[Demiurge]

>FP as I understand it doesn't much oppose imperative or even most of OO

I think this is definitely true, if you take 'oppose' as 'incompatible'. If we look the design concept and plot them, there will be axes, and I think this is what we'll see.

>* Even if you want to encapsulate data and logic together like you suggest, you only need to buy 10% of OO to get that

That's also true, but it's not an argument against the seed of usefulness of the 10%. We have kitchen sinks because people pick and choose features at will.

>* FP, as I practice it, means using the simplest possible thing in a world where simple things compose nicely.

Same here, and everyone should practice this when they look at any function, or object method, trying to make it as atomic as possible.

> This may end up being some kind of full-scale OO

I don't think it's any more useful shoehorning a program into "everything objects" more than "everything is functions".

My point was to say that the core goal of OOP is a great one, it simply codifies something that people have done in C when they put structs and procedures in the same header file. Let's not throw the baby out with the bath water.

[tel]

My point is merely that the benefits of are gained using essentially nothing more than real ADTs and yet are almost always bundled with harmful other things in practice. FP can include these core ideas nicely, much like how monads include imperative ideas nicely, by focusing on what composes well.

[Demiurge]

As long as I'm not using C++ I don't feel harmed by any OOP features I don't have to use. I don't feel being restricted. I'm curious, what are these commonly harmful things you're talking about?

On the other hand, most FP languages I tried feel very restrictive. The only language that is inclusive of all good ideas, seems to be F#.

[tel]

I'm not a big fan of languages allowing mutability anywhere. You can use it, sure, but advertise in big bold letters as to where. This plays out repeatedly for any side effect.

There are some OO systems with effect types I suppose. Those would be interesting.

There are also OO systems that don't have mutability. But I think that it you remove mutation, all that class nonsense, and are careful with subtyping (since it breaks things badly often) then you basically have (badly typed) ML modules with open recursion.

The open recursion (a.k.a. late binding) bit is probably the most interesting thing but you can play with it in Typed LC just fine. It's just another form of recursion.

[klibertp]

> There are some OO systems with effect types I suppose. Those would be interesting.

Nim has something IIRC, but I didn't look into it yet.