<!doctype linuxdoc system>

<article>

<title>Common LISP Hints

<author>Geoffrey J. Gordon &lt;ggordon@cs.cmu.edu&gt;
        Modified by Bruno Haible &lt;haible@ma2s2.mathematik.uni-karlsruhe.de&gt;
	and Peter Van Eynde &lt;s950045@uia.ua.ac.be&gt;

<date>Friday, February 5, 1993

<toc>
      
<sect> Note for other lisp's.

<p>
Note: This tutorial introduction to Common Lisp was written for the
CMU environment, so some of the details of running lisp toward the end
may differ from site to site.


<sect>Further Information

<p>
The best LISP textbook I know of is

  Guy L. Steele Jr. <em/Common LISP: the Language , The second edition/. Digital Press. 1990. Also called Cltl2.

The first edition is easier to read; the second describes a more recent
standard. (The differences between the two standards shouldn't affect
casual programmers.)

A book by Dave Touretsky has also been recommended to me, although I
haven't read it, so I can't say anything about it.



<sect>Symbols

<p>
A symbol is just a string of characters. There are restrictions on what
you can include in a symbol and what the first character can be, but as
long as you stick to letters, digits, and hyphens, you'll be safe.
(Except that if you use only digits and possibly an initial hyphen,
LISP will think you typed an integer rather than a symbol.) Some
examples of symbols:

<enum>
<item>    a
<item>        b
<item>        c1
<item>        foo
<item>        bar
<item>        baaz-quux-garply
</enum>

Some things you can do with symbols follow. (Things after a <tt/``&gt;''/ prompt
are what you type to the LISP interpreter, while other things are what
the LISP interpreter prints back to you. The <tt/``;''/ is LISP's comment
character: everything from a <tt/``;''/ to the end of line is ignored.)

<tscreen><verb>
> (setq a 5)            ;store a number as the value of a symbol
                        ;cmucl will print a warning, ignore it.
5
> a                     ;take the value of a symbol
5
> (let ((a 6)) a)       ;bind the value of a symbol temporarily to 6
6
> a                     ;the value returns to 5 once the let is finished
5
> (+ a 6)               ;use the value of a symbol as an argument to a function
11
> b                     ;try to take the value of a symbol which has no value
Error: Attempt to take the value of the unbound symbol B
                        ;or in CMUCL:
Error in KERNEL::UNBOUND-SYMBOL-ERROR-HANDLER:  the variable B is unbound.

Restarts:
  0: [ABORT] Return to Top-Level.

Debug  (type H for help)

(EVAL B)
0]
                        ; return to top-level by typing "0" or
                        ; "restart 0"               
</verb></tscreen>

There are two special symbols, <em/t/ and <em/nil/. The value of <tt/t/ is defined
always to be <tt/t/, and the value of <tt/nil/ is defined always to be <tt/nil/. LISP
uses <tt/t/ and <tt/nil/ to represent true and false. An example of this use is
in the if statement, described more fully later:

<tscreen><verb>
> (if t 5 6)
5
> (if nil 5 6)
6
> (if 4 5 6)
5
</verb></tscreen>

The last example is odd but correct: <tt/nil/ means false, and anything else
means true. (Unless we have a reason to do otherwise, we use <tt/t/ to mean
true, just for the sake of clarity.)

Symbols like <tt/t/ and <tt/nil/ are called self-evaluating symbols, because
they evaluate to themselves. There is a whole class of self-evaluating
symbols called keywords; any symbol whose name starts with a colon is a
keyword. (See below for some uses for keywords.) Some examples:

<tscreen><verb>
> :this-is-a-keyword
:THIS-IS-A-KEYWORD
> :so-is-this
:SO-IS-THIS
> :me-too
:ME-TOO
</verb></tscreen>

<sect>Numbers

<p>
An integer is a string of digits optionally preceded by + or -. A real
number looks like an integer, except that it has a decimal point and
optionally can be written in scientific notation. A rational looks like
two integers with a / between them. LISP supports complex numbers,
which are written <tt/&num;c(r i)/ (where r is the real part and i is the
imaginary part). A number is any of the above. Here are some numbers:

<enum>
<item>        5
<item>        17
<item>        -34
<item>        +6
<item>        3.1415
<item>        1.722e-15
<item>        &num;c(1.722e-15 0.75)
</enum>

The standard arithmetic functions are all available: <tt/+/, <tt/-/,
<tt/*/, <tt>/</tt>, <tt/floor/, 
<tt/ceiling/, <tt/mod/, <tt/sin/, <tt/cos/, <tt/tan/, <tt/sqrt/,
<tt/exp/, <tt/expt/, and so forth. All of them
accept any kind of number as an argument. <tt/+/, <tt/-/, <tt/*/, and
<tt>/</tt> return a
number according to type contagion: an integer plus a rational is a
rational, a rational plus a real is a real, and a real plus a complex
is a complex. Here are some examples:

<tscreen><verb>
> (+ 3 3/4)             ;type contagion
15/4 
> (exp 1)               ;e
2.7182817 
> (exp 3)               ;e*e*e
20.085537 
> (expt 3 4.2)          ;exponent with a base other than e
100.90418
> (+ 5 6 7 (* 8 9 10))  ;the fns +-*/ all accept multiple arguments
</verb></tscreen>

There is no limit to the absolute value of an integer except the memory
size of your computer. Be warned that computations with bignums (as
large integers are called) can be slow. (So can computations with
rationals, especially compared to the corresponding computations with
small integers or floats.)

