It's the design patterns if you are trying to use a 100% pure language, which is a waste of time. Here's the thing...at some point being functionally pure stops paying off. At some point you end up paying more to fit your work into a pure model, then getting any benefit from that model. This is the problem I have with CT. It seems cool and may make you feel smart, but it doesn't actually provide any benefit to writing maintainable code
Why is maintainable code the goal and not expressiveness, or being able to write close to the domain?
Not that being maintainable isn't a worthy goal or that expressiveness and DSLs can't be maintainable, just wondering why it would be the primary goal.
IS CT only used when writing a piece of code that's supposed to be long living and have regular updates or something?
Maintainable code is very important, unless you work for some company that just produces code then dumps it on a client. The key is to balance functional purity with understandability.
Let's take something like Clojure. I get immutable data, sane multithreading (I haven't had thread related problems in my code for years), etc. All without monads. Whenever I've started doing things like introducing state monads, or seen code involving applicative and "free" stuff, it basically becomes an opaque blob of functions. Zero ability to have the code self introspect or to leverage that pile of functions.
The better method is immutable data and data-first designs. Data is searchable, introspect-able, and transformable. Most CT stuff is all about composing functions via functions. Making your entire system opaque, at least from the point of view of the program itself.
