Kilo LISP Manual

By Nils M Holm, 2019

0 Quick Summary

For the impatient LISP user:

Kilo LISP is a purely symbolic, lexically scoped LISP system
with tail call elimination, macros, and image files. The only
data types are the atom (symbol) and the pair/list. T is true
and NIL is false. The following special forms exist:


The following functions are predefined:



1 S-Expressions

A Symbol is any combination of these characters:

a b c d e f g h i j k l m n o p q r s t u v w x y z
0 1 2 3 4 5 6 7 8 9 0

Other characters can be included by prefixing them with a /.

An Atom is either a symbol or NIL.

A Pair is an ordered pair of two S-expressions:

(car/-part . cdr/-part)

The spaces before and after the dot are optional, i.e.

(a . b)  =  (a.b)

A pair may contain other pairs:

((a . b) . c)
(a . (b . c))
((a . b) (c . d))

An S-Expression is either an atom or a pair.

A nested pair of the form

(a . (b . c))   may be written as   (a b . c)


(a . nil)       may be written as   (a)

The rules apply inductively, so

(a . (b . (c . nil)))  =  (a b c)

1.1 Lists

A list is an S-expression that is either NIL or a pair whose
cdr part is a list:

List  =  NIL  or  (S-expression . List)

These are Lists:

(foo bar baz)
((a . b) foo (nested list))

A list whose innermost/rightmost cdr part is a symbol is called
a Dotted List. These are dotted lists:

(a b . c)
((foo bar) . baz)

The pair is a two-element dotted list.

Lists of single-character symbols may be abbreviated as follows:

'(a b c)  =  #abc

Abbreviated lists are self-quoting. See QUOTE.

An Association List (or Alist) is a list of pairs (associations),
where the car part of each association is called the Key and the
cdr part the value of the association:

((<key1> . <value1>) ... (<keyN> . <valueN>))

See ASSOC for details.


A comment may be inserted anywhere (even inside of a form, but
not inside of an atom) by including a semicolon (;). Comments
extend to the end of the current line.


(setq foo  ; this is a comment
  (quote bar))

1.3 Unreadable Forms

A form that contains characters other than those listed in the
previous sections are unreadable, i.e. they cannot be read by
the LISP parser.

Unreadable forms are sometimes output by the system to represent
data that have no unambiguous textual representation.

2 Expressions

An Expression is an S-expression with a meaning.

x => y     denotes that x evaluates to y;
           y is the Value of x.

undefined  denotes an undefined value.

There are four kinds of expressions:

Lambda Abstraction
Function Application
Special Forms

Strictly speaking, lambda abstraction is also a special form,
but it will be explained separately.

2.1 Variables

A Variable is a symbol that is bound to a Location. A location
is a container that contains a value.

Variables evaluate to the value that is stored in their location:

Variable  =>  Value

References to unbound symbols (symbols that are not bound to
any location) are undefined:

unbound/-symbol  =>  undefined

The term "a variable V is bound to the value X" is short for
"a variable V is bound to a location holding the value X".

A Constant is a variable that is bound to its own name:

constant  =>  constant

For instance, the canonical "true" value, t, is a constant:

t  =>  t

Variables are created by the SETQ special form or by lambda

2.2 Lambda Abstraction

(Lambda <formals> <body>)

is a Lambda Abstraction. Its value is a Function.

Names enclosed in angle brackets are Meta-Syntactic Variables.
They are not part of LISP, but describe parts of special forms.
They may represent single expressions or sequences of

<Formals> is a list of variables (a.k.a. Formal Arguments) of
that function and <body> is the Term or Body of that function.

<Formals> may be

- a variable
- a list of variables
- a dotted list of variables

<Body> must be at least one expression.

Examples: (lambda (x) x)
          (lambda (f x y) (f (f x) (f y)))
          (lambda () (print (quote hello/ world)))
          (lambda x (f x) x)

The variables of a function are created by the function by
binding their names to locations containing the (actual)
arguments of the function. This happens when the function is
being applied to some arguments.

2.3 Function Application

A variable X is Bound in an expression E, if

- E is a lambda function
- X is a variable of E


X  is     bound in  (lambda (x) (f x))
F  is not bound in  (lambda (x) (f x))

When a variable X is not bound in an expression E,
X is Free in E; X is a free variable of E.