<sect>Conses

<p>
A cons is just a two-field record. The fields are called ``car'' and
``cdr'', for historical reasons. (On the first machine where LISP was
implemented, there were two instructions <tt/CAR/ and <tt/CDR/ which stood for
"Contents of Address portion of Register" and ""Contents of Decrement 
portion of Register". Registers were twice as wide as the width of an
address, so you could have 2 adresses in one register. (sizeof(int)==
2*sizeof(void *) ) Conses were implemented using these two registers.)

Conses are easy to use:

<tscreen><verb>
> (cons 4 5)            ;Allocate a cons. Set the car to 4 and the cdr to 5.
(4 . 5)
                        ;you have just allocated memory and promptly
			;forgotten about it. In C this would cause 
                        ;a memory leak, but not in LISP.
> (cons (cons 4 5) 6)
((4 . 5) . 6)
> (car (cons 4 5))
4
> (cdr (cons 4 5))
5
</verb></tscreen>


<sect>Lists

<p>
You can build many structures out of conses. Perhaps the simplest is a
linked list: the car of each cons points to one of the elements of the
list, and the cdr points either to another cons or to <tt/nil/. You can
create such a linked list with the list fuction:

<tscreen><verb>
> (list 4 5 6)
(4 5 6)
</verb></tscreen>

Notice that LISP prints linked lists a special way: it omits some of
the periods and parentheses. The rule is: if the <tt/cdr/ of a cons is
<tt/nil/, 
LISP doesn't bother to print the period or the <tt/nil/; and if the
<tt/cdr/ of 
cons A is cons B, then LISP doesn't bother to print the period for cons
A or the parentheses for cons B. So:

<tscreen><verb>
> (cons 4 nil)
(4)
> (cons 4 (cons 5 6))
(4 5 . 6)
> (cons 4 (cons 5 (cons 6 nil)))
(4 5 6)
</verb></tscreen>

The last example is exactly equivalent to the call <tt/(list 4 5 6)/. Note
that <tt/nil/ now means the list with no elements: the cdr of (a b), a list
with 2 elements, is (b), a list with 1 element; and the cdr of (b), a
list with 1 element, is <tt/nil/, which therefore must be a list with no
elements.

The car and cdr of <tt/nil/ are defined to be nil.

If you store your list in a variable, you can make it act like a stack:

<tscreen><verb>
> (setq a nil)
NIL
> (push 4 a)
(4)
> (push 5 a)
(5 4)
> (pop a)
5
> a
(4)
> (pop a)
4
> (pop a)
NIL
> a
NIL
</verb></tscreen>


<sect>Functions

<p>
You saw one example of a function above. Here are some more:

<tscreen><verb>
> (+ 3 4 5 6)                   ;this function takes any number of arguments
18
> (+ (+ 3 4) (+ (+ 4 5) 6))     ;isn't prefix notation fun?
22
> (defun foo (x y) (+ x y 5))   ;defining a function
FOO
> (foo 5 0)                     ;calling a function
10
> (defun fact (x)               ;a recursive function
    (if (&gt; x 0) 
      (* x (fact (- x 1)))
      1
  ) )
FACT
> (fact 5)
120
> (defun a (x) (if (= x 0) t (b (- x))))        ;mutually recursive functions
A
> (defun b (x) (if (&gt; x 0) (a (- x 1)) (a (+ x 1))))
B
> (a 5)
T
> (defun bar (x)                ;a function with multiple statements in
    (setq x (* x 3))            ;its body -- it will return the value
    (setq x (/ x 2))            ;returned by its final statement
    (+ x 4))
BAR
> (bar 6)
13
</verb></tscreen>

When we defined <tt/foo/, we gave it two arguments, <tt/x/ and
<tt/y/. Now when we 
call <tt/foo/, we are required to provide exactly two arguments: the first
will become the value of <tt/x/ for the duration of the call to
<tt/foo/, and the 
second will become the value of <tt/y/ for the duration of the call. In
LISP, most variables are lexically scoped; that is, if <tt/foo/ calls <tt/bar/
and <tt/bar/ tries to reference <tt/x/, <tt/bar/ will not get
<tt/foo/'s value for <tt/x/. 

The process of assigning a symbol a value for the duration of some
lexical scope is called binding.

