Hacker News Reader

> func pow(uint base, uint n): n == 0 ? return 1 : return n * pow(base, n-1)

Nitpick: that’s not tail recursive. Something like def pow(base, n, acc=1) = match n case 0 => acc; default => pow(base, n-1, acc*base)

Given that `base` is never used, I suspect that's not the only issue :)

> Of course, this means that the same implementation could also be directly expressed logically in a way that is mutable/iterative.

Yes, compilers exist.

> There is no real "advantage" to, or reason to "sell" anybody on tail call recursion if you are able to easily and clearly represent both implementations, IMO.

Avoiding mutation avoids headaches.

> [0] EXCEPT in practice all implementation differences/optimizations introduce observable side effects that can otherwise impact program correctness or semantics. For example, a program could (perhaps implicitly) rely on the fact that it errors out due to stack overflow when recursing > X times, and so enabling TCO could cause the program to enter new/undesirable states; or a program could rely on a functin F making X floating point operations taking at least Y cycles in at least Z microseconds, and not function properly when F takes less than Z microseconds after enabling vectorization. This is Hyrum's Law [2].

These programs are likely not standards compliant. (And that's not just true for the C++ standard but for basically any language with a standard.)

> And just like TCO, RVO is neat but IMO slightly dangerous because it relies on implicit compiler/runtime optimizations that can be accidentally disabled or made non-applicable as code changes:

Who says TCO has to be always implicit? In eg Scheme it's explicit in the standard, and in other languages you can have annotations.

(Whether a call is in tail position is more or less a property you can ascertain syntactically, so your annotation doesn't even have to be understood by the compiler: the linter is good enough. That would catch your 'accidental changes' part.

Things get more complicated, when you have implicit clean-ups happen after returning from the function. Like calling destructors in C++. Then the function call is arguably not in a tail position anymore, and so TCO doesn't apply. Whether this is detectable reliably at compile time depends on the details of your language.)

Avoiding mutation avoids headaches, but the real headaches are actions (mutations) at a distance, and tail recursion vs loops make no difference there.

No mutation (and no side-effects in general) means no action at a distance.

Loops need mutation to work. The mutation might be benign or it might be headache-inducing. Without further analysis you don't know. If there's no mutation, no need for analysis. Lowering the cognitive load.

Well, replacing loops with tail calls is one tool to get rid of some mutations.

It's basically the same reasoning people give you for not using goto in your program: yes, there are ways to use gotos responsible, and there's ways to end up with spaghetti code.

If you use goto, the reader has to analyse and figure out whether you made spaghetti (and the reader can never be quite sure she understood everything and didn't miss an important caveat). If you express yourself without goto, the need for that analysis largely goes away.

I have a similar attitude about const in C++, I use it almost whenever possible. Less (possible) mutation to worry about. But I also let go of it fairly easily when it gets in the way. And... well... tail recursion doesn't feel very idiomatic in C++.

If you iterate by const reference over a const container, and you make every assign-once variable in the loop body const (or in Rust: just not mut), is there any advantage to tail recursion except someone on the internet said it's the proper functional style?

I think functional programming contains some great ideas to keep state under control, but there is no reason to ape certain superficial aspects. E.g. the praise of currying in Haskell tutorials really grinds my gears, I think it's a "clever" but not insightful idea and it really weirds function signatures.

> If you iterate by const reference over a const container, and you make every assign-once variable in the loop body const (or in Rust: just not mut), is there any advantage to tail recursion except someone on the internet said it's the proper functional style?

Function calls can express all kinds of useful and interesting control flow. They are so useful that even people who love imperative programming use functions in their language. (Early and primitive imperative programming languages like very early Fortran and underpowered dialects of BASIC didn't have user defined functions.)

So we established that you want functions in your language anyway. Well, and once you properly optimise function calls, what's known as tail call optimisation, you notice that you don't need no special purpose loops (nor goto) built into your language. You can define these constructs as syntactic sugar over function calls. Just like you can define other combinators like map or filter or tree traversals.

See how in the bad old days, Go had a handful of generic functions and data structures built-in (like arrays), but didn't allow users to define their own. But once you add the ability for users to define their own, you can remove the special case handling.

