Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Lam is a small language based on the λ-calculus.


Install with cabal install. Make sure ~/.cabal/bin is in your PATH, and then type lam to start the Lam REPL. Alternatively, run cabal build and then ./dist/build/lam/lam.

  • Enter an expression to see it evaluated and printed.
  • Enter an assignment to bind the name on the left-hand side to the result of evaluating the right-hand side. (The result will also be printed.)
  • If you didn't assign the previous result to a name, you can still access it with the % symbol.
  • Enter load foo.lam to evaluate the file foo.lam. It should only contain assignments since expressions will not be printed. You can also pass filenames as command line arguments to lam and they will be loaded on startup.
  • When writing Lam in a file, if you need to continue an expression on another line, indent the continuation lines by at least one space or tab.
  • You should probably load prelude.lam from this repository, since most of Lam itself (the language) is defined there.
  • Enter exit, quit, or use Ctrl-C or Ctrl-D when you're finished.


  • Takes minimalism to the extreme.
  • Uses applicative order evaluation with lazy special form If for recursion.
  • Parser error messages are easy to understand and useful.
  • Command-line history (up/down arrow keys), navigation (left/right arrow keys), and other shortcuts such as Ctrl-A and Ctrl-E work as expected.
  • Loading files is easy with tab completion.
  • Church encoding of common data structures is built in.
  • Non-feature: Lam is an extremely inefficient means of computation. That being said, you should hardly ever notice delays unless you are using Lam to crunch numbers.


Comments are ignored by Lam. They begin with ; and continue until the end of the line. Blank lines and lines containing only whitespace are also ignored.

Tokens are the atomic units of Lam. A token is the only thing that cannot be reduced any further. A token is either one lowercase letter from a to z, or a string of characters not beginning with a lowercase letter. Specifically, the string can include any characters except whitespace and ()\.=. For example, Abc is a single token while abc is actually three tokens side by side.

An assignment is of the form x = e, where x is a token that cannot contain any digit characters and e is an expression. The assignment binds x to the result of evaluating e.


There are three kinds of expressions:

  • a symbol, defined by a token;
  • a function, defined by a token and an expression;
  • an application, defined by two expressions.

A token by itself is interpreted as a symbol. If that token has been bound to an expression previously, the symbol will evaluate to that expression. Otherwise, the symbol evaluates to itself (a terminal symbol). For example, if you enter a = b, then after that a will evaluate to b (whereas before it would just remain a). Evaluation continues as long as possible, so if you enter b = c next, then a will evaluate to c. It is impossible to create a loop in this way because a = a is a no-op (and if you enter c = a, the a will first evaluate to c, so it is the same as c = c).

A function is an expression of the form \x.e, where x is a token and e is any expression. The function binds the variable x, so any occurrences of x in e refer to that bound variable. When a function is evaluated by itself, its body is reduced as much as possible.

An application is an expression of the form f a, where f and a are expressions. If f evaluates to a function, then the application is evaluated by substituting a for all occurrences of the formal parameter in the body expression of f. Otherwise, the expression remains an application object, with both terms reduced as much as possible.


  • Adding extra parentheses is always allowed: (((a))) is the same as a.
  • Applications are left-associative, so abc is equivalent to ((a b) c).
  • Function bodies extend as far to the right as possible, so \x.x x is the same as \x.(x x), while explicit parentheses are needed for (\x.x) x.
  • The application f \x.x is allowed, but be careful: f \x.x \x.x is the same as f \x.(x \x.x), so you can only drop the parentheses from a function argument when it occurs at the end.
  • The form \xyz.e is syntactic sugar for the curried function \x.\y.\z.e.


Lam uses applicative order evaluation. The only exception to this is the special form If. In the expression If c a b, the expressions a and b are never evaluated until c is a concrete Boolean value. Recursive functions must use If to work properly, otherwise evaluation will never end.

Lam does three basic types of reduction:

  • ɑ-conversion renames the formal parameter of a function (including its occurrences in the function body). This is sometimes necessary to avoid unintentionally capturing free variables. For example, \x.x is ɑ-equivalent to \y.y.

  • β-reduction is the function application: the substitution of an expression for all occurrences of the formal parameter in the body of the function.

  • η-conversion converts \x.f x to f whenever x does not occur as a free variable in f.

Church encoding

Assuming the Prelude is loaded, True evaluates to \xy.x and False evaluates to \xy.y. Tokens consisting only of digits are handled specially by the interpreter; they are converted to Church numerals before evaluation. For example, 0 becomes \fx.x and 2 becomes \fx.f (f x).

There are special functions that convert encoded data structures back into human-readable strings. The application ? b evaluates to True if b represents logical truth and False for logical falsity. If b is not a Church boolean, the application is not reduced. Similarly, the application # n converts n into a string of digits if n can be interpreted as a Church numeral.


The Lam Prelude defines the core of the language:

  • Logic: True, False, And, Or, Not
  • Arithmetic: Zero, Succ, Pred, Add, Sub, Mult, Div, Quot, Pow
  • Comparison: Zero?, Eq?, LT?, GT?, LTEq?, GTEq?, Max, Min
  • Lists: Cons, Car, Cdr, Nil, Nil?, Take, Drop, Map, Filter, Range
  • Other: Id, Comp


Lam uses Parsec for parsing and Haskeline for the command-line interface.


© 2015 Mitchell Kember

Lam is available under the MIT License; see LICENSE for details.


Yet another lambda calculus language.







No releases published


No packages published