A function is Applied to (Actual) Arguments using the syntax

(function argument ...)

The following steps are involved in the application:

(1) each expression in the application is evaluated
(2) a location is created for each variable of the function
(3) each argument is stored in the corresponding location
(4) the body of the function is evaluated
(5) the value of the body is returned

Because of (1), functions may be bound to variables. In step (1),
a variable bound to a function will be replaced by that function.

Variables are associated with arguments by position, i.e. the
first argument is stored in the first variable, etc.

For instance:

((lambda (x y) (cons x y)) (quote 1) (quote 2))

; evaluate arguments:  (quote 1) => 1,  (quote 2) => 2
((lambda (x y) (cons x y)) 1 2)

; create variables and store 1 in X and 2 in Y, giving the body
(cons 1 2)

; evaluate the body, giving the value
(1 . 2)

2.4 Optional Arguments

When the list of formal arguments is a proper (NIL-terminated)
list, there has to be exactly one argument per variable:

((lambda (x) x))         =>  undefined
((lambda (x) x) 'a)      =>  A
((lambda (x) x) 'a 'b))  =>  undefined

When the argument list of a function is a dotted list, the
variables before the dot bind like fixed arguments, above, and
the variable after the dot binds to a list containing any excess
arguments that may be passed to the function:

((lambda (x . y) y))            =>  undefined
((lambda (x . y) y) 'a)         =>  nil
((lambda (x . y) y) 'a 'b))     =>  (b)
((lambda (x . y) y) 'a 'b 'c))  =>  (b c)

When there are no fixed arguments, the formal argument list is
replaced with a single variable that binds to a list of all

((lambda x x))         =>  nil
((lambda x x) 'a)      =>  (a)
((lambda x x) 'a 'b))  =>  (a b)

3 Special Forms

A Special Form is an expression of the form

(<keyword> <argument> ...)

The syntax is the same as in function application, but there
are some differences: When a syntactic Keyword is found in the
first slot of a list, neither the keyword nor any <argument>s
will be evaluated. The <argument>s will then be processed by
the LISP system according to the semantics of the individual
special form. These semantics may involve the evaluation of
some of the <argument>s.

A Keyword is sometimes also called a Special Operator. The
special operators of the Kilo LISP language are


3.1 (LAMBDA <formals> EXPR1 EXPR2 ...)  =>  FUN

Create a new function with the formal arguments <formals> and
the body (PROG EXPR1 EXPR2 ...). The PROG special form is
implied, i.e.

(lambda (x) a b c)

will be interpreted as

(lambda (x) (prog a b c))

<Formals> may be a list, a dotted list, or a single variable.
See Function Application, above.

Examples: (lambda (x) x)             ; identity function
          (lambda (x) (car (cdr x))) ; CADR function
          (lambda x x)               ; LIST function


Apply FUN to the given LIST of arguments and return the value
delivered by the function. Basically, APPLY does

X = (cons fun list)

and then evaluates X, but *without* evaluating the arguments in
LIST again.

FUN may be a built-in function, or a lambda function, but not a
special operator (like APPLY itself).

Example: (apply cons '(1 2))  =>  (1 . 2)

3.3 (QUOTE EXPR)  =>  EXPR

Return EXPR unevaluated.

The form  'x  is an abbreviation for (quote x).

Examples: (quote foo)      =>  foo
          (quote (car x))  =>  (car x)
          '(foo bar)       =>  (foo bar)

3.4 (IF EXPR1 EXPR2 EXPR3)  =>  EXPR

Evaluate EXPR1. When its value is not NIL, evaluate EXPR2 and
otherwise evaluate EXPR3. Return the value of the expression
evaluated last.

EXPR1 is called the Predicate of the form.

Examples: (if t 'yes 'no)       =>  yes
          (if nil 'yes 'no)     =>  no
          (if (atom 'x) 'a 'b)  =>  a


Evaluate EXPR1. When its value is not NIL, return EXPR1.
Otherwise evaluate and return EXPR2. In either case EXPR1
is only evaluated once.

