I'm trying to implement a singly-linked list (queue) in OCAML which is very difficult and not as intuitive as I thought it would be. Let me first state the invariants of the queue that I am implementing:
q.head and q.tail are either both None, or
- q.head and q.tail both point to Some nodes, and
- q.tail is reachable by following 'next' pointers
from q.head
- q.tail's next pointer is None.
The queues and their associated nodes are declared as using options as follows:
type 'a qnode = { v: 'a;
mutable next: 'a qnode option }
type 'a queue = { mutable head: 'a qnode option;
mutable tail: 'a qnode option }
I have many test cases but the one that fails first is this one:
let test (): bool =
let q = from_list [5; 6; 7] in
delete 5 q;
valid q && to_list q = [6; 7]
;; run_test "delete element from front of queue" test
Valid () just tests whether the queue satisfies the invariants. And here is my implementation:
let delete (elt: 'a) (q: 'a queue) : unit =
if not (valid q) then failwith "delete: given invalid queue";
let rec loop_delete (elt: 'a) (no: 'a qnode option) : unit =
begin match no with
|None -> ()
|Some n -> (*Item to delete is only item in queue*)
if n.v = elt && n.next = None then
(q.head <- None; q.tail <- None)
else if n.v = elt && n.next = Some n
then (q.head <- n.next; n.next <- None) else loop_delete elt n.next
end
in loop_delete elt q.head
Why is n.next not updating appropriately? Moreover, I'm sure there are many other issues wrong with my implementation. Any help is appreciated.
EDIT:
According to suggestions below, I've made the following adjustments. The comments indicate remaining confusions.
let delete (elt: 'a) (q: 'a queue) : unit =
if not (valid q) then failwith "delete: given invalid queue";
let rec loop_delete (elt: 'a) (prev: 'a qnode option)
(curr: 'a qnode option) : unit =
begin match prev, curr with
|_, None -> ()
|None, Some n -> if n.v = elt && n.next = None then (q.head <- None; q.tail <- None) (*One element in list only...right??*)
else if n.v = elt && n.next != None then (*How to continuously update prev and curr??* else ()
|Some n, Some n1 -> (*??*)
end
in loop_delete elt None q.head
(Sorry my earlier answer was wrong. I missed that you're using q directly in your helper function.)
Your function loop_delete looks for the matching node, but when it finds the matching node it seems to make unwarranted assumptions. If the matching node n is at the end of the queue (n.next = None), it assumes it was the only element of the queue (it sets q.head and q.tail to None). If the matching node isn't at the end of the queue, it seems to assume that the matching node was at the beginning of the queue (it sets q.head to n.next).
As a separate problem, this is not a reasonable test:
n.next = Some n
What it tests is whether the next pointer of n points to n itself. If you just want to know whether n.next points to any node you could say n.next <> None.
As an even more general comment, when deleting from a singly linked list you generally need to keep track of two nodes, the node you're interested in and the previous node in the list. When you find the node you want to delete, you need to change the next value of the previous node.
You shouldn't really need to think about pointers when coding in OCaml, but it's not too wrong to think of None as a null pointer and Some n as a pointer to the node n. For the first call you don't have a previous node, so the call might look like this:
loop_delete elt None q.head
When you want to move to the next node inside loop_delete, the call might look something like this:
loop_delete elt no n.next
In essence you are passing two values of type 'a qnode option. They represent the previous node and the current node. If there's no previous node it's because you're at the beginning of the list. If there's no current node it's because you're at the end of the list.
Sorry for unhelpful comments earlier. Maybe this is a little better.
