Every now and then, I see a thread about how functional programming languages are The Way, wondering why they aren't more popular.
A big problem is that people talk about functional programming as The Way for Everything. FP has its place. All tools have their roles. 🧵 1/
A big problem is that people talk about functional programming as The Way for Everything. FP has its place. All tools have their roles. 🧵 1/
The elegance of functional programming makes it an appealing "silver bullet" candidate.
But here are the domains where FP really shines:
1. Programming education
2. Prototyping language features and analyses
3. Domains where you're mostly specifying transformations
2/
But here are the domains where FP really shines:
1. Programming education
2. Prototyping language features and analyses
3. Domains where you're mostly specifying transformations
2/
I fell in love with Scheme and SML in undergrad. I taught myself Haskell and extended it for my senior thesis. In grad school I worked in Haskell, then OCaml, then Scala.
But by the end of grad school, I was mostly using Python.
Here's what happened.
3/
But by the end of grad school, I was mostly using Python.
Here's what happened.
3/
In the beginning, I wondered why everyone didn't use functional programming. When I interned at Google in 2007 I asked them weekly why we were still using C. My mentor would tell me, we don't even have the 50% memory overhead to use Java.
It's a matter of time, I thought.
4/
It's a matter of time, I thought.
4/
When I started grad school, I was building most of my stuff in Haskell. I loved the elegance of pure functional programming.
Then I decided to implement Earley parsing. After a weekend mucking with monads for a 4D l mutable array, I was over not having mutable state.
5/
Then I decided to implement Earley parsing. After a weekend mucking with monads for a 4D l mutable array, I was over not having mutable state.
5/
Then I switched to OCaml. See my pros-and-cons here:
jxyzabc.blogspot.com/2009/03/haskel…
But to scale OCaml for my data processing needs, I had to:
1. Stop using closures.
2. Use references everywhere.
3. Make foreign function calls out to C++.
In conclusion, I was writing C++.
6/
jxyzabc.blogspot.com/2009/03/haskel…
But to scale OCaml for my data processing needs, I had to:
1. Stop using closures.
2. Use references everywhere.
3. Make foreign function calls out to C++.
In conclusion, I was writing C++.
6/
At this point, I still very much believed in strong, statically typed functional languages, so I decided to try Scala.
I was very into Scala for a while and even got featured on their blog for evangelizing Scala, with this post:
jxyzabc.blogspot.com/2012/11/should…
7/
I was very into Scala for a while and even got featured on their blog for evangelizing Scala, with this post:
jxyzabc.blogspot.com/2012/11/should…
7/
But I again hit a snag. I was prototyping new features for Jeeves, the programming language I created during my PhD. I was embedding it as a DSL in Scala and every experiment was a few hours of wrestling with the Scala typechecker.
One day, I gave up:
jxyzabc.blogspot.com/2013/12/im-usi…
8/
One day, I gave up:
jxyzabc.blogspot.com/2013/12/im-usi…
8/
I'm still a big proponent of strong static types and functional languages. I recommend them as thinking tools without reservation. For building systems, however, it's important to match the tool to the task!
(And FP evangelists would benefit from acknowledge this!)
end 🧵/
(And FP evangelists would benefit from acknowledge this!)
end 🧵/
• • •
Missing some Tweet in this thread? You can try to
force a refresh
