33 KiB
The Slime 1.0 Manual
abstract
\tableofcontents
Lisp languages
Lisp is not one language but rather a family of programming languages. The family is devided by some characteristics. There are Lisp-1 and Lisp-2 dialects and there is a difference between a Lisp with lexical scoping as opposed to dynamic scoping. These differences will be explained in later sections.
Like most Lisps, Slime is dynamically typed. That means that like in statically typed Languages Slime has different data types, but they are associated not with variables but with the Lisp objects themselves. Variables can be assigend Lisp objects of any internal type.
The Lisp language family is known to be highly flexible and applicable in all areas by creating domain specific languages in Lisp itself through a powerful macro system. The central data structure in Lisp is the list. The reason why lisp is so powerful is because the program source code itself is represented as lists. The nested lists make up the syntax tree of the lisp program. It is therfore computationally easy to parse lisp programs as the source code itself is already structured in the form of the syntax tree; allowing for parsing in linear time.
The macro system in Slime works by recognizing macros at parse-time and running them, and replacing the macro call in the program code with the return value of the macro and then checking if further macros have to be expanded in the replaced code. Therefore the macros can be used to pre-compute values or rewrite expressions (creating syntactic sugar) or themselves define macros.
Lists
As mentioned in Lisp languages, the central data structure in all Lisps is the list. Lists are
implemented as singly linked lists, made up of pairs (historically called cons-cells), each pair
has two slots, the first and the rest (historically car and cdr). A linked lsit ist then
constructed by the convention that the first field of a pair points to the first element of the
list and the rest field points to the rest of the list. Following this description, the list is a
recursive data structure. For the end of the list a special value nil is used in the rest field.
A helpful way to visualize lists made up of pairs is using box diagrams. A simple box diagram can be
seen in simpleBoxDiagram. Each rectangle is divided in two. The left part represents the first
field, the right part represents the rest. The arrows point to the values in these fields.
The diagram in simpleBoxDiagram shows a simple list containing the values 1, 2 and 3. The first pair
stores the number 1 its first field and the rest points to the rest of the list. The last pair
points to the special value nil in its rest to denote the end of the list.
{{{ditaa_header}}}
+-----+-----+ +-----+-----+ +-----+-----+
| | | | | | | | |
| | |--->| | |--->| | / |
| | | | | | | | |
+-----+-----+ +-----+-----+ +-----+-----+
| | |
| | |
V V V
1 2 3
However the rest of a pair needs not to be a pair or nil, it could also point to any other value.
By doing this the list is no longer "well formed" but rather "ill formed". Ill formed lsits can be
used as an optimization when using the list for storing data. In illFormedList an ill formed list
can be seen, that also contains the values 1, 2 and 3 but stores them using only two pairs instead
of 3.
{{{ditaa_header}}}
+-----+-----+ +-----+-----+
| | | | | |
| | |--->| | |--->3
| | | | | |
+-----+-----+ +-----+-----+
| |
| |
V V
1 2representing lists in Lisp
In Slime and in most Lisps, lists are represented using round parenthesis where ( denotes the
start of the list and ) denotes the end. Eeach element inside these parenthesis separated by one
or more spaces will be interpreted as an element of that list. For example the list from
simpleBoxDiagram would be represented as (1 2 3). During parse time, the Lisp parser transforms
the parenthesised list into the pairs that are in the end stored in memory.
To also be able to represent ill formed lists in Lisp there is a special syntax using the . (dot
symbol). If the parser encounters a . inside of a list, it will treat the next element as the
rest. If there is no or more than one element after the . an parsing error will be thrown. Using
this syntax we can represent the ill formed list from illFormedList as (1 2 . 3). We can also
write well formed lists using the dot notation if we point the rest to another list. So the well
formed list from simpleBoxDiagram can also be written as \[\texttt{(1 . (2 . (3)))}\]
representing function calls in Lisp
If we tried to enter the Lisp representation of the lists like (1 2 3) discussed in [[representing
lists in Lisp]] directly into an Lisp interpreter we would get an error. That doesn't mean that the
explanation given in the section is wrong, it is in fact correct: the lisp parser will transform the
lisp syntax into the pairs in memory. The reason we would get an error is, that when reading Lisp
code, the Lisp interpreter first parses the code and then tries to evaluate it and return the result
back to the user.
In Lisp by default, a list corresponds to a function call. As mentioned in Lisp languages Lisp
represents lists and Lisp programms as lists. If a list is treated as a function call, the first
element will be treated as the function and the rest of the elements will be the arguments to that
function. If we would wnter (1 2 3) directly into the Lisp interpreter we would get an error
saying it cannot find the function 1.
If we would want to compute the sum of the numbers 5 and 3 we could do this by invoking the +
function with 5 and 3 as its arguments. (+ 5 3) will evaluate to 8. We can also nest functions
calls and use the return values as parameters to other functions: (+ (- 12 4) (/ 24 4)) will
evaluate to 14. The box diagramm showing the internal structure of that computation can be seen in
moreComplexBoxDiagram.
{{{ditaa_header}}}
+-----+-----+ +-----+-----+ +-----+-----+
| | | | | | | | |
| | |--->| | |--------------------------------------->| | / |
| | | | | | | | |
+-----+-----+ +-----+-----+ +-----+-----+
| | |
| | |
V V V
+ +-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+
| | | | | | | | | | | | | | | | | |
| | |--->| | |--->| | / | | | |--->| | |--->| | / |
| | | | | | | | | | | | | | | | | |
+-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+ +-----+-----+
| | | | | |
| | | | | |
V V V V V V
- 12 4 / 24 4(+ (- 12 4) (/ 24 4))
Evaluation order
As a first step of evaluation of a regular function, all its arguments are getting evaluated, and
then the function is applied to the evaluated arguments. For example when evaluating the nested
expression in code:complex-math the outermost function is the + function with three arguments: (*
3 4), (- 100 (+ 12 13 14 15)) and 2. So before the outhermost + gets invoked, the three
arguments are getting evaluated recursively.
{{{slime_header}}}
(print (+ (* 3 4)
(- 100 (+ 12 13 14 15))
2))evaluates to => 60
Special forms
The given evaluation rule – to evaluate all the arguments first and then allpying them to the
funciton – as described in Evaluation order is only valid for regular functions. There is a class
of functions that do not follow this evaluation rule called special forms. Special forms are
needed when you do not wish to evaluate all arguments. For example the built-in if function should
only evaluate the "then-expression" if the condition evaluates to a truthy value and not otherwise.
Consider the example in code:special-forms. The if expression only evaluates the then-expression. If
the if function would follow the evaluation order of regular functions, first all three arguments
(< 1 2), (print "I knew it!!\n") but also (print "Oh, it is not?!\n") could get evaluated and
so both messages would be printed. In the given if expression, the condition evaluates to a truthy
value and only I knew it!! will be printed.
{{{slime_header}}}
(if (< 1 2)
(print "I knew it!!\n")
(print "Oh, it is not?!\n"))evaluates to => I knew it!!
if function is a special form because it does not evaluate all of its arguments
The programmer can also define their own special forms using special-lambda and macros, which will
be explained in Special lambdas and Macros.
Symbols and keywords
Truthyness
Lambdas
Slime allows for creating anonymous functions called lambdas. We did not talk about binding
variables, we will do this in Define, but we can still use lambdas now. Remember that Lisp
interpretes the first argument of a list in the source code as a function and the rest as the
arguments. The lambda special form evaluates to a regular function object that can then stand in
the first position of the function call list. The basic syntax for the lambda special form is:
\[\texttt{(lambda (arg1 arg2 ...) (body1) ...)}\] the first arguemnt to lambda is a list of the
arguments. All the following arguments will be the body of the lambda. They will be executed when
the lambda is invoked. The return value of a lambda is the value of the last evaluated expression in
the body.
Probably the simplest function to write as a lambda is the identity function. It takes one argument and returns it. The identity lambda and a few other simple examples of lambdas can be seen in code:simple-lambdas.
{{{slime_header}}}
(printf ((lambda (x) x) 1))
(printf ((lambda (x y) (+ x y)) 3 5))
(printf ((lambda (x y z) (list x y z)) 1 2 3))evaluates to => 1 8 (1 2 3)
Additionally Slime lambdas have the possibility to take optional arguments in the form of keyword arguemnts as well as a rest argument which allows for accepting any number of arguments. Since these concepts are most useful when the function is actually bound to a variable, they will be introduced when we learned how to do that in Define.
Special lambdas
The lambda special form creates a function object that represents a regular function. So the basic
evaluation rules count: when the lambda is invoked all it's arguments are evaluated and then the
lambda is applied to the evaluated arguments. If this is not wanted in some rare cases, the
programmer also has the possibility to define a special form using special-lambda, which, when
invoked does not evaluare any argument. The programmer has to evaluate the arguments in the body
themselves using eval. The rest of the syntax between lambda and special-lambda are the same.
{{{slime_header}}}
((lambda (x) (printf x)) (+ 1 2))
((special-lambda (x) (printf x)) (+ 1 2))
;; Special lambdas make it possible to write
;; code that inspects code
((special-lambda (expr)
(printf "The function to be called is"
(first expr)
"and the result is"
(eval expr)))
(+ 1 2))evaluates to => 3 (+ 1 2) The function to be called is + and the result is 3.
Define
To assign a value to a symbol you can use the define built-in special form. The syntax for
define is: \[\texttt{(define symbol value)}\] and some usages can be seen in
code:variable-defines.
{{{slime_header}}}
(define var1 1)
(define var2 "Hello World")
(define var3 (+ 1 2))
(printf var1 var2 var3)evaluates to => 1 Hello World 3
Defining functions
In Lambdas we learned how to create function objects using the lambda built-in form. Using
define every Lisp Object can be assigned to a symbol making no exception for the function objects.
In code:lambda-defines you can see what that would look like.
{{{slime_header}}}
(define hypothenuse
(lambda (a b)
(** (+ (* a a) (* b b)) 0.5)))
(printf (hypothenuse 3 4))evaluates to => 5
Since defining functions is so common, there is a syntactic shorthand that does not require to write
out the whole lambda definition. In this case the first argument to the call to define is a
list. The frist element of the list is the name of the function to define and the other elemens are
the arguments to that function. An example can be seen in code:function-defines. Note that the
definition looks like a call to the function we are constructing, making it easier to see what a
call to that function will look like.
{{{slime_header}}}
(define (hypothenuse a b)
(** (+ (* a a) (* b b)) 0.5))
(printf (hypothenuse 3 4))evaluates to => 5
Functions with keyword arguments
A sometimes more convenient way of passing arguments to a function is using keyword arguments. Using
keyword arguments a function call could look like this: \[\texttt{(function :arg1 value1 :arg2
value2)}\] here the function accepts two arguments named arg1 and arg2. The user of this
function can see more clearly excatly which argument will be assigned wich value. This notation also
allows for switching the argument order. The following function call is equivalent to the call
above. \[\texttt{(function :arg2 value2 :arg1 value1)}\].
For this to work however, the function must be defined to accept these keyword arguments. To do this
the special marker :keys has to be inserted into the argument list of a lambda or a function
define. All following arguments must be supplied as keyword arguments, unless they are also
supplied with a default value, in which case they do not need to be supplied. To attach a default
value to a keyword argument, insert :defaults-to <value> after the keyword argument name. An
example of all of this can be seen in code:keyword-args. Important note: keyword arguments must be
defined and supplied after all the regular arguments.
{{{slime_header}}}
(define (complex required1 required2 :keys
key1
key2 :defaults-to 3
key3)
(* (+ required1 required2)
key1
key2
key3))
(printf (complex 1 2 :key1 2 :key2 2 :key3 3))
(printf (complex 1 2 :key1 2 :key3 3))
(printf (complex 1 2 :key3 3 :key1 2))evaluates to => 36 54 54
Functions with rest arguments
If the programmer wants to create a function that can accept any number of arguments, they can use
the rest argument. It is defined after the special marker :rest and after the rest argument, no
other arguments can be defined. In the execution of the fuction, the rest arguent will be assigned
to a list containing all the supplied values. The rest argument can be used in conjunction with the
other argument types, regular arguments and keyword arguments.
{{{slime_header}}}
(define (execute-operation operation
:keys
do-logging :defaults-to ()
:rest values)
(define result (apply operation values))
(when do-logging
(printf "Executing operation"
operation
"agains the values"
values
"yielded:"
result))
result)
(printf (execute-operation '+ 1 2 3))
(printf (execute-operation '*
:do-logging t
10 11))evaluates to => 6 Executing operation * agains values (10 11) yielded: 110 110