I'm trying to implement the FFI part for my codegen, and although passing values like Int works, my biggest issue is figuring out how to pass a function.
So, in the docs there is this example with the JS FFI, that passes a function to the JS world, calls it there, and gets the result back, which is what I'm trying to implement. The code in the docs is this:
function twice(f, x) {
return f(f(x));
}
and the Idris code for the FFI binding is this one:
twice : (Int -> Int) -> Int -> IO Int
twice f x = mkForeign (
FFun "twice(%0,%1)" [FFunction FInt FInt, FInt] FInt
) f x
I don't have access to the mkForeign function, because it looks that is from the JS codegen RTS. So I tried to use foreign from Idris builtins:
twice : (Int -> Int) -> Int -> IO Int
twice =
foreign FFI_AHK "twice"
((Int -> Int) -> Int -> IO Int)
Which errors in:
When checking right hand side of twice with expected type
(Int -> Int) -> Int -> IO Int
When checking argument fty to function foreign:
Can't find a value of type
FTy FFI_AHK [] ((Int -> Int) -> Int -> IO Int)
Which I guess it is due to the last parameter of foreign , which has type
{auto fty : FTy f [] ty}
Documented as:
fty - an automatically-found proof that the Idris type works with the FFI in question
How does one satisfy this proof? I'm pretty lost at this point
The JavaScript FFI descriptor part of the docs was very useful in order to fix this.
What was needed was to provide an additional data type in order to wrap the functions so Idris can autogenerate the fty parameter of foreign.
The complete FFI definition of my code generator looks like this:
mutual
public export
data AutoHotkeyFn t = MkAutoHotkeyFn t
public export
data AHK_FnTypes : Type -> Type where
AHK_Fn : AHK_Types s -> AHK_FnTypes t -> AHK_FnTypes (s -> t)
AHK_FnIO : AHK_Types t -> AHK_FnTypes (IO' l t)
AHK_FnBase : AHK_Types t -> AHK_FnTypes t
public export
data AHK_Types : Type -> Type where
AHK_Str : AHK_Types String
AHK_Int : AHK_Types Int
AHK_Float : AHK_Types Double
AHK_Unit : AHK_Types ()
AHK_Raw : AHK_Types (Raw a)
AHK_FnT : AHK_FnTypes t -> AHK_Types (AutoHotkeyFn t)
public export
data AHK_Foreign
= AHK_DotAccess String String
| AHK_Function String
public export
FFI_AHK : FFI
FFI_AHK = MkFFI AHK_Types AHK_Foreign String
In order to write the twice function above, one can wrap the function type with AutoHotkeyFn and then do the same with the value of the function, by wrapping it in MkAutoHotkeyFn:
twice : (Int -> Int) -> Int -> AHK_IO Int
twice f x =
foreign FFI_AHK (AHK_Function "twice")
(AutoHotkeyFn (Int -> Int) -> Int -> AHK_IO Int)
(MkAutoHotkeyFn f)
x
Related
Is there a difference between
foo: {len : _} -> Int -> Vect len Int
and
foo: Int -> {len : _} -> Vect len Int
and similar for data constructors, type constructors etc? Sometimes I find my code compiles with implicits in one position but not in another, and I'm not quite clear on why.
It would matter if you use the value of one implicit in the type of another like:
x : {n : Nat} -> {ts : Vect n Type} -> HVect ts
In this case n must be before ts.
One minor difference: it appears implicits are only in scope if preceding arguments are also in scope. For example, in
foo : {len : _} -> Int -> Vect len Int
foo = ?rhs -- len IS in scope here
while
foo : Int -> {len : _} -> Vect len Int
foo = ?rhs -- len is NOT in scope here
and
foo : Int -> {len : _} -> Vect len Int
foo x = ?rhs -- len IS in scope here
In C# I can represent a function expression like so
Expression<Func<int, int, int>> add = (a, b) => a + b;
The string representation of the function expression looks like this
(a, b) => (a + b)
In Kotlin I can represent a function expression like so
val add = { a: Int, b: Int -> a + b }
The string representation of the function expression looks like this
(kotlin.Int, kotlin.Int) -> kotlin.Int
Is there a way that Kotlin can represent the function expression more aligned to C#, showing the input parameters and the function body?
val add = { a: Int, b: Int -> a + b }
is equivalent to
Func<int, int, int> add = (a, b) => a + b;
not to Expression<...>. And if you print that, you'll see something like System.Func<int, int, int>, like in Kotlin.
