path: root/doc/refman/Setoid.tex

\newtheorem{cscexample}{Example}

\achapter{\protect{User defined equalities and relations}}
\aauthor{Claudio Sacerdoti Coen\footnote{Based on previous work by
Cl\'ement Renard}}
\label{setoid_replace}
\tacindex{setoid\_replace}

This chapter presents the extension of several equality related tactics to
work over user-defined structures (called setoids) that are equipped with
ad-hoc equivalence relations meant to behave as equalities.
Actually, the tactics have also been generalized to relations weaker then
equivalences (e.g. rewriting systems).

The work generalizes, and is partially based on, a previous implementation of
the \texttt{setoid\_replace} tactic by Cl\'ement Renard.

\asection{Relations and morphisms}

A parametric \emph{relation} \texttt{R} is any term of type
\texttt{forall ($x_1$:$T_1$) \ldots ($x_n$:$T_n$), relation $A$}. The
expression $A$, which depends on $x_1$ \ldots $x_n$, is called the
\emph{carrier} of the relation and \texttt{R} is
said to be a relation over \texttt{A}; the list $x_1,\ldots,x_n$
is the (possibly empty) list of parameters of the relation.

\firstexample
\begin{cscexample}[Parametric relation]
It is possible to implement finite sets of elements of type \texttt{A}
as unordered list of elements of type \texttt{A}. The function
\texttt{set\_eq: forall (A: Type), relation (list A)} satisfied by two lists
with the same elements is a parametric relation over \texttt{(list A)} with
one parameter \texttt{A}. The type of \texttt{set\_eq} is convertible with
\texttt{forall (A: Type), list A -> list A -> Prop}.
\end{cscexample}

An \emph{instance} of a parametric relation \texttt{R} with $n$ parameters
is any term \texttt{(R $t_1$ \ldots $t_n$)}.

Let \texttt{R} be a relation over \texttt{A} with $n$ parameters.
A term is a parametric proof of reflexivity for \texttt{R} if it has type
\texttt{forall ($x_1$:$T_1$) \ldots ($x_n$:$T_n$),
 reflexive (R $x_1$ \ldots $x_n$)}. Similar definitions are given for
parametric proofs of symmetry and transitivity.

\begin{cscexample}[Parametric relation (cont.)]
The \texttt{set\_eq} relation of the previous example can be proved to be
reflexive, symmetric and transitive.
\end{cscexample}

A parametric unary function $f$ of type
\texttt{forall ($x_1$:$T_1$) \ldots ($x_n$:$T_n$), $A_1$ -> $A_2$}
covariantly respects two parametric relation instances $R_1$ and $R_2$ if,
whenever $m, n$ satisfy $R_1~x~y$, their images $(f~x)$ and $(f~y)$ 
satisfy $R_2~(f~x)~(f~y)$ . An $f$ that respects its input and output relations
will be called a unary covariant \emph{morphism}. We can also say that $f$ is
a monotone function with respect to $R_1$ and $R_2$. The sequence $x_1,\ldots x_n$ represents the parameters of the morphism.

Let $R_1$ and $R_2$ be two parametric relations.
The \emph{signature} of a parametric morphism of type
\texttt{forall ($x_1$:$T_1$) \ldots ($x_n$:$T_n$), $A_1$ -> $A_2$} that
covariantly respects two parametric relations that are instances of
$R_1$ and $R_2$ is written $R_1 \texttt{++>} R_2$.
Notice that the special arrow \texttt{++>}, which reminds the reader
of covariance, is placed between the two parametric relations, not
between the two carriers or the two relation instances.

The previous definitions are extended straightforwardly to $n$-ary morphisms,
that are required to be simultaneously monotone on every argument.

Morphisms can also be contravariant in one or more of their arguments.
A morphism is contravariant on an argument associated to the relation instance
$R$ if it is covariant on the same argument when the inverse relation
$R^{-1}$ is considered. The special arrow \texttt{-{}->} is used in signatures
for contravariant morphisms.

Functions having arguments related by symmetric relations instances are both
covariant and contravariant in those arguments. The special arrow
\texttt{==>} is used in signatures for morphisms that are both covariant
and contravariant.