And that's also one thing C++ does well: as much as possible, it tries to put the user of the language on the same footing as the designers.

When 'map' or 'filter' are the best construct to express what you want to say, you should use them. When a 'for'-loop is the best construct, you should use it. (And that for-loop could be defined under the hood as syntactic sugar on top of function calls.) The scenario your concocted is exactly one where a foreach-loop shines.

Though to be a bit contrarian: depending on what your loop does, it might be useful to pick an ever more constrained tool. Eg if all you do run one action for each item, with no early return and you are not constructing a value, you can use something like Rust's 'foreach' (https://docs.rs/foreach/latest/foreach/). If you transform a container into another container (and no early return etc), you can use 'map'. Etc.

The idea is to show the reader as much as possible what to expect without forcing them to dive deep into the logic. The transformation in a 'map' might be very complicated, but you know the shape of the result immediately from just spying that it's a 'map'.

When you see the for-loop version of the above, you have to wade through the (complicated) body of the loop just to convince yourself that there's no early return and that we are producing a new container with exactly the same shape as the input container.

> I think functional programming contains some great ideas to keep state under control, but there is no reason to ape certain superficial aspects. E.g. the praise of currying in Haskell tutorials really grinds my gears, I think it's a "clever" but not insightful idea and it really weirds function signatures.

Yes, that's mixing up two separate things. Haskell doesn't really need currying. All you need for Haskell to work as a language is a convenient way to do partial application. So if Haskell (like OCaml) used tuples as the standard way to pass multiple arguments, and you had a syntactically convenient way to transform the function (a, b, c) -> d into (b, c) -> d by fixing the first argument that would get you virtually all of the benefits Haskell gets from pervasive currying without the weird function signatures.

In practice, people tend to get used to the weird function signatures pretty quickly, so there's not much pressure to change the system.

Even the first release of FORTRAN had statement functions.

I would argue having the parameters that change during the loop be explicit is a very nice advantage. Agree that the things can be equivalent in terms of execution but the readability and explicitness, and the fact that all the parameters are listed in the same place is great.

Agreed. Some people really like FP a lot, and I think it's underrated that the kinds of functions where TCO is applicable tend to be so simple that they are not really that inelegant when expressed imperatively. My true opinion is that relying on TCO is usually choosing ideological adherence to FP (or "code that looks cooler") over reliability/performance/communication.

That said, just as I'd expect experienced C++ programmers to be able to recognize others' code using RVO and be careful not to restructure things to break it, I'd concede that experienced FP programmers might be unlikely to accidentally break others' TCO. It's just that ad absurdum you cannot expect every developer to be able to read every other developers' mind and recognize/workaround all implicit behavior they encounter.

> [...] and I think it's underrated that the kinds of functions where TCO is applicable tend to be so simple that they are not really that inelegant when expressed imperatively.

I suspect you are only thinking of patterns that are basically equivalent to a loop. I might agree with that.

TCO really shines when you want to implement state machines. See eg https://news.ycombinator.com/item?id=43076088 for an example where using tail calls in Python's interpreter loop gives nice performance benefits. Similar also for LUA.

> [...] I'd concede that experienced FP programmers might be unlikely to accidentally break others' TCO.

Compiler (or linter) checkable annotations would help here. You are right that you want to make it possible for programmers to statically assert somehow that their function call is a tail call.

Btw this reminds me: recursion isn't just something you do in computation, but also in data structures (amongst other things). In eg Rust the members of your data structure are typically just laid out one after another, but when you have a recursive structure (and in certain other cases) you need to box it, otherwise you'd get an infinitely large data structure. Boxing is more or less equivalent to using indirection via a pointer.

However, unboxing isn't really like TCO. It's more like in-lining.

RVO and URVO are slightly different from TCO in that’s the language guarantees that they are required to happen. You are correct though that small changes can accidentally turn it off unintentionally.

With Named RVO (I.e. you explicitly `return named_variable;`) copy-elision is actually guaranteed by the standard. I believe returning the return value of a function call is also guaranteed to not do a copy constructor. Anything else is compiler and optimization level dependent.

To nitpick a bit, NRVO is an optimization as there is no guarantee that it will be performed, but RVO is now guaranteed (you can now return temporary non-copyable /non-movable objects for example).