dpans/concept-classes.tex

% -*- Mode: TeX -*-
%% Classes

While the \CLOS\ is general enough to describe all \term{standardized} \term{classes}
(including, for example, \typeref{number}, \typeref{hash-table}, and
\typeref{symbol}), \thenextfigure\ contains a list of \term{classes} that are
especially relevant to understanding the \CLOS.

\DefineFigure{ObjectSystemClasses}
\displaythree{Object System Classes}{
built-in-class&method-combination&standard-object\cr
class&standard-class&structure-class\cr
generic-function&standard-generic-function&structure-object\cr
method&standard-method&\cr
}

\beginsubSection{Introduction to Classes}

A \newterm{class} is an \term{object} that determines the structure and behavior 
of a set of other \term{objects}, which are called its \newtermidx{instances}{instance}.   

A \term{class} can inherit structure and behavior from other \term{classes}.
A \term{class} whose definition refers to other \term{classes} for the purpose 
of inheriting from them is said to be a \term{subclass} of each of
those \term{classes}. The \term{classes} that are designated for purposes of
inheritance are said to be \term{superclasses} of the inheriting \term{class}.
                                              
A \term{class} can have a \term{name}. The \term{function} \funref{class-name} 
takes a \term{class} \term{object} and returns its \term{name}. 
The \term{name} of an anonymous \term{class} is \nil.  A \term{symbol} 
can \term{name} a \term{class}. The \term{function} \funref{find-class} takes a
\term{symbol} and returns the \term{class} that the \term{symbol} names.
A \term{class} has a \term{proper name} if the \term{name} is a \term{symbol}
and if the \term{name} of the \term{class} names that \term{class}.
That is, a \term{class}~$C$ has the \term{proper name}~$S$ if $S=$
\f{(class-name $C$)} and $C=$ \f{(find-class $S$)}.
Notice that it is possible for 
\f{(find-class $S\sub 1$)} $=$ \f{(find-class $S\sub 2$)}
and $S\sub 1\neq S\sub 2$.
If $C=$ \f{(find-class $S$)}, we say that $C$ is the \term{class} \term{named} $S$.

A \term{class} $C\sub{1}$ is 
a \newterm{direct superclass} of a \term{class} $C\sub{2}$
if $C\sub{2}$ explicitly designates $C\sub{1}$ 
as a \term{superclass} in its definition.
In this case $C\sub{2}$ is a \newterm{direct subclass} of $C\sub{1}$.
A \term{class} $C\sub{n}$ is a \newterm{superclass} of 
a \term{class} $C\sub{1}$ if there exists a series of
\term{classes} $C\sub{2},\ldots,C\sub{n-1}$ such that 
$C\sub{i+1}$ is a \term{direct superclass} of $C\sub{i}$ for $1 \leq i<n$.
In this case, $C\sub{1}$ is a \newterm{subclass} of $C\sub{n}$.
A \term{class} is considered neither a \term{superclass} nor a \term{subclass} of itself.
That is, if $C\sub{1}$ is a \term{superclass} of $C\sub{2}$, 
then $C\sub{1} \neq C\sub{2}$.
The set of \term{classes} consisting of some given \term{class} $C$ 
along with all of its \term{superclasses} is called ``$C$ and its superclasses.''

Each \term{class} has a \newterm{class precedence list},
which is a total ordering on the set of the given \term{class} and its \term{superclasses}.
The total ordering is expressed as a list ordered from most specific to least specific.
The \term{class precedence list} is used in several ways.  In general, more
specific \term{classes} can \newterm{shadow}\meaning{1} features that would
otherwise be inherited from less specific \term{classes}.
The \term{method} selection and combination process uses 
the \term{class precedence list} to order \term{methods} 
from most specific to least specific. 
 
When a \term{class} is defined, the order in which its direct \term{superclasses}
are mentioned in the defining form is important.  Each \term{class} has a
\newterm{local precedence order}, which is a \term{list} consisting of the
\term{class} followed by its \term{direct superclasses} in the order mentioned
in the defining \term{form}.

A \term{class precedence list} is always consistent with the
\term{local precedence order} of each \term{class} in the list.  
The \term{classes} in each \term{local precedence order} appear
within the \term{class precedence list} in the same order.  
If the \term{local precedence orders} are inconsistent with each other, 
no \term{class precedence list} can be constructed, and an error is signaled.
The \term{class precedence list} and its computation is discussed
in \secref\DeterminingtheCPL.

\term{classes} are organized into a directed acyclic graph.
There are two distinguished \term{classes}, named \typeref{t} and \typeref{standard-object}.
The \term{class} named \typeref{t} has no \term{superclasses}. 
It is a \term{superclass} of every \term{class} except itself.  
The \term{class} named \typeref{standard-object} is an \term{instance} of 
\theclass{standard-class} and is a \term{superclass} of
every \term{class} that is an \term{instance} of \theclass{standard-class} except itself.

\reviewer{Barmar: This or something like it needs to be said in the introduction.}%!!!
There is a mapping from the object system \term{class} space into
the \term{type} space.  Many of the standard \term{types} specified 
in this document have a corresponding \term{class} that has the same 
\term{name} as the \term{type}. Some \term{types} do not have a
corresponding \term{class}. The integration of the \term{type} and \term{class}
systems is discussed in \secref\IntegratingTypesAndClasses.