An instance of a parametric morphism $f$ with $n$ parameters is any term
\texttt{f $t_1$ \ldots $t_n$}.

\begin{cscexample}[Morphisms]
Continuing the previous example, let
\texttt{union: forall (A: Type), list A -> list A -> list A} perform the union
of two sets by appending one list to the other. \texttt{union} is a binary
morphism parametric over \texttt{A} that respects the relation instance
\texttt{(set\_eq A)}. The latter condition is proved by showing
\texttt{forall (A: Type) (S1 S1' S2 S2': list A), set\_eq A S1 S1' ->
 set\_eq A S2 S2' -> set\_eq A (union A S1 S2) (union A S1' S2')}.

The signature of the function \texttt{union} is
\texttt{set\_eq ==> set\_eq ==> set\_eq}.
\end{cscexample}

\begin{cscexample}[Contravariant morphism]
The division function \texttt{Rdiv: R -> R -> R} is a morphism of
signature \texttt{le ++> le -{}-> le} where \texttt{le} is
the usual order relation over real numbers. Notice that division is
covariant in its first argument and contravariant in its second
argument.
\end{cscexample}

Notice that Leibniz equality is a relation and that every function is a
morphism that respects Leibniz equality. Unfortunately, Leibniz equality
is not always the intended equality for a given structure.

In the next section we will describe the commands to register terms as
parametric relations and morphisms. Several tactics that deal with equality
in \Coq\ can also work with the registered relations.
The exact list of tactic will be given in Sect.~\ref{setoidtactics}.
For instance, the
tactic \texttt{reflexivity} can be used to close a goal $R~n~n$ whenever
$R$ is an instance of a registered reflexive relation. However, the tactics
that replace in a context $C[]$ one term with another one related by $R$
must verify that $C[]$ is a morphism that respects the intended relation.
Currently the verification consists in checking whether $C[]$ is a syntactic
composition of morphism instances that respects some obvious
compatibility constraints.

\begin{cscexample}[Rewriting]
Continuing the previous examples, suppose that the user must prove
\texttt{set\_eq int (union int (union int S1 S2) S2) (f S1 S2)} under the
hypothesis \texttt{H: set\_eq int S2 (nil int)}. It is possible to
use the \texttt{rewrite} tactic to replace the first two occurrences of
\texttt{S2} with \texttt{nil int} in the goal since the context
\texttt{set\_eq int (union int (union int S1 nil) nil) (f S1 S2)}, being
a composition of morphisms instances, is a morphism. However the tactic
will fail replacing the third occurrence of \texttt{S2} unless \texttt{f}
has also been declared as a morphism.
\end{cscexample}

\asection{Adding new relations and morphisms}
A parametric relation
\textit{Aeq}\texttt{: forall ($x_1$:$T_1$) \ldots ($x_n$:$T_n$),
 relation (A $x_1$ \ldots $x_n$)} over \textit{(A $x_1$ \ldots $x_n$)}
can be declared with the following command

\comindex{Add Relation}
\begin{verse}
  \texttt{Add Relation} \textit{A Aeq}\\
  ~\zeroone{\texttt{reflexivity proved by} \textit{refl}}\\
  ~\zeroone{\texttt{symmetry proved by} \textit{sym}}\\
  ~\zeroone{\texttt{transitivity proved by} \textit{trans}}\\
  \texttt{~as} \textit{id}.
\end{verse}
after having required the \texttt{Setoid} module with the
\texttt{Require Setoid} command.

The identifier \textit{id} gives a unique name to the morphism and it is
used by the command to generate fresh names for automatically provided lemmas
used internally.

Notice that \textit{A} is required to be a term having the same parameters
of \textit{Aeq}. This is a limitation of the tactic that is often unproblematic
in practice.

The proofs of reflexivity, symmetry and transitivity can be omitted if the
relation is not an equivalence relation.

If \textit{Aeq} is a transitive relation, then the command also generates
a lemma of type:
\begin{quote}
\texttt{forall ($x_1$:$T_1$)\ldots($x_n$:$T_n$)
 (x y x' y': (A $x_1$ \ldots $x_n$))\\
  Aeq $x_1$ \ldots $x_n$ x' x -> Aeq $x_1$ \ldots $x_n$ y y' ->\\
  (Aeq $x_1$ \ldots $x_n$ x y -> Aeq $x_1$ \ldots $x_n$ x' y')}
\end{quote}
that is used to declare \textit{Aeq} as a parametric morphism of signature
\texttt{Aeq -{}-> Aeq ++> impl} where \texttt{impl} is logical implication
seen as a parametric relation over \texttt{Aeq}.

