I don't understand your argument, in that I really can't see how one can equally argue the opposite. I fully acknowledge that Haskell has been successful as a test bed for PL research, and I also think Haskell is an interesting language worth studying. It is also true that FP ideas have increasingly bled into mainstream languages over the past two decades. But that does in no way imply that Haskell is in any way a good general choice for writing useful software.

And although a lot of practically useful stuff has come out of Haskell-land (property based testing, for example), it's not even obvious to me that Haskell has been a huge part of the shift to more FP-y idioms (compared, to say, scheme or SML). Your own example, fold, has been around decades before Haskell.

What can I say, I didn't read why es6, php or cpp lifted closures into the language. I consider this is pure FP influence which haskell is the most known instance. And yeah scheme and sml were prior art, but sml is not even on the map of PL choices (I saw it mentionned in rust history, but I think it's mostly the eager semantics, otherwise you could replace it with any pure FP language)

What do you mean with es6 lifted closures into the language? As far as I remember, closures have been in JS right from the start (and given that Brendan Eich joined mozilla to "do scheme in the browser", we can be pretty sure where they came from). ES6 just added a more lightweight syntax.

I agree that SML isn't even on the map, in terms of practical programming language choices (one of the most hilariously titled CS books of all time must _ML for the working programmer_ where the final "real project" chapters consist of writing such working programer perennials as a lambda calculus interpreter and a theorem prover). And I believe that even scheme is less suited to write almost any type of practical software than Haskell. But that's precisely my point: a language having a positive influence on the state of applied programming does not mean that it is actually a good choice for solving practical problems with. And vice versa.

You're right, I just meant the lexical closure syntax (..) => (giving a clear fp feel to the language by promoting closure syntactically). and this is part of a wide move from all dyn langs to lift fp features like destructuring. es6 was just one tiny point that I had in mind when commenting above.

btw I love sml. I'm just being realist about it's mainstream existence.

Lastly, it's as if you consider haskell as proof of concept and nothing more. I don't see any real hurdles why haskell is less practical than typescript (except the fact that TS relies on the years of js project they can compile and interop with)

How about: a typical typescript project does not start off with a dozen compiler pragmas non-trivially changing language semantics? Or: no matter their background, it will take at least 10x more time to spin up new developers with Haskell than typescript?

Haskell language extensions do not change the language semantics. They extend the semantics with additional functionality. Existing code does not change its meaning when you enable an extension.

Strict. Not a change in language semantics at all.

Ah, great counterexample! StrictData is another. Do you know any others that change language semantics non-trivially? I would roughly define "non-trivially" to mean "correct programs become incorrect".

By the way, I'm particularly interested because I am writing about Haskell myths. I have a very rough draft at http://h2.jaguarpaw.co.uk/posts/haskell-myths/

There’s also QuasiQuoter which can make valid code with list comprehensions not compile, but it doesn’t just compile it, it gets stuck looking for the end tag. Not sure if that qualifies :P

Oh, I would happily admit that Haskell has had some influence on languages -- as expected from a research language.