Examples: (ifnot 'foo 'bar)  =>  foo
          (ifnot nil 'bar)   =>  bar

3.6 (SETQ <var> EXPR)  =>  <var>

Evaluate EXPR and store its value in the location bound to the
variable <var>. Return <var>. When <var> is not bound to any
location, a fresh location will be created and <var> will be
bound to it.

Example: (setq foo 'bar)  =>  foo
         foo              =>  bar

3.7 (PROG EXPR ...)  =>  EXPR

Evaluate all given expressions from the left to the right and
return the value of the rightmost expression. The special form
(prog) evaluates to NIL.

Examples: (prog '1 '2 '3)      =>  3

          (prog (print 'foo)   ; print foo and then print bar
                (print 'bar))  =>  bar

4 Derived Special Forms

A Derived Special Form is a special form that is being
transformed by the LISP system to an expression that uses only
function application and the (primitive) special forms listed
in the previous chapter.

A function transforming a derived special form to a primitive
expression is called a Macro.

A macro is defined using the SETQ, MACRO, and LAMBDA special

(setq <var>
    (lambda ...)))

The transformation of a derived special form is called Macro
Expansion. It is part of the evaluation process and happens
after reading but before interpreting an expression. The
transformation is performed by passing the arguments of the
derived special form to the macro and then replacing the derived
special form by the result of the macro.

The result of macro expansion can contain more derived special
forms, which will also be expanded.


(setq kwote
    (lambda (x)
      (list 'quote x))))

This macro will expand the derived special form (kwote foo) to
the expression (quote foo), no matter what FOO is. The result
of the form is FOO.

(setq listq
    (lambda x
      (if (null x)
          (list 'cons
                (list 'quote (car x))
                (cons 'listq (cdr x)))))))

This macro will expand the derived special form

(listq a b c)

to           (cons (quote a)
                   (listq b c))

then to      (cons (quote a)
                   (cons (quote b)
                         (listq c)))

then to      (cons (quote a)
                   (cons (quote b)
                         (cons (quote c)

and then to  (cons (quote a)
                   (cons (quote b)
                         (cons (quote c) nil)))

which no longer contains any derived special forms. The last
expression in the process is then interpreted, giving the
result (a b c).

The LISTQ macro can be implemented more comprehensibly using
quasiquotation - see QQUOTE at the end of this chapter.

4.1 (GENSYM)  =>  SYMBOL

Generate a unique symbol name.

Rationale: When derived special forms expand to expressions
containing local variables, these variables should be named
using GENSYM. Otherwise name capturing can occur, as the
following example illustrates:

(setq swap
    (lambda (x y)
      @(let ((L ,x))
         (setq ,x ,y)
         (setq ,y L)))))

It is supposed to swap the values of X and Y, but the local name
L is captured when expanding (swap L M), which does nothing:

(let ((L L)) (setq L M) (setq M L))

(It creates a local L and initializes it with the outer L, then
sets the local L to M and then M to the local L.)

The proper way to implement the SWAP macro would be:

(setq swap
    (lambda (x y)
      (let ((g (gensym)))
        @(let ((,g ,x))
           (setq ,x ,y)
           (setq ,y ,g))))))

Example: (list (gensym)
               (gensym))  =>  (G1 G2 G3)

4.2 (LET <bindings> <body>)  =>  EXPR

<Bindings> has the form ((<var1> EXPR1) ... (<varN> EXPRn)) and
<body> is an implied PROG form.

Create the variables <var1> through <varn>, then evaluate the
expressions in no specific order and bind their values to the
corresponding variables. Finally evaluate <body> and return its

Examples: (let ((a '1) (b '2))
            (cons a b))             =>  (1 . 2)

          (let ((a 'foo)) '1 '2 a)  =>  foo

          (let ((a '1))
            (let ((a (cons a '2)))
              a))                   =>  (1 . 2)

4.3 (LABELS <bindings> <body>)  =>  EXPR

<Bindings> has the form ((<var1> EXPR1) ... (<varN> EXPRn)) and
<body> is an implied PROG form.

Create the variables <var1> through <varN> and then evaluate
the expressions, from left to right, and store their values in
the associated variables. Evaluate <body> and return its value.