I don't think Kotlin has a type like Expression in the standard library, or that you can implement it without language support. Its reflection API is richer than Java's (see KFunction) but doesn't let you access the body. You could do it with byte code manipulation libraries, but it would be a lot of work.
I'm trying to create a functor that makes a polynomial ring out of a ring. My underlying type, Ring_elt, has the following signature:
module type Ring_elt = sig
type t
val add : t -> t -> t
val mul : t -> t -> t
val zer : t
val one : t
val neg : t -> t
end;;
My polynomial functor looks like:
module Make_Poly2(Underlying:Ring_elt) = struct
type t = Poly of Underlying.t list
let rec create lst =
match List.rev lst with
| Underlying.zer :: tl -> create List.rev tl
| _ -> Poly of lst
end;;
(so the 'create' function should take a list, remove the leading zeros, and then return the polynomial of the result). However, I get a syntax error and utop underlines the "zer" after "Underlying."
By comparison, the following code (for making integer polynomials) works:
module Make_int_poly = struct
type t = Poly of int list
let rec create lst =
match List.rev lst with
| 0 :: tl -> create (List.rev tl)
| _ -> Poly lst
end;;
Any idea what's going on?
An OCaml pattern is built from constants, data constructors, and new names bound by the pattern match. Underlying.zer isn't any of those things. But 0 is one of them.
Seems like you can just use an if to compare against Underlying.zer.
Jeffrey's answer is good but instead of correcting it with an if construction, what you should do is the following : use algebraic data types
Instead of writing
val zer : t
val one : t
You could write
module type Ring_elt = sig
type t = Zer | One | Num of t
val add : t -> t -> t
val mul : t -> t -> t
val neg : t -> t
end
module Make_int_poly = struct
type t = Poly of int list
let rec create lst =
match List.rev lst with
| Underlying.Zer :: tl -> create (List.rev tl)
| _ -> Poly lst
end
It's a much better way of doing it since you can easily pattern match on it and even add some constants to your type t without problems.
(Disclaimer: I am fairly certain that this is not idiomatic in any way. If there is some alternative tree-traversal idiom in OCaml, I'm all ears :) )
I am writing a toy compiler in OCaml, and I would like to have a visitor for my large syntax tree type. I wrote one using classes, but I thought it would be fun to try and implement one using modules/functors. My type hierarchy is massive, so let me illustrate what I'm trying to do.
Consider the following type definitions (making these up on the spot):
type expr =
SNum of int
| SVarRef of string
| SAdd of expr * expr
| SDo of stmt list
and stmt =
SIf of expr * expr * expr
| SAssign of string * expr
Let me briefly illustrate usage. Say, for example, I wanted to collect all of the SVarRefs inside of the program. If I had a mapping visitor (one which visits every node of the tree and does nothing by default), I could do the following (in a perfect world):
module VarCollector : (sig
include AST_VISITOR
val var_refs : (string list) ref
end) = struct
include MapVisitor
let var_refs = ref []
let s_var_ref (s : string) =
var_refs := s::!var_refs
SVarRef(s)
end
(* Where my_prog is a stmt type *)
let refs = begin
let _ = VarCollector.visit_stmt my_prog in
VarCollector.!var_refs
end
I should note that the advantage of having functions for each specific variant is that my actual codebase has a type which both has a large amount of variants and does not make sense to fragment. Variant-specific functions allow the avoiding of repeated iteration implementations for the other variants of a type.
