[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

A basic type system on top of Prolog, along with higher-order clauses. Translates to normal Prolog.

Notifications You must be signed in to change notification settings

kyledewey/typed-prolog

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

typed-prolog

A basic type system on top of Prolog, along with higher-order clauses. Translates to normal Prolog.

Central Features

Typed-Prolog provides the following basic features:

  • An algebraic type system with local type inference, based on Hindley-Milner. This means the following two core components:
    • Generic types (introduced with datatype definitions)
    • Parametric polymorphism (only supported at the clause level)
  • Higher-order clauses, the equivalent of higher-order functions in a logic programming setting. Type inference is automatic for these. These functions compile down to standard first-order Prolog, via defunctionalization. In practice, not only does this make such functions safer to use (we do not resort to any eval-like metaprogramming), these also run much faster.
  • Staightforward modules
  • Global variables

Example

Below is an example which uses a variety of features of Typed-Prolog.

module(routines, [myMap/3, sum/2, sumNative/2], [myList]).

datadef(myList, [A], [myCons(A, myList(A)), myNil]).

clausedef(foldRight, [A, B], [myList(A), B, relation([A, B, B]), B]).
foldRight(myNil, Accum, _, Accum).
foldRight(myCons(H, T), Accum, Relation, Result) :-
  foldRight(T, Accum, Relation, TailAccum),
  call(Relation, H, TailAccum, Result).

clausedef(myMap, [A, B], [myList(A), relation([A, B]), myList(B)]).
myMap(Input, Mapper, Output) :-
  foldRight(Input,
            myNil,
            lambda([A, CurList, myCons(B, CurList)],
                   call(Mapper, A, B)),
            Output).

clausedef(sum, [], [myList(int), int]).
sum(Input, Output) :-
  foldRight(Input,
            0,
            lambda([X, Y, Z],
                   Z is X + Y),
            Output).

clausedef(sumNative, [], [list(int), int]).
sumNative(Input, Output) :-
  sumNative(Input, 0, Output).

clausedef(sumNative, [], [list(int), int, int]).
sumNative([], Sum, Sum).
sumNative([H|T], Accum, Sum) :-
  NewAccum is Accum + H,
  sumNative(T, NewAccum, Sum).

The above example, if saved in the file routines.pl, should compile. The above example shows the following parts:

  1. Module definition, using module. This takes three components:
    1. The name of the module, which should match the filename
    2. The export list of procedures, which describe procedures which this module makes public to other modules. Any procedures defined in the file which are not part of this list are considered private to the module, and cannot be accessed from outside of the module. For example, in the above code myMap/3 is public, but foldRight/4 is private.
    3. The export list of datatypes, which describe datatypes which this modules makes public to other modules. All constructors of the given datatype are exported. Any datatypes defined in the file which are not part of this list are considered private to the module, and cannot be accessed from outside of the module. The only datatype in play in the example is myList, which is publicly exported.
  2. Datatype definition, using datadef. This takes three components:
    1. The name of the datatype
    2. A list of generic types which are introduced in the datatype. If the datatype is not generic, an empty list ([]) must be used.
    3. The constructors for the datatype, which are permitted to be recursive.
  3. Clause definition, using clausedef. While clause definitions may be located anywhere in the file, it is good practice to put them immediately before their corresponding clauses, as is done in the example. This takes three components:
    1. The name of the clause
    2. Any type variables the clause introduces. If the clause does not introduce any type variables, then an empty list ([]) must be used.
    3. The types of the parameters to the clause, specified in a list.
  4. The relation([...]) type, which is used for higher-order clauses. Note the use of square braces in the type. Values of type relation([...]) are called with the call built-in, which behaves substantially differently from the typical call metaprogramming routine. Uses of call here ultimately are translated down into typical first-order clauses.
  5. The lambda keyword, which introduces values of type relation([...]). These consist of two components:
    1. A parameter list, which is no more constrained than the parameter list for clauses. That is, unification is employed for the parameters, and they need not be variables (though they may be, if desired). Unlike with clausedef, the types of the parameters are inferred, and there is no way to explicitly annotate parameter types here.
    2. The body, which is a snippet of executable code. If multiple conjuncts are needed, they should be enclosed in an outer set of parentheses (as in, the body should be a functor with the name ,).
  6. The int type, which is short for integer. This behaves with is as one might expect.
  7. The list(...) type, which is for the typical lists built into Prolog. These are referenced with the usual notation (e.g., [H|T])

All procedures must have a clausedef associated with them.

About

A basic type system on top of Prolog, along with higher-order clauses. Translates to normal Prolog.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages