Subsections

Forms and Evaluation

Atoms

A data object other than a cons is always an atom, no matter what complex structure it may have. Note that NIL, which is sometimes noted as () to represent an empty list, is also an atom. Every atom except a symbol is always evaluated to itself, although quoting is required in some other Common Lisp implementations.

Scoping

Every symbol may have associated value. A symbol is evaluated to its value determined in the current binding context. There are two kinds of variable bindings; the lexical or static binding and the special or dynamic binding. Lexically bound variables are introduced by lambda form or let and let* special forms unless they are declared special. Lexical binding can be nested and the only one binding which is introduced innermost level is visible, hiding outer lexical bindings and the special binding. Special variables are used in two ways: one is for global variables, and the other is for dynamically scoped local variables which are visible even at the outside of the lexical scope as long as the binding is in effect. In the latter case, special variables are needed to be declared special. The declaration is recognized not only by the compiler, but also by the interpreter. According to the Common Lisp's terms, special variables are said to have indefinite scope and dynamic extent.

Even if there exists a lexical variable in a certain scope, the same variable name can be redeclared to be special in inner scope. Function symbol-value can be used to retrieve the special values regardless to the lexical scopes. Note that set function works only for special variable, i.e. it cannot be used to change the value of lambda or let variables unless they are declared special.

(let ((x 1))
   (declare (special x))
   (let* ((x (+ x x)) (y x))
      (let* ((y (+ y y)) (z (+ x x)))
         (declare (special x))
         (format t "x=~S y=~s z=~s~%" x y z) ) ) )
--> x=1 y=4 z=2

A symbol can be declared to be a constant by defconstant macro. Once declared, an attempt to change the value signals an error thereafter. Moreover, such a constant symbol is inhibited to be used as the name of a variable even for a local variable. NIL and T are examples of such constants. Symbols in the keyword package are always declared to be constants when they are created. In contrast, defvar and defparameter macro declare symbols to be special variables. defvar initializes the value only if the symbol is unbound, and does nothing when it already has a value assigned, while defparameter always resets the value.

When a symbol is referenced and there is no lexical binding for the symbol, its special value is retrieved. However, if no value has been assigned to its special value yet, unbound variable error is signaled.

Generalized Variables

Generally, any values or attributes are represented in slots of objects (or in stack frames). To retrieve and alter the value of a slot, two primitive operations, access and update, must be provided. Instead of defining two distinct primitives for every slot of objects, EusLisp, like Common Lisp, provides uniform update operations based on the generalized variable concept. In this concept, a common form is recognized either as a value access form or as a slot location specifier. Thus, you only need to remember accessing form for each slot and update is achieved by setf macro used in conjunction with the access form. For example, (car x) can be used to replace the value in the car slot of x when used with setf as in (setf (car '(a b) 'c), as well as to take the car value out of the list.

This method is also applicable to all the user defined objects. When a class or a structure is defined, the access and update forms for each slot are automatically defined. Each of those forms is defined as a macro whose name is the concatenation of the class name and slot name. For example, car of a cons can be addressed by (cons-car '(a b c)).

(defclass person :super object :slots (name age))
(defclass programmer :super person :slots (language machine))
(setq x (instantiate programmer))
(setf (programmer-name x) "MATSUI"
      (person-age x) 30)
(incf (programmer-age x))
(programmer-age x)   --> 31
(setf (programmer-language x) 'EUSLISP
      (programmer-machine x) 'SUN4)

Array elements can be accessed in the same manner.

(setq a (make-array '(3 3) :element-type :float))
(setf (aref a 0 0) 1.0 (aref a 1 1) 1.0 (aref a 2 2) 1.0)
a --> #2f((1.0 0.0 0.0) (0.0 1.0 0.0) (0.0 0.0 1.0))

(setq b (instantiate bit-vector 10))  --> #*0000000000
(setf (bit b 5) 1)
b --> #*0000010000

In order to define special setf methods for particular objects, defsetf macro is provided.

(defsetf symbol-value set)
(defsetf get (sym prop) (val) `(putprop ,sym ,val ,prop))

Special Forms


Table 2: EusLisp's special forms
and flet quote
block function return-from
catch go setq
cond if tagbody
declare labels the
defmacro let throw
defmethod let* unwind-protect
defun progn while
eval-when or


All the special forms are listed in Table 2. macrolet, compiler-let, and progv have not been implemented. Special forms are essential language constructs for the management of evaluation contexts and control flows. The interpreter and compiler have special knowledge to process each of these constructs properly, while the application method is uniform for all functions. Users cannot add their own special form definition.

Macros

Macro is a convenient method to expand language constructs. When a macro is called, arguments are passed to the macro body, which is a macro expansion function, without being evaluated. Then, the macro expansion function expands the arguments, and returns the new form. The resulted form is then evaluated again outside the macro. It is an error to apply a macro or special form to a list of arguments. Macroexpand function can be used for the explicit macro expansion.

Though macro runs slowly when interpreted, it speeds up compiled code execution, because macro expansion is taken at compile-time only once and no overhead is left to run-time. Note that explicit call to eval or apply in the macro function may produce different results between interpreted execution and the compiled execution.

Functions

A function is expressed by a lambda form which is merely a list whose first element is lambda. If a lambda form is defined for a symbol using defun, it can be referred as a global function name. Lambda form takes following syntax.


 (lambda ({var}* 
)}*]

) $\vert$ ((:keyword var) [initform])}*
]
)}*])
{declaration}*
{form}*)

There is no function type such as EXPR, LEXPR, FEXPR, etc.: arguments to a function are always evaluated before its application, and the number of acceptable arguments is determined by lambda-list. Lambda-list specifies the sequence of parameters to the lambda form. Each of &optional, &rest, &key and &aux has special meaning in lambda-lists, and these symbols cannot be used as variable names. Supplied-p variables for &optional or &key parameters are not supported.

Since a lambda form is indistinguishable from normal list data, function special form must be used to inform the interpreter and compiler the form is intended to be a function. 1Function is also important to freeze the environment onto the function, so that all the lexical variables can be accessible in the function even the function is passed to another function of different lexical scope. The following program does not work either interpretedly nor after compiled, since sum from the let is invisible inside lambda form.

(let ((x '(1 2 3)) (sum 0))
  (mapc '(lambda (x) (setq sum (+ sum x))) x))

To get the expected result, it should be written as follows:

(let ((x '(1 2 3)) (sum 0))
   (mapc #'(lambda (x) (setq sum (+ sum x))) x ))

#' is the abbreviated notation of function, i.e. #'(lambda (x) x) is equivalent to (function (lambda (x) x)). Here is another example of what is called a funarg problem:

(defun mapvector (f v)
    (do ((i 0 (1+ i)))
       ((>= i (length v)))
       (funcall f (aref v i))))
(defun vector-sum (v)
    (let ((i 0))
       (mapvector #'(lambda (x) (setq i (+ i x))) v)
       i))
(vector-sum #(1 2 3 4)) --> 10

EusLisp's closure cannot have indefinite extent: i.e. a closure can only survive as long as its outer extent is in effect. This means that a closure cannot be used for programming of “generators". The following program does not work.

(proclaim '(special gen))
(let ((index 0))
   (setq gen #'(lambda () (setq index (1+ index)))))
(funcall gen)

However, the same purpose is accomplished by object oriented programming, because an object can hold its own static variables:

(defclass generator object (index))
(defmethod generator
 (:next () (setq index (1+ index)))
 (:init (&optional (start 0)) (setq index start) self))
(defvar gen (instance generator :init 0))
(send gen :next)
This document was generated using the LaTeX2HTML translator on Sat Feb 5 14:36:57 JST 2022 from EusLisp version 138fb6ee Merge pull request #482 from k-okada/apply_dfsg_patch