Seems simple enough, but here's the catch: there are different types of visitors. MapVisitor returns the original syntax tree, so it has type
sig
(** Dispatches to variant implementations *)
val visit_stmt : stmt -> stmt
val visit_expr : expr -> expr
(** Variant implementations *)
val s_num : int -> expr
val s_var_ref : string -> expr
val s_add : (expr * expr) -> expr
val s_do : stmt list -> expr
val s_if : (expr * expr * expr) -> stmt
val s_assign : (string * expr) -> stmt
end
At the same time, one might imagine a folding visitor in which the return type is some t for every function. Attempting to abstract this as much as possible, here is my attempt:
module type AST_DISPATCHER = sig
type expr_ret
type stmt_ret
val visit_expr : expr -> expr_ret
val visit_stmt : stmt -> stmt_ret
end
(** Concrete type designation goes in AST_VISITOR_IMPL *)
module type AST_VISITOR_IMPL = sig
type expr_ret
type stmt_ret
val s_num : int -> expr_ret
val s_var_ref : string -> expr_ret
val s_add : (expr * expr) -> expr_ret
val s_do : stmt list -> expr_ret
val s_if : (expr * expr * expr) -> stmt_ret
val s_assign : (string * expr) -> stmt_ret
end
module type AST_VISITOR = sig
include AST_VISITOR_IMPL
include AST_DISPATCHER with type expr_ret := expr_ret
and type stmt_ret := stmt_ret
end
(** Dispatcher Implementation *)
module AstDispatcherF(IM : AST_VISITOR_IMPL) : AST_DISPATCHER = struct
type expr_ret = IM.expr_ret
type stmt_ret = IM.stmt_ret
let visit_expr = function
| SNum(i) -> IM.s_num i
| SVarRef(s) -> IM.s_var_ref s
| SAdd(l,r) -> IM.s_add (l,r)
| SDo(sl) -> IM.s_do sl
let visit_stmt = function
| SIf(c,t,f) -> IM.s_if (c,t,f)
| SAssign(s,e) -> IM.s_assign (s,e)
end
module rec MapVisitor : AST_VISITOR = struct
type expr_ret = expr
type stmt_ret = stmt
module D : (AST_DISPATCHER with type expr_ret := expr_ret
and type stmt_ret := stmt_ret)
= AstDispatcherF(MapVisitor)
let visit_expr = D.visit_expr
let visit_stmt = D.visit_stmt
let s_num i = SNum i
let s_var_ref s = SVarRef s
let s_add (l,r) = SAdd(D.visit_expr l, D.visit_expr r)
let s_do sl = SDo(List.map D.visit_stmt sl)
let s_if (c,t,f) = SIf(D.visit_expr c, D.visit_expr t, D.visit_expr f)
let s_assign (s,e) = SAssign(s, D.visit_expr e)
end
Running this gives me the following error message, however:
Error: Signature Mismatch:
Values do not match:
val visit_expr : expr -> expr_ret
is not included in
val visit_expr : expr -> expr_ret
I know this means that I am not expressing the relationship between the types correctly, but I cannot figure out what the fix is in this case.
Disclaimer: Modules are just records of values accompanied with type definitions. Since there are no types in your modules there is no need to use them at all, just use plain old record types, and you will get one of the idiomatic AST traversal pattern. Soon, you will find out that you need an open recursion, and will switch to class-based approach. Anyway, this was the main reason, why classes were added to OCaml. You may also wrap your classes into a state monad, to fold AST with arbitrary user data.
What concerning your error message, then it is simple, you hide your types with signatures, a common mistake. The easiest solution is to omit an annotation of the return type of a functor at all, or to propagate type equality with a with type expr = expr annotations.
If you need examples of more idiomatic approaches, then for records you can go for ppx mappers, here is an example of different visitors implemented with classes, including those that are wrapped into a state monad.
I currently have two "layers" of modules that represent identifier-data relationships in a database.
The first layer defines identifier types, such as IdUser.t or IdPost.t while the second layer defines data types such as User.t or Post.t. I need all the modules of the first layer to be compiled before the modules of the second layer, because a Post.t must hold the IdUser.t of its author and the User.t holds the IdPost.t of the last five posts he visited.
Right now, IdUser.t provides functionality that should only ever be used by User.t, such as the ability to transform an IdUser.t into an IdUser.current: for security reasons, this transform must only ever be performed by the function User.check_password. Since IdUser and User are independent modules, I need to define those features as public functions and rely on conventions to avoid calling them anywhere outside of User, which is rather dirty. A symmetrical situation happens in IdPost.mine:
module IdUser : sig
type t
type current
val current_of_t : t -> current (* <--- Should not be public! *)
end = struct
type t = string
type current = string
let current_of_t x = x
end
module IdPost : sig
type t
type mine
val mine_of_t : t -> mine (* <--- Should not be public! *)
end = struct
type t = string
type mine = string
let mine_of_t x = x
end
module Post : sig
(* Should not "see" IdUser.current_of_t but needs IdPost.mine_of_t *)
val is_mine : IdUser.current -> IdPost.t -> IdPost.mine
end
module User : sig
(* Should not "see" IdPost.mine_of_t but needs IdUser.current_of_t *)
val check_password : IdUser.t -> password:string -> IdUser.current
end
Is there a way of defining an current_of_t : t -> current function in IdUser that can only be called from within module User ?
