> Just like LLMs took off recently but still play a minor role in almost every area I know.
Okay, let's do the comparison.
LLMs became a big story less than a year ago. Since then, they are, among other things, powering the fastest growing app of all time. They are poised to take over pretty much every NLU functionality, products are being developed every day that wouldn't be able to otherwise, and major players in the industry throw billions at them. Their, and other ML achievements, impact on society are estimated to be so profound, that they have caused ripples in international politics, rekindled debates about UBI, made geopolitical positions about chip maufacturing and AI research even more relevant, and set lawmakers all around the globe scrambling to figure out how to deal with this new reality, regarding everything from copyright laws to regulations.
So, what is going on in the purely-functional-PL space?
Well, the flagship of these languages, Haskell, had it's stable release 12 years ago. Since then, the only software product that I can name without googling that's written in it, is pandoc. The language is currently Number 34 on the Tiobe index. In fact, I am not sure any of the current top20 could be called "purely functional", which would place these languages below Scratch, Fortran, and Classic Visual Basic in terms of that index.
Regarding usefullness, what have purely functional PLs demonstrated? That they can be challenging to use coming from an imperative background, and that they can be a drag on performance. On the plus side they offer ... yeah, what exactly is it they offer, that imperative languages adopting some of their features, like functions as first class objects, cannot?
So I have a hard time seeing any indicators of an impending take-off for purely functional PLs.
I am aware that most mainstream languages incorporated FP features. And they are the better for it.
But the topic of this discussion is PURE functional programming. See the following quote:
"same reason why it took some time for pure functional programming to take off too."
So when did that happen? Almost every codebase I see is imperative, either procedural or OO. They use mutable state, they use functions with side effects.
The reason why my argument above is built around PFP Languages, is because they are the only ones where pure functional codebases seem to exist from my PoV. I am happy to change my mind if you can point me towards major projects written in a PFP style in non PFP PLs.
> But the topic of this discussion is PURE functional programming.
Yes, exactly. It is NOT about "pure functional programming languages" nor is it about (what's nowadays called "functional programming" features). But you seem to be talking about those two things.
> So when did that happen? Almost every codebase I see is imperative, either procedural or OO. They use mutable state, they use functions with side effects.
Then maybe you need to broaden your horizon. I've worked exclusively(!) in different code bases over the last 5+ years that followed a pure(!) functional style, maybe with very few exceptions for performance or sometimes to make things simpler since it was not Haskell.
ZIO is a PFP library for Scala (even though they don't market it as that to be more beginner-friendly, but read the documentation and you'll see that it's essentially an IO-monad on steroids). And it is a rather new one but already got lots of traction. And even typescript has now fp-ts which is used much more than I expected myself.
Of course, being a ZIO adopter does not necessarily mean the whole codebase is using ZIO (and hence following a PFP approach) but the chance is high and I personally know two more companies that use ZIO and are not in the list, so there are very likely much more companies that use ZIO. Also, those companies were using a PFP style pretty much everywhere, in every team (backend only though). So it's not like there was just one project in that style or something like that.
When I compare that to 10 years ago, there was pretty much nothing besides Haskell and Haskell is, as you correctly said, not very prevalent in visible production projects. But checkout the list and to your surprise you might find some familiar names in there.
> (even though they don't market it as that to be more beginner-friendly, but read the documentation and you'll see that it's essentially an IO-monad on steroids)
Okay, let's do the comparison.
LLMs became a big story less than a year ago. Since then, they are, among other things, powering the fastest growing app of all time. They are poised to take over pretty much every NLU functionality, products are being developed every day that wouldn't be able to otherwise, and major players in the industry throw billions at them. Their, and other ML achievements, impact on society are estimated to be so profound, that they have caused ripples in international politics, rekindled debates about UBI, made geopolitical positions about chip maufacturing and AI research even more relevant, and set lawmakers all around the globe scrambling to figure out how to deal with this new reality, regarding everything from copyright laws to regulations.
So, what is going on in the purely-functional-PL space?
Well, the flagship of these languages, Haskell, had it's stable release 12 years ago. Since then, the only software product that I can name without googling that's written in it, is pandoc. The language is currently Number 34 on the Tiobe index. In fact, I am not sure any of the current top20 could be called "purely functional", which would place these languages below Scratch, Fortran, and Classic Visual Basic in terms of that index.
https://www.tiobe.com/tiobe-index/
Regarding usefullness, what have purely functional PLs demonstrated? That they can be challenging to use coming from an imperative background, and that they can be a drag on performance. On the plus side they offer ... yeah, what exactly is it they offer, that imperative languages adopting some of their features, like functions as first class objects, cannot?
So I have a hard time seeing any indicators of an impending take-off for purely functional PLs.