When the EXPRs are lambda special forms, then each EXPR may
refer to each VAR, thereby allowing to create mutually recursive
functions. In addition, each EXPRi can refer to any <varJ>,
without any further restrictions, as long as J<i.

Examples: (labels ((e (lambda (x)
                        (if (null x) t   (o (cdr x)))))
                   (o (lambda (x)
                        (if (null x) nil (e (cdr x))))))
            (e #1234))                                    =>  t

          (labels ((complement (lambda (p)
                                 (lambda (x)
                                   (not (p x)))))
                   (pair (complement atom)))
            (pair '(1 . 2)))                              =>  t

4.4 (AND EXPR ...)  =>  EXPR

Evaluate the given expressions, from left to right, until one
expression evaluates to NIL. In this case return NIL. When none
of the EXPRs evaluates to NIL, return the value of the last
expression. When no arguments are given, return T.

This special form implements the short-circuit logical AND.

Examples: (and)           =>  t
          (and '1 '2 '3)  =>  3
          (and nil 'foo)  =>  nil

4.5 (OR EXPR ...)  =>  EXPR

Evaluate the given expressions, from left to right, until one
expression evaluates to a value other than NIL. In this case
return the value. When none of the EXPRs evaluates to non-NIL,
return NIL. When no arguments are given, return NIL.

This special form implements the short-circuit logical OR.

Examples: (or)           =>  nil
          (or '1 '2 '3)  =>  1
          (or nil 'foo)  =>  foo
          (or nil nil)   =>  nil

4.6 (COND <clause> ...)  =>  EXPR

Each <clause> has either of these two forms:

(EXPR1 EXPR2 ...)
(ELSE EXPR2 ...)

COND proceeds as follows:

The predicate EXPR1 of the first clause is evaluated. When its
value is not NIL, there are two cases to distinguish:

(1) When the clause contains multiple expressions, the
expressions EXPR2 ... will be evaluated from the left to the
right and the value of the last expression will be returned.
No other clauses will be examined.

(2) When the clause contains only the predicate EXPR1, the value
of the predicate will be returned. No other clauses will be

When the predicate of the first clause is NIL, the predicate of
the next clause will be evaluated. Evaluation of COND ends when
either a clause with a true predicate is found or no clauses
are left to evaluate.

When there are no clauses left to evaluate, NIL is returned.

When a clause has a predicate of the form ELSE, then the
predicate will be assumed to be true *and* the clause must be
the last clause in the COND form.

Examples: (cond (t 'first) (t 'second))    =>  first
          (cond (nil 'a) (nil 'b) (t 'c))  =>  c
          (cond (nil 'a) (else 'b))        =>  b
          (cond (nil 'foo))                =>  nil
          (cond ((cons '1 '2)))            =>  (1 . 2)
          (cond)                           =>  nil

4.7 (LOOP <name> <bindings> <body>)  =>  EXPR

<Bindings> has the form ((<var1> EXPR1) ... (<varN> EXPRn)) and
<body> is an implied PROG form.

<Name> is a variable that will be bound to the function

(lambda (<var1> ... <varN>) <body>)

and this function will then be applied to the arguments EXPR1
through EXPRn. Hence applying <name> to new arguments inside of
<body> will re-enter the loop with new values bound to the
given variables.

Examples: (loop next ((a '(3 2 1)))   ; print (3 2 1) (2 1) (1)
            (cond ((null a))
                  (else (prin a)
                        (next (cdr a)))))  =>  t

          (loop next () (next))       ; infinite loop

4.8 (QQUOTE <template>)  =>  EXPR

Create an S-expression from a Quasiquotation template.

Quasiquotation is like quotation, but allows to "unquote" parts
of the quoted S-expression, thereby inserting dynamic elements
into an otherwise static template.

The following abbreviations can be used:

@expr   =   (qquote expr)
,expr   =   (unquote expr)
,@expr  =   (splice expr)

(qquote template) quasiquotes an S-expression.
(unquote expr) unquotes an S-expression contained in a template.
(splice expr) is like UNQUOTE, but splices the expression.


    @a    =  'a
   @,a    =  a
  @(a b)  =  (cons 'a (cons 'b nil))
 @(,a b)  =  (cons a (cons 'b nil))
@(,@a b)  =  (conc a (cons 'b nil))

More intuitively, QQUOTE creates a (mostly nested) list
structure and UNQUOTE replaces an element in that structure
with the value of its argument. SPLICE replaces an element
with multiple elements passed to it as a list. For instance:

@(a ,(list 'b 'c) d)  =>  (a (b c) d)


@(a ,@(list 'b 'c) d)  =>  (a b c d)

At the *end* of a list, a dot followed by UNQUOTE can be used
in the place of SPLICE, thereby saving one application of CONC:

(x ,@y)  =  (x . ,y)

Quasiquotation is mostly used to create expressions in macros.
For instance, the KWOTE macro from the beginning of section 4
can be written as follows using QQUOTE:

(setq kwote
    (lambda (x)
      @(quote ,x))))

and the LISTQ example from the same section can be written as

(setq listq
    (lambda x
      (if (null x)
          @(cons (quote ,(car x))
                 (listq . ,(cdr x)))))))

5 List Functions

5.1 (CONS EXPR1 EXPR1)  =>  EXPR

Create a fresh pair with EXPR1 as its car part and EXPR2 as its
cdr part.

Examples: (cons '1 '2)      =>  (1 . 2)
          (cons '1 '(2 3))  =>  (1 2 3)      ; (1 . (2 3))
          (cons '(1 2) '3)  =>  ((1 2) . 3)

5.2 (CAR PAIR)  =>  EXPR
    (CDR PAIR)  =>  EXPR

CAR extracts the CAR part and CDR the cdr part of a pair.

Examples: (car '(1 . 2))  =>  1
          (car '(1))      =>  1
          (car '(1 2 3))  =>  1
          (cdr '(1 . 2))  =>  2
          (cdr '(1))      =>  nil
          (cdr '(1 2 3))  =>  (2 3)

5.3 (CAAR PAIR)   =>  EXPR
    (CDDDR PAIR)  =>  EXPR

These functions extract elements from nested pairs. These
elements are extracted by the C..R functions:

((caar . cdar) . (cadr . cddr))

And these elements by C...R:

 (  ((caaar . cdaar) . (cadar . cddar))
  . ((caadr . cdadr) . (caddr . cdddr))  )

I.e, CAAR extracts the car field of a car field, CADR the car
field of a cdr field, etc. The most commonly used variants are

CADR   the second element of a list
CDDR   the tail of a list after the second element
CADDR  the third element of a list
CDDDR  the tail of a list after the third element

Examples: (cadr '(1 2 3 4))  =>  2
          (cddr '(1 2 3 4))  =>  (3 4)

          (cadar '((lambda (x) x) 'foo))  =>  (x)


SETCAR replaces the car part and SETCDR replaces the cdr part
of the given pair with EXPR. The pair will be modified! Both
functions return the modified pair.

Examples: (setcar '(foo . bar) '0)  =>  (0 . bar)
          (setcdr '(foo . bar) '1)  =>  (foo . 1)

          (let ((x '(1)))      ; infinite
            (setcdr x x))  =>  (1 1 1 1 1 1 ...)

5.5 (LIST EXPR ...)  =>  LIST

Create a fresh list containing the values of the given

Examples: (list)              =>  nil
          (list '1 '2 '3)     =>  (1 2 3)

          (list (cons '1 '2)
                (atom nil)
                (null 'x))    =>  ((1 . 2) t nil)


Create a fresh list containing the elements of LIST in reverse
order and concatenate that list to EXPR.

NRECONC is similar to RECONC, but reverses and concatenates its
arguments destructively, so the original argument values will be

Examples: (reconc '(1 2 3) '(a b c))  =>  (3 2 1 a b c)
          (reconc '(1 2 3) '(a . b))  =>  (3 2 1 a . b)
          (reconc '(1 2 3) 'a))       =>  (3 2 1 . a)
          (reconc '(1 2 3) nil)       =>  (3 2 1)
          (reconc nil 'a)             =>  a
          (reconc nil '(a b c))       =>  (a b c)
          (reconc nil nil)            =>  nil

5.7 (CONC LIST ...)   =>  LIST
    (NCONC LIST ...)  =>  LIST

Create a fresh list containing the elements of all LIST
arguments in the given order. All LISTs must be NIL-terminated.