You can specify optional arguments for your functions. Any argument
after the symbol <tt/&amp;optional/ is optional:

<tscreen><verb>
> (defun bar (x &amp;optional y) (if y x 0))
BAR
> (defun baaz (&amp;optional (x 3) (z 10)) (+ x z))
BAAZ
> (bar 5)
0
> (bar 5 t)
5
> (baaz 5)
15
> (baaz 5 6)
11
> (baaz)
13
</verb></tscreen>

It is legal to call the function <tt/bar/ with either one or two arguments.
If it is called with one argument, <tt/x/ will be bound to the value of that
argument and <tt/y/ will be bound to <tt/nil/; if it is called with two
arguments, <tt/x/ and <tt/y/ will be bound to the values of the first
and second 
argument, respectively.

The function <tt/baaz/ has two optional arguments. It specifies a default
value for each of them: if the caller specifies only one argument, <tt/z/
will be bound to 10 instead of to <tt/nil/, and if the caller specifies no
arguments, <tt/x/ will be bound to 3 and <tt/z/ to 10.

You can make your function accept any number of arguments by ending its
argument list with an <tt/&amp;rest/ parameter. LISP will collect all arguments
not otherwise accounted for into a list and bind the <tt/&amp;rest/
parameter to 
that list. So:

<tscreen><verb>
> (defun foo (x &amp;rest y) y)
FOO
> (foo 3)
NIL
> (foo 4 5 6)
(5 6)
</verb></tscreen>

Finally, you can give your function another kind of optional argument
called a keyword argument. The caller can give these arguments in any
order, because they're labelled with keywords.

<tscreen><verb>
> (defun foo (&amp;key x y) (cons x y))
FOO
> (foo :x 5 :y 3)
(5 . 3)
> (foo :y 3 :x 5)
(5 . 3)
> (foo :y 3)
(NIL . 3)
> (foo)
(NIL)
</verb></tscreen>

An <tt/&amp;key/ parameter can have a default value too:

<tscreen><verb>
> (defun foo (&amp;key (x 5)) x)
FOO
> (foo :x 7)
7
> (foo)
5
</verb></tscreen>


<sect>Printing

<p>
Some functions can cause output. The simplest one is <tt/print/, which
prints its argument and then returns it.

<tscreen><verb>
> (print 3)
3
3
</verb></tscreen>

The first 3 above was printed, the second was returned.

If you want more complicated output, you will need to use <tt/format/.
Here's an example:

<tscreen><verb>
> (format t "An atom: &tilde;S&tilde;&percnt;and a list: &tilde;S&tilde;&percnt;and an integer: &tilde;D&tilde;&percnt;"
          nil (list 5) 6)
An atom: NIL
and a list: (5)
and an integer: 6
</verb></tscreen>

The first argument to format is either <tt/t/, <tt/nil/, or a
stream. <tt/T/ specifies 
output to the terminal. <tt/Nil/ means not to print anything but to return a
string containing the output instead. Streams are general places for
output to go: they can specify a file, or the terminal, or another
program. This handout will not describe streams in any further detail.

The second argument is a formatting template, which is a string
optionally containing formatting directives.

All remaining arguments may be referred to by the formatting
directives. LISP will replace the directives with some appropriate
characters based on the arguments to which they refer and then print
the resulting string.

Format always returns <tt/nil/ unless its first argument is <tt/nil/, in which
case it prints nothing and returns the string instead of printing it.

There are three different directives in the above example:
<tt/&tilde;S/, <tt/&tilde;D/, and 
<tt/&tilde;&percnt;/. The first one accepts any LISP object and is
replaced by a printed 
representation of that object (the same representation which is
produced by print). The second one accepts only integers. The third one
doesn't refer to an argument; it is always replaced by a carriage
return.

Another useful directive is <tt/&tilde;&tilde;/, which is replaced by
a single &tilde;. 

Refer to a LISP manual for (many, many) additional formatting
directives.



<sect>Forms and the Top-Level Loop

<p>
The things which you type to the LISP interpreter are called forms; the
LISP interpreter repeatedly reads a form, evaluates it, and prints the
result. This procedure is called the read-eval-print loop.

Some forms will cause errors. After an error, LISP will put you into
the debugger so you can try to figure out what caused the error. LISP
debuggers are all different; but most will respond to the command
"help" or ":help" by giving some form of help.

In general, a form is either an atom (for example, a symbol, an
integer, or a string) or a list. If the form is an atom, LISP evaluates
it immediately. Symbols evaluate to their value; integers and strings
evaluate to themselves. If the form is a list, LISP treats its first
element as the name of a function; it evaluates the remaining elements
recursively, and then calls the function with the values of the
remaining elements as arguments.

For example, if LISP sees the form <tt/(+ 3 4)/, it treats <tt/+/ as
the name of 
a function. It then evaluates 3 to get 3 and 4 to get 4; finally it
calls <tt/+/ with 3 and 4 as the arguments. The <tt/+/ function
returns 7, which 
LISP prints.

The top-level loop provides some other conveniences; one particularly
convenient convenience is the ability to talk about the results of
previously typed forms. LISP always saves its most recent three
results; it stores them as the values of the symbols <tt/*/, <tt/**/,
and <tt/***/. 
For example:

<tscreen><verb>
> 3
3
> 4
4
> 5
5
> ***
3
> ***
4
> ***
5
> **
4
> *
4
</verb></tscreen>


<sect>Special forms

<p>
There are a number of special forms which look like function calls but
aren't. These include control constructs such as <tt/if/ statements and do
loops; assignments like <tt/setq/, <tt/setf/, <tt/push/, and <tt/pop/;
definitions such as 
<tt/defun/ and <tt/defstruct/; and binding constructs such as
<tt/let/. (Not all of 
these special forms have been mentioned yet. See below.)

One useful special form is the <tt/quote/ form: <tt/quote/ prevents
its argument 
from being evaluated. For example:

<tscreen><verb>
> (setq a 3)
3
> a
3
> (quote a)
A
> 'a                    ;'a is an abbreviation for (quote a)
                        ;it's the quote next to the enter key
                        ;on a qwerty keyboard
A
</verb></tscreen>

Another similar special form is the <tt/function/ form: <tt/function/
causes its 
argument to be interpreted as a function rather than being evaluated.
For example:

<tscreen><verb>
> (setq + 3)
3
> +
3
> '+
+
> (function +)
#<Function + @ #x-fbef9de>
> #'+                   ;#'+ is an abbreviation for (function +)
#<Function + @ #x-fbef9de>
</verb></tscreen>

The <tt/function/ special form is useful when you want to pass a function as
an argument to another function. See below for some examples of
functions which take functions as arguments.

<sect>Binding

<p>
Binding is lexically scoped assignment. It happens to the variables in
a function's parameter list whenever the function is called: the formal
parameters are bound to the actual parameters for the duration of the
function call. You can bind variables anywhere in a program with the
let special form, which looks like this:

<tscreen><verb>
(let ((var1 val1)
      (var2 val2)
      ...
     )
    body)
</verb></tscreen>

(It's considered bad style to leave a single ``)'' on a line)

Let binds <tt/var1/ to <tt/val1/, <tt/var2/ to <tt/val2/, and so
forth; then it executes 
the statements in its body. The body of a let follows exactly the same
rules that a function body does. Some examples:

<tscreen><verb>
> (let ((a 3)) (+ a 1))
4
> (let ((a 2) 
        (b 3)
        (c 0))
    (setq c (+ a b))
    c)
5
> (setq c 4)
4                     ;and a warning from CMUCL, ignore it    
> (let ((c 5)) c)
5
> c
4
</verb></tscreen>

Instead of <tt/(let ((a nil) (b nil)) ...)/, you can write <tt/(let (a b) ...)/.

The <tt/val1/, <tt/val2/, etc. inside a let cannot reference the
variables <tt/var1/, 
<tt/var2/, etc. that the let is binding. For example,

<tscreen><verb>
> (let ((x 1)
        (y (+ x 1)))
    y)
Error: Attempt to take the value of the unbound symbol X
                      ;in CMUCL:

In: LET ((X 1) (Y (+ X 1)))
  (LET ((X 1) (Y #))
    Y)
Warning: Variable X defined but never used.


Warning: This variable is undefined:
  X

Error in KERNEL::UNBOUND-SYMBOL-ERROR-HANDLER:  the variable X is
unbound.

Restarts:
  0: [ABORT] Return to Top-Level.

Debug  (type H for help)

(EVAL::LEAF-VALUE
 #<C::REF #x9009B15  LEAF= #<C::GLOBAL-VAR #x9009AD5  NAME= X  KIND=
 :GLOBAL>>
 0
 #())
0]
</verb></tscreen>

If the symbol <tt/x/ already has a global value, stranger happenings will
result:

<tscreen><verb>
> (setq x 7)
7                     ;again a warning in CMUCL:
Warning:  Declaring X special.
                      ;x is automaticly declared special
> (let ((x 1)
        (y (+ x 1)))
    y
  )
8
</verb></tscreen>

The <tt/let*/ special form is just like <tt/let/ except that it allows
values to 
reference variables defined earlier in the <tt/let*/. For example,

<tscreen><verb>
> (setq x 7)
7
> (let* ((x 1)
         (y (+ x 1)))
    y
  )
2
</verb></tscreen>

The form

<tscreen><verb>
(let* ((x a)
       (y b))
       ...
) 
</verb></tscreen>

is equivalent to

<tscreen><verb>
(let ((x a))
     (let ((y b))
          ...
) )
</verb></tscreen>

<sect>Dynamic Scoping

<p>
The <tt/let/ and <tt/let*/ forms provide lexical scoping, which is what you
expect if you're used to programming in C or Pascal. Dynamic scoping is
what you get in BASIC: if you assign a value to a dynamically scoped
variable, every mention of that variable returns that value until you
assign another value to the same variable.

In LISP, dynamically scoped variables are called <em/special variables/. You
can declare a special variable with the defvar special form. Here are
some examples of lexically and dynamically scoped variables.

In this example, the function <tt/check-regular/ references a regular (ie,
lexically scoped) variable. Since <tt/check-regular/ is lexically outside of
the <tt/let/ which binds <tt/regular/, <tt/check-regular/ returns the
variable's global value.

<tscreen><verb>
> (set 'regular 5)    ;setq would make it special in CMUCL!
5 
> (defun check-regular () regular)
CHECK-REGULAR 
> (check-regular)
5 
> (let ((regular 6)) (check-regular))
5
</verb></tscreen>

In this example, the function <tt/check-special/ references a special (ie,
dynamically scoped) variable. Since the call to <tt/check-special/ is
temporally inside of the <tt/let/ which binds special,
<tt/check-special/ returns 
the variable's local value.

<tscreen><verb>
> (defvar *special* 5)
*SPECIAL*
> (defun check-special () *special*)
CHECK-SPECIAL
> (check-special)
5
> (let ((*special* 6)) (check-special))
6
</verb></tscreen>

By convention, the name of a special variable begins and ends with a *.
Special variables are chiefly used as global variables, since
programmers usually expect lexical scoping for local variables and
dynamic scoping for global variables.

For more information on the difference between lexical and dynamic
scoping, see <em/Common LISP: the Language/.


<sect>Arrays

<p>
The function <tt/make-array/ makes an array. The <tt/aref/ function
accesses its 
elements. All elements of an array are initially set to <tt/nil/. For
example:

<tscreen><verb>
> (make-array '(3 3))
#2a((NIL NIL NIL) (NIL NIL NIL) (NIL NIL NIL))
> (aref * 1 1)
NIL
> (make-array 4)        ;1D arrays don't need the extra parens
#(NIL NIL NIL NIL)
</verb></tscreen>

Array indices always start at 0.

See below for how to set the elements of an array.



<sect>Strings

<p>
A string is a sequence of characters between double quotes. LISP
represents a string as a variable-length array of characters. You can
write a string which contains a double quote by preceding the quote
with a backslash; a double backslash stands for a single backslash. For
example:

<tscreen><verb>
"abcd" has 4 characters
"\"" has 1 character
"\\" has 1 character
</verb></tscreen>

Here are some functions for dealing with strings:

<tscreen><verb>
> (concatenate 'string "abcd" "efg")
"abcdefg"
> (char "abc" 1)
#\b                     ;LISP writes characters preceded by #\
> (aref "abc" 1)
#\b                     ;remember, strings are really arrays
</verb></tscreen>

The <tt/concatenate/ function can actually work with any type of sequence:

<tscreen><verb>
> (concatenate 'string '(#\a #\b) '(#\c))
"abc"
> (concatenate 'list "abc" "de")
(#\a #\b #\c #\d #\e)
> (concatenate 'vector '#(3 3 3) '#(3 3 3))
#(3 3 3 3 3 3)
</verb></tscreen>


<sect>Structures

<p>
LISP structures are analogous to C structs or Pascal records. Here is
an example:

<tscreen><verb>
> (defstruct foo
    bar
    baaz
    quux)
FOO
</verb></tscreen>

This example defines a data type called <tt/foo/ which is a structure
containing 3 fields. It also defines 4 functions which operate on this
data type: <tt/make-foo/, <tt/foo-bar/, <tt/foo-baaz/, and
<tt/foo-quux/. The first one 
makes a new object of type <tt/foo/; the others access the fields of an
object of type <tt/foo/. Here is how to use these functions:

<tscreen><verb>
> (make-foo)
#s(FOO :BAR NIL :BAAZ NIL :QUUX NIL) 
> (make-foo :baaz 3)
#s(FOO :BAR NIL :BAAZ 3 :QUUX NIL) 
> (foo-bar *)
NIL
> (foo-baaz **)
3
</verb></tscreen>

The <tt/make-foo/ function can take a keyword argument for each of the
fields a structure of type <tt/foo/ can have. The field access functions
each take one argument, a structure of type <tt/foo/, and return the
appropriate field.

See below for how to set the fields of a structure.



<sect>Setf

<p>
Certain forms in LISP naturally define a memory location. For example,
if the value of <tt/x/ is a structure of type <tt/foo/, then
<tt/(foo-bar x)/ defines 
the <tt/bar/ field of the value of <tt/x/. Or, if the value of <tt/y/
is a one-dimensional array, <tt/(aref y 2)/ defines the third element
of <tt/y/. 

The <tt/setf/ special form uses its first argument to define a place in
memory, evaluates its second argument, and stores the resulting value
in the resulting memory location. For example,

<tscreen><verb>
> (setq a (make-array 3))
#(NIL NIL NIL)
> (aref a 1)
NIL
> (setf (aref a 1) 3)
3
> a
#(NIL 3 NIL)
> (aref a 1)
3
> (defstruct foo bar)
FOO
> (setq a (make-foo))
#s(FOO :BAR NIL)
> (foo-bar a)
NIL
> (setf (foo-bar a) 3)
3
> a
#s(FOO :BAR 3)
> (foo-bar a)
3
</verb></tscreen>

<tt/Setf/ is the only way to set the fields of a structure or the elements
of an array.

Here are some more examples of <tt/setf/ and related functions.

<tscreen><verb>
> (setf a (make-array 1))       ;setf on a variable is equivalent to setq
#(NIL)
> (push 5 (aref a 1))           ;push can act like setf
(5)
> (pop (aref a 1))              ;so can pop
5
> (setf (aref a 1) 5)
5
> (incf (aref a 1))             ;incf reads from a place, increments,
6                               ;and writes back
> (aref a 1)
6
</verb></tscreen>


<sect>Booleans and Conditionals

<p>
LISP uses the self-evaluating symbol <tt/nil/ to mean false. Anything other
than <tt/nil/ means true. Unless we have a reason not to, we usually use the
self-evaluating symbol <tt/t/ to stand for true.

LISP provides a standard set of logical functions, for example
<tt/and/, <tt/or/, 
and <tt/not/. The <tt/and/ and <tt/or/ connectives are
short-circuiting: and will not 
evaluate any arguments to the right of the first one which evaluates to
<tt/nil/, while <tt/or/ will not evaluate any arguments to the right
of the first 
one which evaluates to <tt/t/ (with <tt/t/ we mean here ``non-<tt/nil/'').

LISP also provides several special forms for conditional execution. The
simplest of these is <tt/if/. The first argument of <tt/if/ determines whether
the second or third argument will be executed:

<tscreen><verb>
> (if t 5 6)
5
> (if nil 5 6)
6
> (if 4 5 6)
5
</verb></tscreen>

If you need to put more than one statement in the then or else clause
of an if statement, you can use the <tt/progn/ special
form. <tt/Progn/ executes 
each statement in its body, then returns the value of the final one.

<tscreen><verb>
> (setq a 7)
7
> (setq b 0)
0
> (setq c 5)
5
> (if (&gt; a 5)
    (progn
      (setq a (+ b 7))
      (setq b (+ c 8)))
    (setq b 4))
13
</verb></tscreen>

An <tt/if/ statement which lacks either a then or an else clause can be
written using the <tt/when/ or <tt/unless/ special form:

<tscreen><verb>
> (when t 3)
3
> (when nil 3)
NIL
> (unless t 3)
NIL
> (unless nil 3)
3
</verb></tscreen>

<tt/When/ and <tt/unless/, unlike <tt/if/, allow any number of
statements in their 
bodies. (Eg, <tt/(when x a b c)/ is equivalent to 
<tt/(if x (progn a b c))/.)

<tscreen><verb>
> (when t
    (setq a 5)
    (+ a 6))
11
</verb></tscreen>

More complicated conditionals can be defined using the <tt/cond/ special
form, which is equivalent to an if ... else if ... fi construction.

A <tt/cond/ consists of the symbol <tt/cond/ followed by a number of cond
clauses, each of which is a list. The first element of a cond clause is
the condition; the remaining elements (if any) are the action. The cond
form finds the first clause whose condition evaluates to true (ie,
doesn't evaluate to nil); it then executes the corresponding action and
returns the resulting value. None of the remaining conditions are
evaluated; nor are any actions except the one corresponding to the
selected condition. For example:

<tscreen><verb>
> (setq a 3)
3
> (cond
   ((evenp a) a)        ;if a is even return a
   ((&gt; a 7) (/ a 2))    ;else if a is bigger than 7 return a/2
   ((&lt; a 5) (- a 1))    ;else if a is smaller than 5 return a-1
   (t 17))              ;else return 17
2
</verb></tscreen>

If the action in the selected cond clause is missing, <tt/cond/ returns what
the condition evaluated to:

<tscreen><verb>
> (cond ((+ 3 4)))
7
</verb></tscreen>

Here's a clever little recursive function which uses <tt/cond/. You might be
interested in trying to prove that it terminates for all integers <tt/x/ at
least 1. (If you succeed, please publish the result.)

<tscreen><verb>
> (defun hotpo (x steps)        ;hotpo stands for Half Or Triple Plus One
    (cond
     ((= x 1) steps)
     ((oddp x) (hotpo (+ 1 (* x 3)) (+ 1 steps)))
     (t (hotpo (/ x 2) (+ 1 steps)))))
A
> (hotpo 7 0)
16
</verb></tscreen>

The LISP case statement is like a C switch statement:

<tscreen><verb>
> (setq x 'b)
B
> (case x
   (a 5)
   ((d e) 7)
   ((b f) 3)
   (otherwise 9))
3
</verb></tscreen>

The otherwise clause at the end means that if <tt/x/ is not <tt/a/,
<tt/b/, <tt/d/, <tt/e/, or <tt/f/, the case statement will return 9.