\term{Classes} are represented by \term{objects} that are themselves
\term{instances} of \term{classes}. 
The \term{class} of the \term{class} of an \term{object} is termed
the \newterm{metaclass} of that \term{object}. When no misinterpretation is
possible, the term \term{metaclass} is used to refer to a \term{class}
that has \term{instances} that are themselves \term{classes}. The \term{metaclass}
determines the form of inheritance used by the \term{classes} that are its
\term{instances} and the representation of the \term{instances} of those \term{classes}.
The \CLOS\ provides a default \term{metaclass}, \typeref{standard-class}, that is
appropriate for most programs.
%The meta-object protocol provides
%mechanisms for defining and using new metaclasses.

Except where otherwise specified, all \term{classes} mentioned in this
standard are \term{instances} of \theclass{standard-class},
all \term{generic functions} are \term{instances} 
of \theclass{standard-generic-function},
and all \term{methods} are \term{instances} of \theclass{standard-method}.


\endsubSection%{Classes}
%\beginsubsubsection{Metaclasses}
%??? Is the following paragraph necessary in this specification???
%
%The \newterm{metaclass} of an object is the class of its class.  The
%metaclass determines the representation of instances of its instances and
%the forms of inheritance used by its instances for slot descriptions and
%method inheritance.  The metaclass mechanism can be used to provide
%particular forms of optimization or to tailor the \CLOS\ for particular
%uses.  
%The protocol for defining metaclasses is discussed in the chapter
%``The \CLOS\ Meta-Object Protocol.''

%\endsubsubsection%{Metaclasses}

\beginsubsubsection{Standard Metaclasses}

The \CLOS\ provides a number of predefined \term{metaclasses}. 
These include the \term{classes} \typeref{standard-class}, 
\typeref{built-in-class}, and \typeref{structure-class}:

\beginlist

\itemitem{\bull}
\Theclass{standard-class} is the default \term{class} of 
\term{classes} defined by \macref{defclass}.
                        
\itemitem{\bull} \Theclass{built-in-class} is the \term{class} whose
\term{instances} are \term{classes} that have special implementations with
restricted capabilities.  Any \term{class} that corresponds to a standard
\term{type} might be an \term{instance} of \typeref{built-in-class}.
The predefined \term{type} specifiers that are required to have
corresponding \term{classes} are listed in \figref\ClassTypeCorrespondence.  
It is \term{implementation-dependent} whether each of these \term{classes} 
is implemented as a \term{built-in class}.

\itemitem{\bull}                     
All \term{classes} defined by means of \macref{defstruct} are
\term{instances} of \theclass{structure-class}.
\endlist

\endsubsubsection%{Standard Metaclasses}

\beginsubSection{Defining Classes}
           
The macro \macref{defclass} is used to define a new named \term{class}.  
%The syntax for \macref{defclass} is given in Figure ??

The definition of a \term{class} includes:

\beginlist

\itemitem{\bull} The \term{name} of the new \term{class}. 
  For newly-defined \term{classes} this \term{name} is a \term{proper name}.

\itemitem{\bull} The list of the direct \term{superclasses} of the new \term{class}. 

\itemitem{\bull} A set of \newtermidx{slot specifiers}{slot specifier}.
  Each \term{slot specifier} includes the \term{name} of the \term{slot} 
  and zero or more \term{slot} options.  A \term{slot} option pertains 
  only to a single \term{slot}.  If a \term{class} definition contains
  two \term{slot specifiers} with the same \term{name}, an error is signaled.

\itemitem{\bull} A set of \term{class} options.  
  Each \term{class} option pertains to the \term{class} as a whole.  

\endlist
                                              
The \term{slot} options and \term{class} options of 
the \macref{defclass} form provide mechanisms for the following:

\beginlist

\itemitem{\bull} Supplying a default initial value \term{form} 
for a given \term{slot}.  

\itemitem{\bull} Requesting that \term{methods} for \term{generic functions}
be automatically generated for reading or writing \term{slots}. 

\itemitem{\bull} Controlling whether a given \term{slot} is shared by 
all \term{instances}
of the \term{class} or whether each 
\term{instance} of the \term{class} has its own \term{slot}.

\itemitem{\bull} Supplying a set of initialization arguments and initialization
argument defaults to be used in \term{instance} creation.

%\itemitem{\bull} Requesting that a constructor function be automatically
%generated for making instances of the new class.

\itemitem{\bull} Indicating that the \term{metaclass} is to be other 
than the default.  The \kwd{metaclass} option is reserved for future use; 
an implementation can be extended to make use of the \kwd{metaclass}
option.

\itemitem{\bull} Indicating the expected \term{type} for the value stored
in the \term{slot}.

\itemitem{\bull} Indicating the \term{documentation string} for the \term{slot}.

\endlist 

\endsubSection%{Defining Classes}

\goodbreak

\beginsubSection{Creating Instances of Classes}
                      
The generic function \funref{make-instance} creates and returns a new
\term{instance} of a \term{class}.  
The \OS\ provides several mechanisms for
specifying how a new \term{instance} is to be initialized.  For example, it
is possible to specify the initial values for \term{slots} in newly created
\term{instances} 
either by giving arguments to \funref{make-instance} or by
providing default initial values.  Further initialization activities
can be performed by \term{methods} written for \term{generic functions} 
that are
part of the initialization protocol.  The complete initialization
protocol is described in \secref\ObjectCreationAndInit.

\endsubSection%{Creating Instances of Classes}

\beginsubSection{Inheritance}
\DefineSection{Inheritance}
                                              
A \term{class} can inherit \term{methods}, \term{slots}, 
and some \macref{defclass} options from its \term{superclasses}.  
Other sections describe the inheritance of \term{methods}, 
the inheritance of \term{slots} and \term{slot} options, 
and the inheritance of \term{class} options.
 

\beginsubsubsection{Examples of Inheritance}

\code
 (defclass C1 () 
     ((S1 :initform 5.4 :type number) 
      (S2 :allocation :class)))
 
 (defclass C2 (C1) 
     ((S1 :initform 5 :type integer)
      (S2 :allocation :instance)
      (S3 :accessor C2-S3)))
\endcode

\term{Instances} of the class \f{C1} have a \term{local slot} named \f{S1},
whose default initial value is 5.4 and
whose \term{value} should always be a \term{number}.
The class \f{C1} also has a \term{shared slot} named \f{S2}.

There is a \term{local slot} named \f{S1} in \term{instances} of \f{C2}.
The default initial value of \f{S1} is 5.
The value of \f{S1} should always be of type \f{(and integer number)}.
There are also \term{local slots} named \f{S2} and \f{S3} in \term{instances} of \f{C2}.
The class \f{C2} has a \term{method} for \f{C2-S3} for reading the value of slot \f{S3};
there is also a \term{method} for \f{(setf C2-S3)} that writes the value of \f{S3}.

\endsubsubsection%{Examples of Inheritance}

\beginsubsubsection{Inheritance of Class Options}
     
The \kwd{default-initargs} class option is inherited.  The set of
defaulted initialization arguments for a \term{class} is the union of the
sets of initialization arguments supplied in
the \kwd{default-initargs} class options of the \term{class} and its \term{superclasses}.
When more than one default initial value \term{form} is supplied for a given
initialization argument, the default initial value \term{form} that is used
is the one supplied by the \term{class} that is most specific according to
the \term{class precedence list}.


If a given \kwd{default-initargs} class option specifies an
initialization argument of the same \term{name} more than once, an
error \oftype{program-error} is signaled.

\endsubsubsection%{Inheritance of Class Options}


\endsubSection%{Inheritance}


\beginsubSection{Determining the Class Precedence List}
\DefineSection{DeterminingtheCPL}

The \macref{defclass} form for a \term{class} provides a total ordering
on that \term{class} and its direct \term{superclasses}.  This ordering is
called the \newterm{local precedence order}.  It is an ordered list of the
\term{class} and its direct \term{superclasses}. The
\newterm{class precedence list} for a class $C$ is a total ordering on
$C$ and its \term{superclasses} that is consistent with the
\term{local precedence orders} for each of $C$ and its \term{superclasses}.

A \term{class} precedes its direct \term{superclasses}, 
and a direct \term{superclass} precedes all other 
direct \term{superclasses} specified to its right 
in the \term{superclasses} list of the \macref{defclass} form.  
For every class $C$, define $$R\sub C=\{(C,C\sub 1),(C\sub 1,C\sub
2),\ldots,(C\sub {n-1},C\sub n)\}$$ where $C\sub 1,\ldots,C\sub n$ are
the direct \term{superclasses} of $C$ in the order in which
they are mentioned in the \macref{defclass} form. These ordered pairs
generate the total ordering on the class $C$ and its direct
\term{superclasses}.

Let $S\sub C$ be the set of $C$ and its \term{superclasses}. Let $R$ be
$$R=\bigcup\sub{c\in {S\sub C}}R\sub c$$.

\reviewer{Barmar: ``Consistent'' needs to be defined, or maybe we should say
``logically consistent''?}%!!!

The set $R$ might or might not generate a partial ordering, depending on
whether the $R\sub c$, $c\in S\sub C$, are 
consistent; it is assumed
that they are consistent and that $R$ generates a partial ordering.
When the $R\sub c$ are not consistent, it is said that $R$ is inconsistent.

%This partial ordering is generated by taking the the transitive
%closure of the set $R\cup \{(c,c) \vert c\in {S\sub C}\}$.  When
%$(C\sub 1,C\sub 2)\in R$\negthinspace, it is said that $C\sub 1$ 
%{\bit precedes or equals} $C\sub 2$.  Intuitively, $C\sub 1$ precedes
%or equals $C\sub 2$ when $C\sub 1=C\sub 2$ or $C\sub 2$ is a
%superclass of $C\sub 1$.  
%
%Recall that a partial ordering of the set $S$ is a relation between 
%objects of $S$ that is transitive, reflexive, and antisymmetric. The 
%set $\{(c,c) \vert c\in {S\subC}\}$ was added to the transitive 
%closure of the set $R$ in order to make the relation reflexive.  In 
%the remainder of this section the set of precedence relations $R$ and 
%not the partial ordering will be used.

To compute the \term{class precedence list} for~$C$\negthinspace,
topologically sort the elements of $S\sub C$ with respect to the
partial ordering generated by $R$\negthinspace.  When the topological
sort must select a \term{class} from a set of two or more 
\term{classes}, none of
which are preceded by other \term{classes} with respect to~$R$\negthinspace,
the \term{class} selected is chosen deterministically, as described below.

If $R$ is inconsistent, an error is signaled.


\goodbreak

\beginsubsubsection{Topological Sorting}

% !!!
% Barmar: Sometimes $C$ refers  to the original class and sometimes it refers
%         to an element of $S\sub C$.
% KAB: Really? Where?  I don't see any.

%I didn't have time to figure out this comment but maybe someone will another time.
% -kmp 11-Oct-90
Topological sorting proceeds by finding a class $C$ in~$S\sub C$ such
that no other \term{class} precedes that element according to the elements
in~$R$\negthinspace.  The class $C$ is placed first in the result.
Remove $C$ from $S\sub C$, and remove all pairs of the form $(C,D)$,
$D\in S\sub C$, from $R$\negthinspace. Repeat the process, adding
\term{classes} with no predecessors to the end of the result.  Stop when no
element can be found that has no predecessor.

If $S\sub C$ is not empty and the process has stopped, the set $R$ is
inconsistent. If every \term{class} in the finite set of 
\term{classes} is preceded
by another, then $R$ contains a loop. That is, there is a chain of
classes $C\sub 1,\ldots,C\sub n$ such that $C\sub i$ precedes
$C\sub{i+1}$, $1\leq i<n$, and $C\sub n$ precedes $C\sub 1$.

Sometimes there are several \term{classes} from $S\sub C$ with no
predecessors.  In this case select the one that has a direct
\term{subclass} rightmost in the \term{class precedence list} computed so far.
%Because a direct superclass precedes all other direct superclasses to
%its right, there can be only one such candidate class. 
%%Barmar thinks the next sentence is redundant with previous paragraph,
%but I think it's useful for emphasis. I put it in parens to deemphasize it. -kmp 11-Oct-90
(If there is no such candidate \term{class}, $R$ does not generate 
a partial ordering---the $R\sub c$, $c\in S\sub C$, are inconsistent.)

In more precise terms, let $\{N\sub 1,\ldots,N\sub m\}$, $m\geq 2$, be
the \term{classes} from $S\sub C$ with no predecessors.  Let $(C\sub
1\ldots C\sub n)$, $n\geq 1$, be the \term{class precedence list}
constructed so far.  $C\sub 1$ is the most specific \term{class}, and $C\sub
n$ is the least specific.  Let $1\leq j\leq n$ be the largest number
such that there exists an $i$ where $1\leq i\leq m$ and $N\sub i$
is a direct \term{superclass} of $C\sub j$; $N\sub i$ is placed next.

% The effect of this rule for selecting from a set of \term{classes} with no
% predecessors is that the \term{classes} in a simple \term{superclass} chain are
% adjacent in the \term{class precedence list} and that \term{classes} in each
% relatively separated subgraph are adjacent in the \term{class precedence list}.
% For example, let $T\sub 1$ and $T\sub 2$ be subgraphs
% whose only element in common is the class $J$\negthinspace. Suppose
% that no \term{superclass} of $J$ appears in either $T\sub 1$ or $T\sub 2$.
% Let $C\sub 1$ be the bottom of $T\sub 1$; and let $C\sub 2$ be the
% bottom of $T\sub 2$.  Suppose $C$ is a \term{class} whose direct \term{superclasses}
% are $C\sub 1$ and $C\sub 2$ in that order, then the 
% \term{class precedence list}
% for $C$ will start with $C$ and will be followed by all \term{classes}
% in $T\sub 1$ except $J$. All the \term{classes} of $T\sub 2$ will be next.
% The class $J$ and its \term{superclasses} will appear last.
%
%% Replaced as follows per RPG. -kmp 13-Jan-92

The effect of this rule for selecting from a set of \term{classes} with no
predecessors is that the \term{classes} in a simple \term{superclass} chain are
adjacent in the \term{class precedence list} and that \term{classes} in each
relatively separated subgraph are adjacent in the \term{class precedence list}.
For example, let $T\sub 1$ and $T\sub 2$ be subgraphs whose only
element in common is the class $J$\negthinspace.
% Suppose that no \term{superclass} of $J$ appears in 
% either $T\sub 1$ or $T\sub 2$ and that $J$ 
% is an indirect superclass of every class in both $T\sub 1$ and $T\sub 2$.
%% Rewritten by RPG to avoid use of "indirect superclass". -kmp 13-Jan-92
Suppose that no superclass of $J$ appears in either $T\sub 1$ or $T\sub 2$,
and that $J$ is in the superclass chain of every class in both $T\sub 1$ and $T\sub 2$.
    Let $C\sub 1$ be the bottom of $T\sub 1$; 
and let $C\sub 2$ be the bottom of $T\sub 2$.
Suppose $C$ is a \term{class} whose direct \term{superclasses}
are $C\sub 1$ and $C\sub 2$ in that order, then the \term{class precedence list}
for $C$ starts with $C$ and is followed by
all \term{classes} in $T\sub 1$ except $J$. 
All the \term{classes} of $T\sub 2$ are next.
The \term{class} $J$ and its \term{superclasses} appear last.

\endsubsubsection%{Topological Sorting}

\beginsubsubsection{Examples of Class Precedence List Determination}

This example determines a \term{class precedence list} for the
class \f{pie}.  The following \term{classes} are defined:

\code
 (defclass pie (apple cinnamon) ())
 
 (defclass apple (fruit) ())
 
 (defclass cinnamon (spice) ())
 
 (defclass fruit (food) ())

 (defclass spice (food) ())

 (defclass food () ())
\endcode

% $S$ => "$S\sub{pie}$ per Barmar and Laubsch.
The set $S\sub{pie}$~$=$ $\{${\tt pie, apple, cinnamon, fruit, spice, food,
standard-object, t}$\}$. The set $R$~$=$ $\{${\tt (pie, apple),
(apple, cinnamon), (apple, fruit), (cinnamon, spice), \hfil\break
(fruit, food), (spice, food), (food, standard-object), (standard-object,
t)}$\}$.

The class \f{pie} is not preceded by anything, so it comes first;
the result so far is {\tt (pie)}.  Remove \f{pie} from $S$ and pairs
mentioning \f{pie} from $R$ to get $S$~$=$ $\{${\tt apple, cinnamon,
fruit, spice, food, standard-object, t}$\}$ and $R$~$=$~$\{${\tt
(apple, cinnamon), (apple, fruit), (cinnamon, spice),\hfil\break (fruit,
food), (spice, food), (food, standard-object),
(standard-object, t)}$\}$.

The class \f{apple} is not preceded by anything, so it is next; the
result is {\tt (pie apple)}. Removing \f{apple} and the relevant
pairs results in $S$~$=$ $\{${\tt cinnamon, fruit, spice, food,
standard-object, t}$\}$ and $R$~$=$ $\{${\tt (cinnamon, spice),
(fruit, food), (spice, food), (food, standard-object),\hfil\break
(standard-object, t)}$\}$.

The classes \f{cinnamon} and {\tt fruit} are not preceded by
anything, so the one with a direct \term{subclass} rightmost in the 
\term{class precedence list} computed so far goes next.  The class \f{apple} is a
direct \term{subclass} of {\tt fruit}, and the class \f{pie} is a direct
\term{subclass} of \f{cinnamon}.  Because \f{apple} appears to the right
of \f{pie} in the \term{class precedence list}, 
{\tt fruit} goes next, and the
result so far is {\tt (pie apple fruit)}.  $S$~$=$ $\{${\tt cinnamon,
spice, food, standard-object, t}$\}$; $R$~$=$ $\{${\tt (cinnamon,
spice), (spice, food),\hfil\break (food, standard-object),
(standard-object, t)}$\}$.

The class \f{cinnamon} is next, giving the result so far as {\tt
(pie apple fruit cinnamon)}.  At this point $S$~$=$ $\{${\tt spice,
food, standard-object, t}$\}$; $R$~$=$ $\{${\tt (spice, food), (food,
standard-object), (standard-object, t)}$\}$.

The classes \f{spice}, \f{food}, \typeref{standard-object}, and 
\typeref{t} are added in that order, and the \term{class precedence list} 
is \f{(pie apple fruit cinnamon spice food standard-object t)}.

It is possible to write a set of \term{class} definitions that cannot be 
ordered.   For example: 

\code
 (defclass new-class (fruit apple) ())
 
 (defclass apple (fruit) ())
\endcode

The class \f{fruit} must precede \f{apple} 
because the local ordering of \term{superclasses} must be preserved.
The class \f{apple} must precede \f{fruit} 
because a \term{class} always precedes its own \term{superclasses}.
When this situation occurs, an error is signaled, as happens here
when the system tries to compute the \term{class precedence list} 
%Barmar suggested we add:
of \f{new-class}.

The following might appear to be a conflicting set of definitions:

\code
 (defclass pie (apple cinnamon) ())
 
 (defclass pastry (cinnamon apple) ())
 
 (defclass apple () ())
 
 (defclass cinnamon () ())
\endcode

The \term{class precedence list} for \f{pie} is 
\f{(pie apple cinnamon standard-object t)}.

The \term{class precedence list} for \f{pastry} is  
\f{(pastry cinnamon apple standard-object t)}.

It is not a problem for \f{apple} to precede \f{cinnamon} in the
ordering of the \term{superclasses} of \f{pie} but not in the ordering for
\f{pastry}.  However, it is not possible to build a new \term{class} that
has both \f{pie} and \f{pastry} as \term{superclasses}.

\endsubsubsection%{Examples of Class Precedence List Determination}

\endsubSection%{Determining the Class Precedence List}
\beginsubSection{Redefining Classes}  
\DefineSection{ClassReDef}
                                
%"instance" => "direct instance" per Barrett,Barmar,Moon.
A \term{class} that is a \term{direct instance} of \typeref{standard-class} can
be redefined if the new \term{class} is also
%"instance" => "direct instance" per Barrett,Barmar,Moon.
a \term{direct instance} of \typeref{standard-class}.
Redefining a \term{class} modifies the existing
\term{class} \term{object} to reflect the new \term{class} definition; it does not
create a new \term{class} \term{object} for the \term{class}.  
Any \term{method} \term{object} created by a \kwd{reader}, \kwd{writer}, 
or \kwd{accessor} option specified by the old \macref{defclass} form is
removed from the corresponding \term{generic function}.
\term{Methods} specified by the new \macref{defclass} form are added.

% any function specified by the \kwd{constructor} option of the
% old \macref{defclass} form is removed from the corresponding symbol function cell.

When the class $C$ is redefined, changes are propagated to its \term{instances}
and to \term{instances} of any of its \term{subclasses}.  Updating such an
\term{instance} occurs at an \term{implementation-dependent} time, but no later than
the next time a \term{slot} 
of that \term{instance} is read or written.  Updating an
\term{instance} 
does not change its identity as defined by \thefunction{eq}.
The updating process may change the \term{slots} of that
particular \term{instance}, 
but it does not create a new \term{instance}.  Whether
updating an \term{instance} consumes storage is \term{implementation-dependent}.

Note that redefining a \term{class} may cause \term{slots} to be added or 
deleted.  If a \term{class} is redefined in a way that changes the set of
\term{local slots} \term{accessible} in \term{instances}, the \term{instances} 
are updated.  It is \term{implementation-dependent} whether \term{instances} 
are updated if a \term{class} is redefined in a way that does not change 
the set of \term{local slots} \term{accessible} in \term{instances}.

The value of a \term{slot} 
that is specified as shared both in the old \term{class}
and in the new \term{class} is retained.  
If such a \term{shared slot} was unbound
in the old \term{class}, it is unbound in the new \term{class}.  
\term{Slots} that
were local in the old \term{class} and that are shared in the new 
\term{class} are
initialized.  Newly added \term{shared slots} are initialized.

Each newly added \term{shared slot} is set to the result of evaluating the
%captured \kwd{initform} form
\term{captured initialization form} for the \term{slot} that was specified 
in the \macref{defclass} \term{form} for the new \term{class}.  
%If there is no \kwd{initform} form,
If there was no \term{initialization form}, the \term{slot} is unbound.

If a \term{class} is redefined in such a way that the set of
\term{local slots} \term{accessible} in an \term{instance} of the \term{class} 
is changed, a two-step process of updating the \term{instances} of the
\term{class} takes place.  The process may be explicitly started by 
invoking the generic function \funref{make-instances-obsolete}.  This
two-step process can happen in other circumstances in some implementations.
For example, in some implementations this two-step process is
triggered if the order of \term{slots} in storage is changed.

The first step modifies the structure of the \term{instance} by adding new
\term{local slots} and discarding \term{local slots} that are not
defined in the new version of the \term{class}.  The second step
initializes the newly-added \term{local slots} and performs any other
user-defined actions. These two steps are further specified
in the next two sections.

\beginsubsubsection{Modifying the Structure of Instances}

\reviewer{Barmar: What about shared slots that are deleted?}%!!!

The first step modifies the structure of \term{instances} of the redefined
\term{class} to conform to its new \term{class} definition.  
\term{Local slots} specified
by the new \term{class} definition that are not specified as either local or
shared by the old \term{class} are added, and \term{slots} 
not specified as either
local or shared by the new \term{class} definition that are specified as
local by the old \term{class} are discarded. 
The \term{names} of these added and discarded
\term{slots} are passed as arguments 
to \funref{update-instance-for-redefined-class}
as described in the next section.

The values of \term{local slots} specified by both the new and old
\term{classes} are retained. If such a \term{local slot} was unbound,
it remains unbound.

The value of a \term{slot} that is specified as shared in the old 
\term{class} and as local in the new \term{class} is retained.  If such 
a \term{shared slot} was unbound, the \term{local slot} is unbound.

\endsubsubsection%{Modifying the Structure of the Instance}


\beginsubsubsection{Initializing Newly Added Local Slots}

The second step initializes the newly added \term{local slots} and performs
any other user-defined actions.  This step is implemented by the generic
function \funref{update-instance-for-redefined-class}, which is called after
completion of the first step of modifying the structure of the
\term{instance}.

The generic function \funref{update-instance-for-redefined-class} takes
four required arguments: the \term{instance} being updated after it has
undergone the first step, a list of the names of \term{local slots} that were
added, a list of the names of \term{local slots} that were discarded, and a
property list containing the \term{slot} names and values of 
\term{slots} that were
discarded and had values.  Included among the discarded \term{slots} are
\term{slots} that were local in the old \term{class} and that are shared in the new
\term{class}.
                      
The generic function \funref{update-instance-for-redefined-class} also
takes any number of initialization arguments.  When it is called by
the system to update an \term{instance} whose \term{class} 
has been redefined, no
initialization arguments are provided.
                                               
There is a system-supplied primary \term{method} for 
\funref{update-instance-for-redefined-class} whose \term{parameter specializer}
for its \term{instance} argument is \theclass{standard-object}.  
First this \term{method} checks the validity of initialization arguments and signals an
error if an initialization argument is supplied that is not declared
as valid.  (For more information, \seesection\DeclaringInitargValidity.)
Then it calls the generic function
\funref{shared-initialize} with the following arguments: the 
\term{instance},
the list of \term{names} of 
the newly added \term{slots}, and the initialization
arguments it received.

\endsubsubsection%{Initializing Newly added Local Slots}

\beginsubsubsection{Customizing Class Redefinition}
             
\reviewer{Barmar: This description is hard to follow.}%!!!

\term{Methods} for \funref{update-instance-for-redefined-class} may be 
defined to specify actions to be taken when an \term{instance} is updated.
If only \term{after methods} for \funref{update-instance-for-redefined-class} are
defined, they will be run after the system-supplied primary \term{method} for
initialization and therefore will not interfere with the default
behavior of \funref{update-instance-for-redefined-class}.  Because no
initialization arguments are passed to \funref{update-instance-for-redefined-class}
when it is called by the system, the 
%\kwd{initform} forms 
\term{initialization forms} for \term{slots} 
that are filled by \term{before methods} for \funref{update-instance-for-redefined-class} 
will not be evaluated by \funref{shared-initialize}.

\term{Methods} for \funref{shared-initialize} may be defined to customize
\term{class} redefinition.  For more information, \seesection\SharedInitialize.

\endsubsubsection%{Customizing Class Redefinition}

%% The following was removed by request of Symbolics, who point out that other 
%% extensions may well be permitted too, and there's no reason to give special
%% advertising to these. -kmp 9-Oct-90
%
% \beginsubsubsection{Extensions}
% 
% There are two allowed extensions to \term{class} redefinition: 
% 
% \beginlist
% 
% \itemitem{\bull} The \OS\ may be extended to permit the new \term{class}
% to be an \term{instance} of a \term{metaclass} 
% other than the \term{metaclass} of the
% old \term{class}.
% 
% \itemitem{\bull} The \OS\ may be extended to support an updating process
% when either the old or the new \term{class} is an \term{instance} of a
% \term{class} 
% other than \typeref{standard-class} that is not a \term{built-in class}.
% 
% \endlist
% 
% \endsubsubsection%{Extensions}

\beginsubSection{Integrating Types and Classes} 
\DefineSection{IntegratingTypesAndClasses}
                                              
The \CLOS\ maps the space of \term{classes} into the space of \term{types}.
Every \term{class} that has a proper name has a corresponding \term{type} 
with the same \term{name}.  

The proper name of every \term{class} is a valid \term{type specifier}.  In
addition, every \term{class} \term{object} is a valid \term{type specifier}.  
Thus the expression \f{(typep \param{object} \param{class})} evaluates to 
\term{true} if the \term{class} of \param{object} is \param{class} itself or 
a \term{subclass} of \term{class}.  The evaluation of the expression
\f{(subtypep class1 class2)} returns the values 
%was {\tt t~t} but came out as "t~t" in formatted form. -kmp 10-Oct-90
%\t~\t\ if \f{class1} is a subclass of \f{class2} or if they are the
\term{true} and \term{true} if \f{class1} is a subclass of \f{class2} or if they are the
same \term{class}; otherwise it returns the values 
%was {\tt nil~t}
%\nil~\t.
\term{false} and \term{true}.
% If $I$ is an \term{instance} of some \term{class}
% $C$ named $S$ and $C$ is an \term{instance} of \typeref{standard-class}, 
% the evaluation of the expression \f{(type-of $I$\/)} will return $S$ if
% $S$ is the proper name of $C$\negthinspace; if $S$ is not the proper
% name of $C$\negthinspace, the expression \f{(type-of $I$\/)} will
% return $C$\negthinspace.
%% Barmar thought the above was too complicated.
%% This is a partial simplification, not quite what he wanted, but hopefully
%% an improvement.
If  $I$ is an \term{instance} of some \term{class} $C$ named $S$ 
and $C$ is an \term{instance} of \typeref{standard-class}, 
the evaluation of the expression \f{(type-of $I$\/)} returns $S$ 
if $S$ is the \term{proper name} of $C$; 
otherwise, it returns $C$.

Because the names of \term{classes} 
and \term{class} \term{objects} are \term{type specifiers}, they may
be used in the special form \specref{the} and in type declarations.
                                   
Many but not all of the predefined \term{type specifiers} have a
corresponding \term{class} with 
the same proper name as the \term{type}.  These type
specifiers are listed in \figref\ClassTypeCorrespondence.
For example, \thetype{array} has 
a corresponding \term{class} named \typeref{array}.  
No \term{type specifier} that is a
list, such as {\tt (vector double-float 100)}, has a corresponding \term{class}.
The \term{operator} \macref{deftype} does not create any \term{classes}.
                                            
Each \term{class} that corresponds to a predefined \term{type specifier} can
be implemented in one of three ways, at the discretion of each implementation.
It can be a \term{standard class},
%% Not necessary. -kmp 12-Feb-92
% (of the kind defined by \macref{defclass}), 
a \term{structure class},
%% Not necessary. -kmp 12-Feb-92
% (defined by \macref{defstruct}), 
\issue{METACLASS-OF-SYSTEM-CLASS:UNSPECIFIED}
%or a \term{built-in class}.
or a \term{system class}.
\endissue{METACLASS-OF-SYSTEM-CLASS:UNSPECIFIED}
%% Not necessary. -kmp 12-Feb-92
% (implemented in a special, non-extensible way).

%"instances" => "generalized instances" per Barmar,Moon.
A \term{built-in class} is one whose \term{generalized instances} have restricted capabilities 
or special representations.  Attempting to use \macref{defclass} to define 
\term{subclasses} of a \typeref{built-in-class} signals an error.
%"instance" => "generalized instance" per Barmar,Moon.
Calling \funref{make-instance} to create a \term{generalized instance} of a 
\term{built-in class} signals an error.  Calling \funref{slot-value} on a
%"instance" => "generalized instance" per Barmar,Moon.
\term{generalized instance} of a \term{built-in class} signals an error.
Redefining a \term{built-in class} or using \funref{change-class} to change
%"instance" => "object" per Barmar,Moon.
the \term{class} of an \term{object} to or from a \term{built-in class} signals an error.
However, \term{built-in classes} can be used as \term{parameter specializers} 
in \term{methods}.
                                        
%The \OS\ specifies that all predefined \term{type specifiers}
%listed in \figref\ClassTypeCorrespondence\ are built-in classes, but a particular
%implementation is allowed to extend the \OS\ to define some of them as
%standard classes or as structure classes.

It is possible to determine whether a \term{class} is a \term{built-in class}
by checking the \term{metaclass}.
A \term{standard class}  is an \term{instance} of \theclass{standard-class},
a \term{built-in class}  is an \term{instance} of \theclass{built-in-class}, and
a \term{structure class} is an \term{instance} of \theclass{structure-class}.
                                
Each \term{structure} \term{type} created by \macref{defstruct} without 
using the \kwd{type} option has a corresponding \term{class}.  
%"instance" => "generalized instance" per Barmar,Moon.
This \term{class} is a \term{generalized instance} of \theclass{structure-class}.  
%A portable program must assume that
%\typeref{structure-class} is a subclass of \typeref{built-in-class} and has the
%same restrictions as built-in classes.  Whether \theclass{structure-class}
%in fact is a subclass of \typeref{built-in-class} is
%\term{implementation-dependent}. 
The \kwd{include} option of \macref{defstruct} creates a direct
\term{subclass} of the \term{class} 
that corresponds to the included \term{structure} 
%Added "type" -kmp
\term{type}.

% I moved the following two paragraphs here from the dictionary entry
% for CLASS.  --sjl 7 Mar 92
\issue{CONDITION-SLOTS:HIDDEN}
It is \term{implementation-dependent} whether \term{slots} are involved in the
operation of \term{functions} defined in this specification
on \term{instances} of \term{classes} defined in this specification,
except when \term{slots} are explicitly defined by this specification.

If in a particular \term{implementation} a \term{class} defined in this specification
has \term{slots} that are not defined by this specfication, the names of these \term{slots}
must not be \term{external symbols} of \term{packages} defined in this specification nor
otherwise \term{accessible} in \thepackage{cl-user}.
\endissue{CONDITION-SLOTS:HIDDEN}

                                                    
The purpose of specifying that many of the standard \term{type specifiers} have a
corresponding \term{class} is to enable users to write \term{methods} that
discriminate on these \term{types}.  \term{Method} selection requires that a 
\term{class precedence list} can be determined for each \term{class}. 

The hierarchical relationships among the \term{type specifiers} are mirrored by
relationships among the \term{classes} corresponding to those \term{types}.  
%The existing type hierarchy is used for determining the
%\term{class precedence list} 
%for each \term{class} that corresponds to a predefined
%\term{type}.  In some cases, 
%a \term{local precedence order} is not specifiied for two \term{supertypes} 
%of a
%given \term{type specifier}.  For example, \typeref{null} is a \term{subtype} 
%of both \typeref{symbol} and \typeref{list}, but it is not specified
%whether \typeref{symbol} is more specific or less
%specific than \typeref{list}.  The \CLOS\ defines those
%relationships for all such \term{classes}.
                 
\figref\ClassTypeCorrespondence\ lists the set of \term{classes} 
that correspond to predefined \term{type specifiers}.

\issue{REAL-NUMBER-TYPE:X3J13-MAR-89}
\DefineFigure{ClassTypeCorrespondence}
\displaythree{Classes that correspond to pre-defined type specifiers}{
arithmetic-error&generic-function&simple-error\cr
array&hash-table&simple-type-error\cr
bit-vector&integer&simple-warning\cr
broadcast-stream&list&standard-class\cr
built-in-class&logical-pathname&standard-generic-function\cr
cell-error&method&standard-method\cr
character&method-combination&standard-object\cr
class&null&storage-condition\cr
complex&number&stream\cr
concatenated-stream&package&stream-error\cr
condition&package-error&string\cr
cons&parse-error&string-stream\cr
control-error&pathname&structure-class\cr
division-by-zero&print-not-readable&structure-object\cr
echo-stream&program-error&style-warning\cr
end-of-file&random-state&symbol\cr
error&ratio&synonym-stream\cr
file-error&rational&t\cr
file-stream&reader-error&two-way-stream\cr
float&readtable&type-error\cr
floating-point-inexact&real&unbound-slot\cr
floating-point-invalid-operation&restart&unbound-variable\cr
floating-point-overflow&sequence&undefined-function\cr
floating-point-underflow&serious-condition&vector\cr
function&simple-condition&warning\cr
}
\endissue{REAL-NUMBER-TYPE:X3J13-MAR-89}

The \term{class precedence list} information specified in the entries for
each of these \term{classes} are those that are required by the \OS.

Individual implementations may be extended to define other type
specifiers to have a corresponding \term{class}.  Individual implementations
may be extended to add other \term{subclass} relationships and to add other
\term{elements} to the \term{class precedence lists} as long as
they do not violate the type relationships and disjointness
requirements specified by this standard.
A standard \term{class} defined with no direct \term{superclasses} is guaranteed to
be disjoint from all of the \term{classes} in the table, except for the
class named \typeref{t}.
 
\endsubSection%{Integrating Types and Classes}