That's the neat part : nowhere. (Well kinda, but no spoil).

Well, I have a working thing but I'm not sure it's what was expected. I'll look at the other solutions when I'm done with this kata.

Ps : does your version of join really go through all elements ? If that's not the case then the tests may take elements forever.

Currently, I am battling a bit with Haskell so my join function doesn't do much; but when it "worked", it indeed went through all elements, except that you would see 2^n elements of the first stream before seeing the first of the n-th stream, which is probably why it broke down (NB. when I click on Attempt, it successfully passes the first batch of tests for LazyThunk, but not for LazyOption, because it times out — so the implementation is correct, it's just too long)... Whatever, I guess I'll just change the order, even if that's a bit longer to code.

About Haskell, I'm a noob too. I was hinting at the documentation eg. zipWith.

I don't even know where to put the initialization !

That's the neat part : nowhere. (Well kinda, but no spoil).

Finally about join : the order doesn't matter because it is not tested, but the tests still need to access some elements so they have to be at a reasonable position (something exponential would not be reasonable).

Ps : does your version of join really go through all elements ? If that's not the case then the tests may take elements forever.

This kata is quite under specified, yeah. In case you (or anyone) want to know what a function is supposed to do, it’s likely to be equivalent to the homonymous Haskell function.

Unfortunately, I know almost nothing about Haskell (if I'm currently 4 kyu in Haskell it's merely because I found Kata were you don't need to know how to code in Haskell to finish them).

For fib_aux, let’s say it’s a bit like building a list recursively, but adding stuff to the tail side lazily in the future

My problem about that is how do I reuse already built information. I mean, I don't even know where to put the initializatoin!

tail and join, can (and should) be really lazy. After all it’s a ’a L.t (or a Stream of it), so no evaluation is needed right now.

Modifying signatures in general will most likely stop anything from even compiling. (but refactoring modules is possible, as written in the solution setup).

After some thoughs, I realized what I was supposed to do for these.

The order of the element doesn’t matter in join, you just need to have every element of the « 2D stream » in the output « 1D stream ».

Except it does :') I tested my solution and it [stack overlowed / exceeded the allowed runtime] because the order I chose made it so that you have to wait very long before you actually reach the first element of the n-th stream. But I guess that's easy to fix (just a bit sad that my unpractical solution is rejected). I realized after posting that it was not because a given order was enforced, but rather because mine was really bad.

Thanks for the tips! I think the only thing I am still struggling with is the fibonacci one...

• The order of the element doesn’t matter in join, you just need to have every element of the «  2D stream » in the output « 1D stream ». Trying to exhaust the first stream, then the second, etc … will not work because they are, well, infinite.

• Modifying signatures in general will most likely stop anything from even compiling. (but refactoring modules is possible, as written in the solution setup).

• tail and join, can (and should) be really lazy. After all it’s a ’a L.t (or a Stream of it), so no evaluation is needed right now.

• For fib_aux, let’s say it’s a bit like building a list recursively, but adding stuff to the tail side lazily in the future :

(* With list syntax : *)
(* Instead of *)
new_thing :: already_built
(* You do *)
currently_built :: will_do_in_the_future

• About tie specificaly, it’s hard to wrap your head around how it works - at least that’s my experience. So sorry I can’t help.

• This kata is quite under specified, yeah. In case you (or anyone) want to know what a function is supposed to do, it’s likely to be equivalent to the homonymous Haskell function.

I’ll try to answer some of your points without revealing too much of the solution. (Just in case I’ll split the answer so a spoiler flag will not hide everything).

After some more attempts, I have circumvented the tail issue, by realizing it was my join which was not lazy enough! And that's because, even though it was successfully type-checked, it didn't do what it was supposed to because there wasn't a clear explanation about that :| This kata could really benefit from being more precise.

I have some trouble with this Kata. First of all, it's not clear at all what the tie function should do (or, really, any other function). It's relatively straightforward to "guess" what other functions should do from their signature, but tie's signature is not enough for me. I guess it should be something like "create a dummy thunk in a reference, pass it to the function given and store the result where the dummy thunk was, then return it" but I'm not sure. And, because I'm unsure about that, I'm a little bit confused about what I should do for the fibonacci one.

Secondly, I don't understand how it is possible in OCaml to actually make tail "lazy enough" with the provided signature. That is, the tests require the head not to be evaluated if I only want to tail of the stream, but since both the head and the tail are in the same thunk, it's impossible to evaluate one without the other. I tried changing the signature of Stream to provide two thunks, but then it doesn't work because the constructor is used directly in the tests (which is strange since the module also provides an interface to create streams).

Finally, I'm not sure about which order to use in the join function. Since there are several ways to do that, and that each provides a different stream in the end, it's not clear which one should be chosen.

I have the impression that this solution is wrong... Consider "1110000000111" (3x1, 7x0, 3x1). The shortest sequence of zeros or ones is of length 3, and yet the time unit is 1. You should take the gcd instead of the min.

A match over a boolean value is just an if.