<sect>Iteration

<p>
The simplest iteration construct in LISP is <tt/loop/: a <tt/loop/ construct
repeatedly executes its body until it hits a <tt/return/ special form. For
example,

<tscreen><verb>
> (setq a 4)
4
> (loop 
   (setq a (+ a 1))
   (when (&gt; a 7) (return a)))
8
> (loop
   (setq a (- a 1))
   (when (&lt; a 3) (return)))
NIL
</verb></tscreen>

The next simplest is <tt/dolist/: <tt/dolist/ binds a variable to the
elements of 
a list in order and stops when it hits the end of the list.

<tscreen><verb>
> (dolist (x '(a b c)) (print x))
A 
B 
C 
NIL 
</verb></tscreen>

<tt/Dolist/ always returns <tt/nil/. Note that the value of <tt/x/ in the above
example was never <tt/nil/: the<tt/ NIL/ below the C was the value
that <tt/dolist/ 
returned, printed by the read-eval-print loop.

The most complicated iteration primitive is called <tt/do/. A <tt/do/ statement
looks like this: 

<tscreen><verb>
> (do ((x 1 (+ x 1))
       (y 1 (* y 2)))
      ((&gt; x 5) y)
    (print y)
    (print 'working)
  )
1 
WORKING 
2 
WORKING 
4 
WORKING 
8 
WORKING 
16 
WORKING 
32 
</verb></tscreen>

The first part of a <tt/do/ specifies what variables to bind, what their
initial values are, and how to update them. The second part specifies a
termination condition and a return value. The last part is the body. A
do form binds its variables to their initial values like a <tt/let/, then
checks the termination condition. As long as the condition is false, it
executes the body repeatedly; when the condition becomes true, it
returns the value of the return-value form.

The <tt/do/* form is to <tt/do/ as <tt/let/* is to <tt/let/.



<sect>Non-local Exits

<p>
The <tt/return/ special form mentioned in the section on iteration is an
example of a nonlocal return. Another example is the <tt/return/-from form,
which returns a value from the surrounding function:

<tscreen><verb>
> (defun foo (x)
    (return-from foo 3)
    x)
FOO
> (foo 17)
3
</verb></tscreen>

Actually, the <tt/return-from/ form can return from any named block -- it's
just that functions are the only blocks which are named by default. You
can create a named block with the block special form:

<tscreen><verb>
> (block foo
    (return-from foo 7)
    3)
7
</verb></tscreen>

The <tt/return/ special form can return from any block named
<tt/nil/. Loops are 
by default labelled <tt/nil/, but you can make your own nil-labelled blocks:

<tscreen><verb>
> (block nil
    (return 7)
    3)
7
</verb></tscreen>

Another form which causes a nonlocal exit is the <tt/error/ form:

<tscreen><verb>
> (error "This is an error")
Error: This is an error
</verb></tscreen>

The <tt/error/ form applies format to its arguments, then places you in the
debugger.



<sect>Funcall, Apply, and Mapcar

<p>
Earlier I promised to give some functions which take functions as
arguments. Here they are:

<tscreen><verb>
> (funcall &num;'+ 3 4)
7
> (apply &num;'+ 3 4 '(3 4))
14
> (mapcar &num;'not '(t nil t nil t nil))
(NIL T NIL T NIL T)
</verb></tscreen>

<tt/Funcall/ calls its first argument on its remaining arguments.

<tt/Apply/ is just like <tt/funcall/, except that its final argument
should be a 
list; the elements of that list are treated as if they were additional
arguments to a <tt/funcall/.

The first argument to <tt/mapcar/ must be a function of one argument;
<tt/mapcar/ 
applies this function to each element of a list and collects the
results in another list.

<tt/Funcall/ and <tt/apply/ are chiefly useful when their first argument is a
variable. For instance, a search engine could take a heuristic function
as a parameter and use <tt/funcall/ or <tt/apply/ to call that function on a
state description. The sorting functions described later use <tt/funcall/
to call their comparison functions.

<tt/Mapcar/, along with nameless functions (see below), can replace many
loops.



<sect>Lambda

<p>
If you just want to create a temporary function and don't want to
bother giving it a name, <tt/lambda/ is what you need.

<tscreen><verb>
> #'(lambda (x) (+ x 3))
(LAMBDA (X) (+ X 3))
> (funcall * 5)               ;* is the last result,remember?
8
<tscreen><verb>

The combination of <tt/lambda/ and <tt/mapcar/ can replace many loops. For
example, the following two forms are equivalent:

<tscreen><verb>
> (do ((x '(1 2 3 4 5) (cdr x))
       (y nil))
      ((null x) (reverse y))
    (push (+ (car x) 2) y))
(3 4 5 6 7)
> (mapcar #'(lambda (x) (+ x 2)) '(1 2 3 4 5))
(3 4 5 6 7)
</verb></tscreen>

However the first version will run significantly faster when compiled
by CMUCL. But speed isn't everything and the second version is a lot
clearer and easier to debug.