Related
The following code will not work as I expected it to:
// Gets all or nothing. If predicate is false for at least one item
// in the sequence, returns None. If seq is empty, returns the empty seq.
let tryGetAllOrNone predicate (source: seq<_>) =
let mutable condition = true
let resultSeq =
seq {
use e = source.GetEnumerator()
while condition && e.MoveNext() && (condition <- predicate e.Current; condition) do
yield e.Current
}
if condition then Some resultSeq
else None
The reason is quite clear: a sequence is lazily evaluated, which means that here the if statement will be evaluated first, returning the sequence. Then, when we consume the resulting sequence we will always get Some results, until the condition turns false:
// expect: None (predicate is false for the first item)
> [1;2;3] |> tryGetAllOrNone (fun x -> x = 2);;
val it : seq<int> option = Some (seq [])
// expect None (predicate is false for the second item)
> [1;2;3] |> tryGetAllOrNone (fun x -> x = 1);;
val it : seq<int> option = Some (seq [1])
// correct (predicate is always true)
> [1;2;3] |> tryGetAllOrNone (fun x -> x > 0);;
val it : seq<int> option = Some (seq [1; 2; 3])
I might just have to consume the sequence first, i.e. by using [...yield ...] instead of seq { .. yield ...}, but maybe there's a simpler solution that retains the laziness (just asking the question makes it sound backwards, so my gut tells me: consume first, right)?
EDIT: thinking about this is a tad longer has me come to the conclusion that what I'm asking is not possible. You cannot first lazily return one after the other from a sequence and then, upon hitting an invalid item, say: "hey, all those items you got thus far, give them back, if one is invalid, all are invalid!".
Leaving it out here nonetheless in case it helps someone else or in case someone has a better idea ;).
You are right that, in general, you will need to iterate over the whole sequence until you can return None or Some. If the last element of the sequence does not satsify the condition, then you need to first read all the preceding elements until you know that you have failed.
The only optimization you can do is that you can return None as soon as you find a first element that does not satisfy the condition. This is slightly better than building the whole sequence and then checking if the condition was false for any of the elemnetns (in that, if an earlier element fails to satisfy the condition, you can return None sooner). The following implementation does that:
// Gets all or nothing. If predicate is false for at least one item
// in the sequence, returns None. If seq is empty, returns the empty seq.
let tryGetAllOrNone predicate (source: seq<_>) =
let results = ResizeArray<_>()
let mutable failed = false
use e = source.GetEnumerator()
while not failed && e.MoveNext() do
if not (predicate e.Current) then failed <- true
else results.Add(e.Current)
if failed then None
else Some(seq results)
Anyone have a decent example, preferably practical/useful, they could post demonstrating the concept?
I came across this term somewhere that I’m unable to find, probably it has to do something with a function returning a function while enclosing on some mutable variable. So there’s no visible mutation.
Probably Haskell community has originated the idea where mutation happens in another area not visible to the scope. I maybe vague here so seeking help to understand more.
It's a good idea to hide mutation, so the consumers of the API won't inadvartently change something unexpectedly. This just means that you have to encapsulate your mutable data/state. This can be done via objects (yes, objects), but what you are referring to in your question can be done with a closure, the canonical example is a counter:
let countUp =
let mutable count = 0
(fun () -> count <- count + 1
count)
countUp() // 1
countUp() // 2
countUp() // 3
You cannot access the mutable count variable directly.
Another example would be using mutable state within a function so that you cannot observe it, and the function is, for all intents and purposes, referentially transparent. Take for example the following function that reverses a string not character-wise, but rather by taking individual text elements (which, depending on language, can be more than one character):
let reverseStringU s =
if Core.string.IsNullOrEmpty s then s else
let rec iter acc (ee : System.Globalization.TextElementEnumerator) =
if not <| ee.MoveNext () then acc else
let e = ee.GetTextElement ()
iter (e :: acc) ee
let inline append x s = (^s : (member Append : ^x -> ^s) (s, x))
let sb = System.Text.StringBuilder s.Length
System.Globalization.StringInfo.GetTextElementEnumerator s
|> iter []
|> List.fold (fun a e -> append e a) sb
|> string
It uses a StringBuilder internally but you cannot observe this externally.
I have a list of type (string * (int * int)) list. I want to be able to search through the list, finding the right element by it's string identifier, do a calculation on one of the ints, and then return the full, modified list.
