Skip to content

burz/Feval

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

71 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Feval - evaluation using f-algebras

Author: Anthony Burzillo


Feval is a statically typed functional programming language that uses f-algebras as opposed to classic recursion to solve the problem of evaluation and typechecking, which allows the compiler to perform better optimizations.

The overall language of Feval is EF, which is an extension of the smaller language F. In order to run a program written in EF, we first use an f-algebra to transform the AST to F, and then we transform the result (via an f-algebra) into the individual AST's used by the evaluator's f-algebra and the typechecker's f-algebra. The reason we need to use seperate f-algebras is because we need to stall the processing of certain subtrees of the AST in order to ensure correct evaluation and typechecking. For instance, we cannot evaluate the expression of a anonymous function until we have obtained its argument. Similarly, we cannot typecheck a function until we have assigned a type hypothesis to the variable.

For more information on how we solve these problems, check out the article Feval: F-Algebras for expression evaluation. For an in-depth explanation of parsing see Feval: Parsing a functional language with Parsec

Usage

To build Feval run cabal configure && cabal build && cabal install. Then you can run Feval which acts as a REPL:

$ ./feval
Function x -> x && True
  => Function x -> x && True
    : Bool -> Bool
(Function x -> Function y -> x + y / 50) 5
  => Function y -> 5 + y / 50
    : Int -> Int
Let f x = If x = 0 Then 1 Else x * f (x - 1) In f 6
  => 720
    : Int
Case [1, 2, 3, 4] Of [] -> 0 | (x : xs) -> x + 6
  => 7
    : Int
Function x -> Case True : x Of [] -> True | (y : ys) -> True || !(y || False)
  => Function x -> Case True Of [] -> True | (y, ys) -> True || !(y || False)
    : [Bool] -> Bool

To quit simply press ctrl-d.

Since the REPL can only handle one line expressions, we also allow feval to take a file to execute as an argument. For instance, with the provided file mergesort.fvl:

$ feval mergesort.fvl
  => [-34, 3, 4, 23]
    : [Int]

Expressions

Boolean Operations

We allow conjunction (&&), disjunction (||), and negation (!) expressions.

Integer Operations

The operations of addition (+), subtraction (-), multiplication (*), division (/), and modulus (%) evaluate to integer values. On the other hand, the comparison operators of equality (=), less-than (<), less-than-or-equal (<=), greater-than (>), and greater-than-or-equal (>=) all evaluate to boolean values.

Functional Operations

We allow the creation of anonymous functions via Function x -> e where e is some expression, and x is the argument to the function. We can create multiple argument anonymous functions via Function x -> Function y -> e where x is the first argument and y is the second, etc.

To apply a function f simply use f e where e is the expression for the first argument, or f e1 e2 for the first argument e1 and second argument e2.

If Expressions

Use If e1 Then e2 Else e3.

Let Expressions

We allow let expressions to define constants and functions (possibly recursive) via

Let x = 4 In x + 54

and

Let f x y = If x = 0 Then 0 Else y + f (x - 1) y In f 3 4

Semi-colon Expressions

An expression of the form e1; e2 first evaluates e1 then e2 and returns the result of e2.

List Expressions

You can create empty lists [] or lists with values [1, 2, 3, 4]. You can cons values onto a list via

5 : [4, 5, 6]
  => [5, 4, 5, 6] 
    : [Int]

Finally, we can match on lists via a case expression via

Case e1 Of [] -> e2 | (x : xs) -> e3

Releases

No releases published

Packages

No packages published