EDIT: this was a simplified example of one pair of modules, but there's an obvious solution for a single pair that cannot be generalized to multiple pairs and I need to solve this for multiple pairs — about 18 pairs, actually... So, I've extended it to be an example of two pairs.
So IdUser is in reality an existential type: For User there exists a type
IdUser.current such that the public IdUser.t can be lifted to it. There are a couple of ways to encode this: either functorize User as Gasche shows if statically managing the dependence is sufficient, or use first-class modules or objects if you need more dynamism.
I'll work out Gasche's example a bit more, using private type abbreviations for convenience and to show how to leverage translucency to avoid privatizing implementation types too much. First, and this might be a limitation, I want to declare an ADT of persistent IDs:
(* File id.ml *)
module type ID = sig
type t
type current = private t
end
module type PERSISTENT_ID = sig
include ID
val persist : t -> current
end
With this I can define the type of Posts using concrete types for the IDs but with ADTs to enforce the business rules relating to persistence:
(* File post.ml *)
module Post
(UID : ID with type t = string)
(PID : PERSISTENT_ID with type t = int)
: sig
val is_mine : UID.current -> PID.t -> PID.current
end = struct
let is_mine uid pid =
if (uid : UID.current :> UID.t) = "me" && pid = 0
then PID.persist pid
else failwith "is_mine"
end
The same thing with Users:
(* File user.ml *)
module User
(UID : PERSISTENT_ID with type t = string)
: sig
val check_password : UID.t -> password:string -> UID.current
end = struct
let check_password uid ~password =
if uid = "scott" && password = "tiger"
then UID.persist uid
else failwith "check_password"
end
Note that in both cases I make use of the concrete but private ID types. Tying all together is a simple matter of actually defining the ID ADTs with their persistence rules:
module IdUser = struct
type t = string
type current = string
let persist x = x
end
module IdPost = struct
type t = int
type current = int
let persist x = x
end
module MyUser = User (IdUser)
module MyPost = Post (IdUser) (IdPost)
At this point and to fully decouple the dependencies you will probably need signatures for USER and POST that can be exported from this module, but it's a simple matter of adding them in.
One way that seems to work at least on your simplified example is to group IdUser and User inside a same module:
module UserAndFriends : sig ... end = struct
module IdUser : sig
...
end = struct
...
end
module User = struct
...
end
end
module Post : sig
val create : (* <--- Should not "see" IdUser.current_of_t *)
author:IdUser.current -> title:string -> body:string -> IdPost.t
end
Hiding the dangerous function(s) in the signature of UserAndFriends gives the result you desire. If you do not want to make a big file containing both IdUser and User, you can use option -pack of ocamlc to create UserAndFriends. Note that in this case, you must craft your Makefile carefully so that the .cmi files of IdUser and User are not visible when compiling Post. I am not the Makefile specialist for Frama-C, but I think we use separate directories and position the compiler option -I carefully.
I suggest you parametrize Post (and possibly User for consistency) by a signature for the IdUser module : you would use a signature with current_of_t for User, and one without for Post.
This guarantee that Post doesn't use IdUser private features, but the public interface of IdUser is still too permissive. But with this setup, you have reversed the dependencies, and IdUser (the sensitive part) can control its use directly, give itself (with the private part) to IdUser and restrict the public signature to the public parts.
module type PrivateIdUser = sig
val secret : unit
end
module type PublicIdUser = sig
end
module type UserSig = sig
(* ... *)
end
module MakeUser (IdUser : PrivateIdUser) : UserSig = struct
(* ... *)
end
module IdUser : sig
include PublicIdUser
module User : UserSig
end
= struct
module IdUser = struct
let secret = ()
end
module User = MakeUser(IdUser)
include IdUser
end
module Post = struct
(* ... *)
end
Edit : Pascal Cuoq's concurrent -- in the temporal sense -- solution is alos very nice. Actually it's simpler and has less boilerplate. My solution adds an abstraction that allows for slightly more modularity, as you can define User independently of IdUser.
I think which solution is best probably depends on the specific application. If you have a lot of different modules that use PrivateIdUser private information, then using functors to write them separately instead of bundling everyone in the same module can be a good idea. If only User needs to be in the "private zone" and it's not very big, then Pascal's solution is a better choice.
Finally, while being forced to explicit Private and Public interfaces can be seen as an additional burden, it is also a way to make the access properties of different modules more explicit that using the position inside the module hierarchy.
It's possible to achieve fine-grained control over signatures with a combination of recursive modules, first-class modules and GADTs, but the limitation would be that all modules should then be inside the same top-level module and unpackings of first-class modules inside the recursive modules should be done in each function separately (not on the module-level as it would cause runtime exception Undefined_recursive_module):
module rec M1 : sig
module type M2's_sig = sig
val a : int
val c : float
end
module type M3's_sig = sig
val b : string
val c : float
end
type _ accessor =
| I'm_M2 : M2.wit -> (module M2's_sig) accessor
| I'm_M3 : M3.wit -> (module M3's_sig) accessor
val access : 'a accessor -> 'a
type wit
val do_it : unit -> unit
end = struct
module type M2's_sig = sig
val a : int
val c : float
end
module type M3's_sig = sig
val b : string
val c : float
end
type _ accessor =
| I'm_M2 : M2.wit -> (module M2's_sig) accessor
| I'm_M3 : M3.wit -> (module M3's_sig) accessor
module M1 = struct
let a = 1
let b = "1"
let c = 1.
end
let access : type a. a accessor -> a =
function
| I'm_M2 _ -> (module M1)
| I'm_M3 _ -> (module M1)
type wit = W
let do_it () =
let (module M2) = M2.(access ## I'm_M1 W) in
let (module M3) = M3.(access ## I'm_M1 W) in
Printf.printf "M1: M2: %d %s M3: %d %s\n" M2.a M2.b M3.a M3.b
end
and M2 : sig
module type M1's_sig = sig
val a : int
val b : string
end
module type M3's_sig = sig
val b : string
val c : float
end
type _ accessor =
| I'm_M1 : M1.wit -> (module M1's_sig) accessor
| I'm_M3 : M3.wit -> (module M3's_sig) accessor
val access : 'a accessor -> 'a
type wit
val do_it : unit -> unit
end = struct
module type M1's_sig = sig
val a : int
val b : string
end
module type M3's_sig = sig
val b : string
val c : float
end
type _ accessor =
| I'm_M1 : M1.wit -> (module M1's_sig) accessor
| I'm_M3 : M3.wit -> (module M3's_sig) accessor
module M2 = struct
let a = 2
let b = "2"
let c = 2.
end
let access : type a. a accessor -> a =
function
| I'm_M1 _ -> (module M2)
| I'm_M3 _ -> (module M2)
type wit = W
let do_it () =
let (module M1) = M1.(access ## I'm_M2 W) in
let (module M3) = M3.(access ## I'm_M2 W) in
Printf.printf "M2: M1: %d %f M3: %d %f\n" M1.a M1.c M3.a M3.c
end
and M3 : sig
module type M1's_sig = sig
val a : int
val b : string
end
module type M2's_sig = sig
val a : int
val c : float
end
type _ accessor =
| I'm_M1 : M1.wit -> (module M1's_sig) accessor
| I'm_M2 : M2.wit -> (module M2's_sig) accessor
val access : 'a accessor -> 'a
type wit
val do_it : unit -> unit
end = struct
module type M1's_sig = sig
val a : int
val b : string
end
module type M2's_sig = sig
val a : int
val c : float
end
type _ accessor =
| I'm_M1 : M1.wit -> (module M1's_sig) accessor
| I'm_M2 : M2.wit -> (module M2's_sig) accessor
module M3 = struct
let a = 3
let b = "3"
let c = 3.
end
let access : type a. a accessor -> a =
function
| I'm_M1 _ -> (module M3)
| I'm_M2 _ -> (module M3)
type wit = W
let do_it () =
let (module M1) = M1.(access ## I'm_M3 W) in
let (module M2) = M2.(access ## I'm_M3 W) in
Printf.printf "M3: M1: %s %f M2: %s %f\n" M1.b M1.c M2.b M2.c
end
let () =
M1.do_it ();
M2.do_it ();
M3.do_it ()