Example:
Given a list
let st = [("a1",(100,10)); ("a2",(50,20)); ("a3",(25,40))]
I'm trying to make a function which gets one of the elements and subtracts number from one of the ints in the tuple.
get ("a2",10) st
//Expected result: st' = [("a1",(100,10)); ("a2",(40,20)); ("a3",(25,40))]
I feel I'm almost there, but am a little stuck with the following function:
let rec get (a,k) st =
match st with
| (a',(n',p'))::rest when a'=a && k<=n' -> (n'-k,p')::rest
| (a',(n',p'))::rest -> (n',p')::get (a,k) rest
| _ -> failwith "Illegal input"
This returns [("a2",(40,20)); ("a3",(25,40))] and is thus missing the first a1 element. Any hints?
Lists are immutable, so if you want to "change one element" you are really creating a new list with one element transformed. The easiest way to do a transformation like this is to use List.map function. I would write something like:
let updateElement key f st =
st |> List.map (fun (k, v) -> if k = key then k, f v else k, v)
updateElement is a helper that takes a key, update function and an input. It returns list where the element with the given key has been transformed using the given function. For example, to increment the first number associated with a2, you can write:
let st = [("a1",(100,10)); ("a2",(50,20)); ("a3",(25,40))]
st |> updateElement "a2" (fun (a, b) -> a + 10, b)
I was looking for a function which would update an element in a list based on the element's data. I couldn't find one in F#5, so wrote one using Tomas' solution:
let updateAt (elemFindFunc: 'a -> bool) (newElem: 'a) (source: 'a list) : 'a list =
source
|> List.map
(fun elem ->
let foundElem = elemFindFunc elem
if foundElem then newElem else elem)
elemFindFunc is the function which consumes an element and returns true if this is the element we want to replace. If this function returns true for multiple elements, then those will be replaced by newElem. Also, if elemFindFunc evaluates to false for all elements, the list will be unaltered.
newElem is the new value you want to replace with. newElem could be replaced by a function like valueFunc: 'a -> 'a if you want to process the element before inserting it.
If I have a type named Person, and list of functions, for example...
let checks = [checkAge; checkWeight; checkHeight]
...where each function is of the type (Person -> bool), and I want to do the equivalent of...
checkAge >> checkWeight >> checkHeight
...but I don't know in advance what functions are in the list, how would I do it?
I tried the following...
checks |> List.reduce (>>)
...but this gives the following error...
error FS0001: Type mismatch. Expecting a
(Person -> bool) -> (Person -> bool) -> Person -> bool
but given a
(Person -> bool) -> (bool -> 'a) -> Person -> 'a
The type 'Person' does not match the type 'bool'
What am I doing wrong?
It looks like Railway oriented programming would be a good fit here.
If you choose to go this route, you basically have two options.
You can either go all in, or the quick route.
Quick route
You rewrite your validation functions to take a Person option instead of just plain Person.
let validAge (record:Record option) =
match record with
| Some rec when rec.Age < 65 && rec.Age > 18 -> record
| None -> None
Now you should be able to easily chain your function.
checks |> List.reduce (>>)
All in
Alternatively, if you are lazy and don't want to match .. with in every validation function, you can write some more code.
(samples taken from [1])
First there's a bit of setup to do.
We'll define a special return type, so we can get meaningful error messages.
type Result<'TSuccess,'TFailure> =
| Success of 'TSuccess
| Failure of 'TFailure
A bind function, to bind the validations together
let bind switchFunction =
function
| Success s -> switchFunction s
| Failure f -> Failure f
You'll have to rewrite your validation functions as well.
let validAge (record:Record) =
if record.Age < 65 && record.Age > 18 then Success input
else Failure "Not the right age bracket"
Now combine with
checks |> List.reduce (fun acc elem -> acc >> bind elem)
Either way, check out the original article.
There's much more there you might be able to use :)
Edit: I just noticed that I was too slow in writing this answer once again.
Besides, I think Helge explained the concetp better than I did as well.
You may somehow have stumbled upon a dreaded concept. Apperently its the Voldemort (dont say his name!) of functional programming.