<sect>Sorting

<p>
LISP provides two primitives for sorting: <tt/sort/ and <tt/stable-sort/.

<tscreen><verb>
> (sort '(2 1 5 4 6) #'<)
(1 2 4 5 6)
> (sort '(2 1 5 4 6) #'>)
(6 5 4 2 1)
</verb></tscreen>

The first argument to <tt/sort/ is a list; the second is a comparison
function. The <tt/sort/ function does not guarantee stability: if there are
two elements <tt/a/ and <tt/b/ such that <tt/(and (not (< a b)) (not
(< b a)))/, sort 
may arrange them in either order. The <tt/stable-sort/ function is exactly
like <tt/sort/, except that it guarantees that two equivalent elements
appear in the sorted list in the same order that they appeared in the
original list.

Be careful: <tt/sort/ is allowed to destroy its argument, so if the original
sequence is important to you, make a copy with the <tt/copy-list/ or
<tt/copy/-seq/ function.



<sect>Equality

<p>
LISP has many different ideas of equality. Numerical equality is
denoted by <tt/=/. Two symbols are <tt/eq/ if and only if they are
identical. Two 
copies of the same list are not <tt/eq/, but they are equal.

<tscreen><verb>
> (eq 'a 'a)
T
> (eq 'a 'b)
NIL
> (= 3 4)
T
> (eq '(a b c) '(a b c))
NIL
> (equal '(a b c) '(a b c))
T
> (eql 'a 'a)
T
> (eql 3 3)
T
</verb></tscreen>

The <tt/eql/ predicate is equivalent to <tt/eq/ for symbols and to
<tt/=/ for numbers. 

The <tt/equal/ predicate is equivalent to <tt/eql/ for symbols and
numbers. It is 
true for two conses if and only if their cars are equal and their cdrs
are equal. It is true for two structures if and only if the structures
are the same type and their corresponding fields are equal.



<sect>Some Useful List Functions

<p>
These functions all manipulate lists.

<tscreen><verb>
> (append '(1 2 3) '(4 5 6))    ;concatenate lists
(1 2 3 4 5 6)
> (reverse '(1 2 3))            ;reverse the elements of a list
(3 2 1)
> (member 'a '(b d a c))        ;set membership -- returns the first tail
(A C)                           ;whose car is the desired element
> (find 'a '(b d a c))          ;another way to do set membership
A
> (find '(a b) '((a d) (a d e) (a b d e) ()) :test #'subsetp)
(A B D E)                       ;find is more flexible though
> (subsetp '(a b) '(a d e))     ;set containment
NIL
> (intersection '(a b c) '(b))  ;set intersection
(B)
> (union '(a) '(b))             ;set union
(A B)
> (set-difference '(a b) '(a))  ;set difference
(B)
</verb></tscreen>

<tt/Subsetp/, <tt/intersection/, <tt/union/, and <tt/set-difference/
all assume that each 
argument contains no duplicate elements -- <tt/(subsetp '(a a) '(a b b))/ is
allowed to fail, for example.

<tt/Find/, <tt/subsetp/, <tt/intersection/, <tt/union/, and
<tt/set-difference/ can all take a
<tt/:test/ keyword argument; by default, they all use <tt/eql/.



<sect>Getting Started with Emacs

<p>
You can use Emacs to edit LISP code: most Emacses are set up to enter
LISP mode automatically when they find a file which ends in .lisp, but
if yours isn't, you can type <tt/M-x lisp-mode/. M-x means: meta-key
plus x. If there is a meta (or ALT) key try it. If this fails M-x can
be typed by typing ESC followed by x. If your keyboard lacks an ESC
key, curse DEC, and type control-&lsqb; followed by x.

You can run LISP under Emacs, too: make sure that there is a command in
your path called "lisp" which runs your favorite LISP. For example, you
could type

<tscreen><verb>
ln -s /usr/local/bin/clisp &tilde;/bin/lisp
</verb></tscreen>

This isn't needed in CMUCL. Lisp is called lisp there.

Then in Emacs type <tt/M-x run-lisp/. You can send LISP code to the LISP you
just started, and do all sorts of other cool things; for more
information, type <tt/C-h m/ from any buffer which is in LISP
mode. This means: press control <em/and/ h, release both and press m.

Actually, you don't even need to make a link. Emacs has a variable
called inferior-lisp-program; so if you add the line

<tscreen><verb>
(setq inferior-lisp-program "/usr/local/bin/clisp")
</verb></tscreen>

to your .emacs file, Emacs will know where to find CLISP when
you type <tt/M-x run-lisp/.


</article>