NCONC is similar to CONC, but concatenates its arguments
destructively, so the original argument values will be lost.

Examples: (conc)                      =>  nil
          (conc '(a b) '(c d))        =>  (a b c d)
          (conc '(a) '(b) '(c) '(d))  =>  (a b c d)
          (conc '(a) nil nil '(b))    =>  (a b)

5.8 (REVER LIST)   =>  LIST

Create a fresh list containing the elements of LIST in reverse
order. The list must be a NIL-terminated list.

NREVER is similar to REVER, but reverses its argument
destructively, so the original argument value will be lost.

Examples: (rever '(1 2 3))  =>  (3 2 1)
          (rever nil)       =>  nil


Find the given EXPR in the given LIST. Return the tail of LIST
that starts with the first occurrence of EXPR in LIST. When the
expression is not contained in the list, return NIL. EXPR must
be an atom.

Examples: (memb 'c '(a b c d e f))  =>  (c d e f)
          (memb 'x '(a b c d e f))  =>  nil


Find an association with the key EXPR in ALIST and return it.
When no association in ALIST has EXPR as its key, return NIL.
EXPR must be an atom.

Examples: (assoc 'b '((a.1) (b.2) (c.3))  =>  (b . 2)
          (assoc 'x '((a.1) (b.2) (c.3))  =>  nil

6 Higher-Order Functions

6.1 (MAP FUN LIST1)        =>  LIST

When one list is supplied, apply FUN to each of its elements,
returning a fresh list containing the values of the function
applications. I.e.:

(map f (a1 ... aN))  equals  (list (f a1) ... (f aN))

When two lists are given, FUN is applied to elements of the
lists pairwise:

        (map f (a1 ... aN) (b1 ... bN))
equals  (list (f a1 b1) ... (f aN bN))

Examples: (map atom '(a (b) nil))       =>  (t nil t)
          (map cons '(a b c) '(1 2 3))  =>  ((a.1) (b.2) (c.3))

7 Predicates

A Predicate is a function that returns T or NIL.

7.1 (ATOM EXPR)  =>  T or NIL

Return T, if EXPR is an atom, i.e. NIL or a symbol.

Examples: (atom 'foo)  =>  t
          (atom nil)   =>  t
          (atom '(x))  =>  nil

7.2 (NULL EXPR)  =>  T or NIL
    (NOT EXPR)   =>  T or NIL

Return T, if EXPR is NIL. Both predicates are the same function,
the difference between them is intensional: NULL tests whether
is argument is the empty list and NOT negates the truth value of
its argument.

Examples: (null nil)       =>  t
          (not (atom 'x))  =>  nil

7.3 (EQ EXPR1 EXPR2)  =>  T or NIL

Return T, if EXPR1 and EXPR2 are the same object. This is the
case, if EXPR1 and EXPR2 are both

- the same symbol
- bound to the same variable

In particular,

(EQ (cons x nil) (cons x NIL))     =>  nil


(let ((y (cons x nil))) (eq y y))  =>  t

for any S-expression X.

Examples: (eq 'foo 'foo)  =>  t
          (eq 'foo 'bar)  =>  nil
          (eq 'foo nil)   =>  nil
          (eq t t)        =>  t

7.4 (EQUAL EXPR1 EXPR2)  =>  T or NIL

Return T, if EXPR1 and EXPR2 are structurally equal. This is the
case, if 

- (eq EXPR1 EXPR) => t
- (equal (car EXPR1) (car EXPR2))
  (equal (cdr EXPR1) (cdr EXPR2))

(Eq a b) implies (equal a b), but the converse is not necessarily

Informally, two  S-expressions are EQUAL, if PRIN emits the same
output for them.

Examples: (equal '(a (b) c) '(a (b) c))  =>  t
          (equal '(a (b) c) '(a (X) c))  =>  nil

7.5 (EOFP EXPR)  =>  T or NIL

Return T, if EXPR signals the end of input, as returned by the
READ function.