With no further ado lets walk right into the code:
type Person =
{ Name : string
Age : int
Weight : int
Height : int }
type Result =
| Ok of Person
| Fail
let bind f m =
match m with
| Ok p -> f p
| _ -> Fail
let (>=>) f1 f2 = f1 >> (bind f2)
let checkAge p =
if p.Age > 18 then Ok(p)
else Fail
let checkWeight p =
if p.Weight < 80 then Ok(p)
else Fail
let checkHeight p =
if p.Height > 150 then Ok(p)
else Fail
let checks = [ checkAge; checkWeight; checkHeight ]
let allcheckfunc = checks |> List.reduce (>=>)
let combinedChecks =
checkAge
>=> checkWeight
>=> checkHeight
let p1 =
{ Name = "p1"
Age = 10
Weight = 20
Height = 110 }
let p2 =
{ Name = "p2"
Age = 19
Weight = 65
Height = 180 }
allcheckfunc p1
combinedChecks p1
allcheckfunc p2
combineChecks p2
At this point I could throw around a lot of weirdo lingo (not really true, I couldnt...), but lets just look at what I have done.
I dropped your return value of bool and went for another type (Result) with either (mark that keyword!) Ok or Fail.
Then made a helper (bind) to wrap and unwrapp stuff from that Result-type.
And a new operator (>=>) to combine the stuff in reduce.
Mind that the first check-function to Fail will shortcut the entire chain and more or less immediately (not calling the other functions) return Fail. In addition you will not know where in this chain it did Fail or which functions ahead of any Fail did actually Ok.
There are methods to also accumulate the errors as you go along, so that you get get a feedback of type: "the checkAge returned Fail, but the others was great success"
The code is mostly stolen from here: http://fsharpforfunandprofit.com/posts/recipe-part2/
And you may want to read about the entire website of Wlaschin and even a lot more to get into the finer and harder details if wanted.
Good luck on your journey to the upper floors of the Ivory Tower. ;-)
Footnote: This is called an Either-monad usually. Its not entirely finished and what not in the above code, but never mind... I think it will work in your case...
The >> operator is useful if you have functions that perform some transformation. For example, if you had a list of functions Person -> Person that turn one person into another.
In your case, it seems that you have functions Person -> bool and you want to build a composed function that returns true if all functions return true.
Using List.reduce you can write:
checks|> List.reduce (fun f g -> (fun p -> f p && g p))
Perhaps an easier option is to just write a function that takes a person and uses List.forall:
let checkAll checks person = checks |> List.forall (fun f -> f person)
On my journey to learning F#, I've run into a problem I cant solve. I have defined a custom type:
type BinTree =
| Node of int * BinTree * BinTree
| Empty
I have made a function which takes a tree, traverses it, and adds the elements it visits to a list, and returns it:
let rec inOrder tree =
seq{
match tree with
| Node (data, left, right) ->
yield! inOrder left
yield data;
yield! inOrder right
| Empty -> ()
}
|> Seq.to_list;
Now I want to create a function, similar to this, which takes a tree and a function, traverses it and applies a function to each node, then returns the tree:
mapInOrder : ('a -> 'b) -> 'a BinTree -> 'b BinTree
This seems easy, and it probably is! But I'm not sure how to return the tree. I've tried this:
let rec mapInOrder f tree =
match tree with
| Node(data, left, right) ->
mapInOrder f left
Node(f(data), left, right)
mapInOrder f right
| Empty -> ()
but this returns a unit. I havent worked with custom types before, so I'm probably missing something there!
Try this:
let rec mapInOrder f = function
| Node(a,l,r) ->
let newL = mapInOrder f l
let b = f a
let newR = mapInOrder f r
Node(b,newL,newR)
| Empty -> Empty
If the function is side-effect free, then traversal order is unimportant and you can instead write:
let rec map f = function
| Node(a,l,r) -> Node(f a, map f l, map f r)
| Empty -> Empty
A match is an expression. It returns the value of the matching case. That implies that all match cases must have the same type. The match expression itself then has that type.
In your first attempt, your Empty clause returned (), and thus had unit type--not the tree type you were looking for.
Since mapInOrder just returns the match result, it too took on unit return type.
The Node clause was fine because its return value is the result of calling mapInOrder, so it also took on unit type and the requirement that all match clauses have the same type was satisfied.
A key change in kvb's suggestion was making the Empty clause return a tree instead of unit. Once you do that, you get compiler errors and warnings pointing to the other problems.
You can often work through issues like this by explicitly coding the type you'd like, and then seeing where the compile errors and warnings show up.