

> There is a large amount of evidence that shows that homeopathy isn't effective. That's not quite how evidence for the existence of effects works. Evidence for lack of an effect is the attempt, and subsequent failure, to find it. Few studies have been able to substantiate the hypothesis that homeopathy is very effective, just like the case for FP. It's just that there have been far fewer studies overall testing the second hypothesis than the first. So while we can be more confident that, indeed, homeopathy isn't effective than in the case of FP, there is still no weight to the claim that FP is effective when it comes to correctness. > there is more evidence pointing that FP leads to correctness than otherwise There isn't. It's perfectly OK that people write about how much they like FP or OOP because the paradigm fits well with how they think, they like its aesthetics, or any similar reasons. It's not OK to try and justify the preference by making up completely unsupported empirical claims. And by the way, even when there is no evidence one way or the other, you still cannot claim there is an effect.

Tehnix 8y ago
>> there is more evidence pointing that FP leads to correctness than otherwise > There isn't. But there is. Take this paper for example, http://web.cs.ucdavis.edu/~filkov/papers/lang_github.pdf http://web.cs.ucdavis.edu/~filkov/papers/lang_github.pdf "A Large Scale Study of Programming Languages and Code Quality in Github". "The data indicates functional languages are better than procedural languages; it suggests that strong typing is better than weak typing; that static typing is better than dynamic; and that managed memory usage is better than un- managed. Further, that the defect proneness of languages in general is not associated with software domains. Also, languages are more related to individual bug categories than bugs overall."
- pron 8y ago
  That study: 1. Found a statistically significant, but a rather small effect (low single digit % IIRC) that cannot justify language/paradigm choice (choose FP for 1.5% fewer bugs!). If anything, it's evidence against a large effect. 2. Had most of even that small effect disappear on reproduction, which increases the evidence against a large effect: https://arxiv.org/abs/1901.10220 https://arxiv.org/abs/1901.10220
  - cannabis_sam 8y ago
    Which puts types in a different category than homeopathy, despite your attempts to associate the two.
    - pron 8y ago
      1. When did I mention types? I was talking about FP vs imperative. There are typed and untyped languages in both paradigms. There actually have been a couple of studies that found a positive effect for types on correctness. The largest effect (15%!) was reported in this paper: http://earlbarr.com/publications/typestudy.pdf http://earlbarr.com/publications/typestudy.pdf but it compared only TypeScript and JavaScript. 2. Why the different category? Both effects were looked for and not found. Homeopathy was just looked for much more, so the certainty it has no effect (beyond placebo) is more certain. Again, the effect of FP on correctness is currently between completely unknown and some evidence against. Repeating this claim without supporting evidence is ridiculous.
      - cannabis_sam 8y ago
        1. Here: https://news.ycombinator.com/item?id=19586915 https://news.ycombinator.com/item?id=19586915 > Explaining why it is unlikely (from a theory perspective) that languages like Haskell have a significant effect on correctness is easier [2] > It's relatively easy to classify which program properties can be assisted by the language and which cannot. Those that can (e.g. memory safety and type safety) are called inductive (or compositional). They can be helpful, but the vast majority of correctness properties aren't inductive. Inductive/compositional means that the property is preserved by all primitive operations in the language. 2. Because homeopathy's claims and method of action can be dismissed a priori, due to the fact that they don't follow the laws of nature or the chemical principles of solvation. Further, your are conflating the question " could (typed) FP be use for correctness" with " should (typed) FP be use for correctness", and while the answer to the first one is trivially "yes", you appear to be chasing the answer you want ("no") to the second question. This puts your approach closer to that of the homeopaths.

by pron 8y ago

Tehnix 8y ago

>> there is more evidence pointing that FP leads to correctness than otherwise > There isn't. But there is. Take this paper for example, http://web.cs.ucdavis.edu/~filkov/papers/lang_github.pdf http://web.cs.ucdavis.edu/~filkov/papers/lang_github.pdf "A Large Scale Study of Programming Languages and Code Quality in Github". "The data indicates functional languages are better than procedural languages; it suggests that strong typing is better than weak typing; that static typing is better than dynamic; and that managed memory usage is better than un- managed. Further, that the defect proneness of languages in general is not associated with software domains. Also, languages are more related to individual bug categories than bugs overall."

pron 8y ago
That study: 1. Found a statistically significant, but a rather small effect (low single digit % IIRC) that cannot justify language/paradigm choice (choose FP for 1.5% fewer bugs!). If anything, it's evidence against a large effect. 2. Had most of even that small effect disappear on reproduction, which increases the evidence against a large effect: https://arxiv.org/abs/1901.10220 https://arxiv.org/abs/1901.10220
- cannabis_sam 8y ago
  Which puts types in a different category than homeopathy, despite your attempts to associate the two.
  - pron 8y ago
    1. When did I mention types? I was talking about FP vs imperative. There are typed and untyped languages in both paradigms. There actually have been a couple of studies that found a positive effect for types on correctness. The largest effect (15%!) was reported in this paper: http://earlbarr.com/publications/typestudy.pdf http://earlbarr.com/publications/typestudy.pdf but it compared only TypeScript and JavaScript. 2. Why the different category? Both effects were looked for and not found. Homeopathy was just looked for much more, so the certainty it has no effect (beyond placebo) is more certain. Again, the effect of FP on correctness is currently between completely unknown and some evidence against. Repeating this claim without supporting evidence is ridiculous.
    - cannabis_sam 8y ago
      1. Here: https://news.ycombinator.com/item?id=19586915 https://news.ycombinator.com/item?id=19586915 > Explaining why it is unlikely (from a theory perspective) that languages like Haskell have a significant effect on correctness is easier [2] > It's relatively easy to classify which program properties can be assisted by the language and which cannot. Those that can (e.g. memory safety and type safety) are called inductive (or compositional). They can be helpful, but the vast majority of correctness properties aren't inductive. Inductive/compositional means that the property is preserved by all primitive operations in the language. 2. Because homeopathy's claims and method of action can be dismissed a priori, due to the fact that they don't follow the laws of nature or the chemical principles of solvation. Further, your are conflating the question " could (typed) FP be use for correctness" with " should (typed) FP be use for correctness", and while the answer to the first one is trivially "yes", you appear to be chasing the answer you want ("no") to the second question. This puts your approach closer to that of the homeopaths.