Example: (eofp (read)) %  =>  t

8 Program Loading

8.1 (LOAD SYMBOL)  =>  T

Read and evaluate expressions from the file SYMBOL. The input
from the file will be processed as if typed in at the LISP
prompt, but without printing any values.

LOAD can in principle be nested (files being loaded may contain
LOAD expressions), but when nesting is too deep, the system
stack may overflow, causing unpredictable results. It is hence
recommended not to use LOAD in files.

Example: (load 'program)  =>  t


Write a core image to the file SYMBOL. The file will contain the
state of the LISP system at REPL level, so it can be reinstated
later by loading the image. (It is not like the MACLISP SUSPEND
function, which captures the *entire* state of the system).

An image is loaded by passing its name to the LISP system when
starting it (e.g., as a command line argument).

The default image file name is "klisp".

Example:  (suspend 'klisp)  =>  t

9 Input and Output

These functions translate between the Internal and External
Representations of S-expressions.

The external representation of an S-expressions is what a user
types in at the LISP prompt in order to create an LISP object,
like a symbol or a list. The internal representation of that
object is what the LISP system uses internally to store the
object -- typically a set of pointers or indexes and small

9.1 (READ)  =>  EXPR

Read the external representation of one object from the current
input source, convert it to internal representation and return

Note that the value returned by READ on the REPL will then be
printed by PRINT (see below), which converts it to external
representation again. Hence the internal representation of an
object will never be visible to the programmer.

READ will skip over leading white space characters and comments
while reading input. It will only return when a complete
S-expression has been read. When reading lists or pairs, the
input may span multiple lines.

When no input is available on the current input source, READ
will return a special object call the End of Transmission Marker
(or EOT marker). The EOT marker will print as *eot* on the LISP

An EOT marker can normally generated using the keyboard, but the
exact keys depend on the operating environment. On a Unix system
one would press "Control" and "d" (control-D). Kilo LISP will
also interpret the "%" character as an EOT marker.

Examples: (read) foo  =>  foo

          (read) (a
                  c)  =>  (a b c)

          (read) %    =>  *eot*

9.2 (PRIN EXPR)   =>  EXPR
    (PRIN1 EXPR)  =>  EXPR
    (PRINT EXPR)  =>  EXPR

All of these functions print the external representation of the
given S-expression on the current output device. They return the
object being printed.

PRIN1 just prints EXPR.
PRIN  prints a trailing blank after EXPR.
PRINT advances to the next line after printing EXPR.

Example: (prin 'foo)  =>  foo  ; prints "foo "

10 Misc. Functions


Signal an error by printing a message and aborting the
computation in progress. The message will consists of a question
mark and the given SYMBOL. When a second argument is given, it
will print after the SYMBOL, separated from it by a colon. E.g.

(error 'don/'t/ do/ this/!)   will print   ? don't do this!

(error 'wrong 'foo)           will print   ? wrong: foo

The ERROR function delivers no value, because it never returns.

Example: (if (atom x)
             (error 'need/ a/ list x)
             (do/-something x))

10.2 (GC)  =>  SYMBOL

Perform a garbage collection and return an uninterned symbol
representing the number of free nodes in the system.

Note that garbage collections are triggered automatically. This
function merely informs the programmer about memory usage.

Because the returned symbol is "uninterned" (it cannot be used
to form a variable), its name will be recycled when it is no
longer used.

Example: (gc)  =>  6810   ; actual number may differ

11 The REPL

The Read Eval Print Loop (REPL) is the interface to the LISP
system. It prints a prompt (*), reads an expressions via READ,
evaluates it, prints its value using PRINT, and starts over.

While the REPL is parsing input, partially read input can be
abandoned by pressing the "interrupt" key followed by ENTER.

The "interrupt" key depends on the operating environment. On a
Unix system it is typically the combination control-C or the
key labeled "DEL".

Once a computation is in progress, it can be aborted by pressing
the "interrupt" key.

The value returned by a computation will be bound to the
variable IT, so it can be used in the expression typed next:

* (cons '1 '(2 3))
(1 2 3)
* (cons '0 it)
(0 1 2 3)

To end the LISP session, send an EOT marker to the system.

contact  |  privacy