Some tactics
(\texttt{reflexivity}, \texttt{symmetry}, \texttt{transitivity}) work only
on relations that respect the expected properties. The remaining tactics
(\texttt{replace}, \texttt{rewrite} and derived tactics such as
\texttt{autorewrite}) do not require any properties over the relation.
However, they are able to replace terms with related ones only in contexts
that are syntactic compositions of parametric morphism instances declared with
the following command.

\comindex{Add Morphism}
\begin{verse}
  \texttt{Add Morphism} \textit{f}\\
  \texttt{~with signature} \textit{sig}\\
  \texttt{~as id}.\\
  \texttt{Proof}\\
  ~\ldots\\
  \texttt{Qed}
\end{verse}

The command declares \textit{f} as a parametric morphism of signature
\textit{sig}. The identifier \textit{id} gives a unique name to the morphism
and it is used by the command to generate fresh names for automatically
provided lemmas used internally. The number of parameters for \textit{f}
is inferred by comparing its type with the provided signature.
The command asks the user to prove interactively that \textit{f} respects
the relations identified from the signature.

\begin{cscexample}
We start the example by assuming a small theory over homogeneous sets and
we declare set equality as a parametric equivalence relation and
union of two sets as a parametric morphism.
\begin{verbatim}
Require Export Relation_Definitions.
Require Export Setoid.
Set Implicit Arguments.
Set Contextual Implicit.
Parameter set: Type -> Type.
Parameter empty: forall A, set A.
Parameter eq_set: forall A, set A -> set A -> Prop.
Parameter union: forall A, set A -> set A -> set A.
Axiom eq_set_refl: forall A, reflexive _ (eq_set (A:=A)).
Axiom eq_set_sym: forall A, symmetric _ (eq_set (A:=A)).
Axiom eq_set_trans: forall A, transitive _ (eq_set (A:=A)).
Axiom empty_neutral: forall A (S: set A), eq_set (union S empty) S.
Axiom union_compat:
 forall (A : Type),
  forall x x' : set A, eq_set x x' ->
  forall y y' : set A, eq_set y y' ->
   eq_set (union x y) (union x' y').

Add Relation set eq_set
 reflexivity proved by (@eq_set_refl)
 symmetry proved by (@eq_set_sym)
 transitivity proved by (@eq_set_trans)
 as eq_set_rel.

Add Morphism union
 with signature eq_set ==> eq_set ==> eq_set
 as union_mor.
Proof.
 exact union_compat.
Qed.
\end{verbatim}

We proceed now by proving a simple lemma performing a rewrite step
and then applying reflexivity, as we would do working with Leibniz
equality. Both tactic applications are accepted
since the required properties over \texttt{eq\_set} and
\texttt{union} can be established from the two declarations above.

\begin{verbatim}
Goal forall (S: set nat),
 eq_set (union (union S empty) S) (union S S).
Proof.
 intros.
 rewrite (@empty_neutral).
 reflexivity.
Qed.
\end{verbatim}
\end{cscexample}

The tables of relations and morphisms are compatible with the \Coq\ 
sectioning mechanism. If you declare a relation or a morphism inside a section,
the declaration will be thrown away when closing the section.
And when you load a compiled file, all the declarations
of this file that were not inside a section will be loaded.

\asection{Rewriting and non reflexive relations}
To replace only one argument of an n-ary morphism it is necessary to prove
that all the other arguments are related to themselves by the respective
relation instances.

\begin{cscexample}
To replace \texttt{(union S empty)} with \texttt{S} in
\texttt{(union (union S empty) S) (union S S)} the rewrite tactic must
exploit the monotony of \texttt{union} (axiom \texttt{union\_compat} in
the previous example). Applying \texttt{union\_compat} by hand we are left
with the goal \texttt{eq\_set (union S S) (union S S)}.
\end{cscexample}

When the relations associated to some arguments are not reflexive, the tactic
cannot automatically prove the reflexivity goals, that are left to the user.

Setoids whose relation are partial equivalence relations (PER)
are useful to deal with partial functions. Let \texttt{R} be a PER. We say
that an element \texttt{x} is defined if \texttt{R x x}. A partial function
whose domain comprises all the defined elements only is declared as a
morphism that respects \texttt{R}. Every time a rewriting step is performed
the user must prove that the argument of the morphism is defined.

\begin{cscexample}
Let \texttt{eqO} be \texttt{fun x y => x = y $\land$ ~x$\neq$ 0} (the smaller PER over
non zero elements). Division can be declared as a morphism of signature
\texttt{eq ==> eq0 ==> eq}. Replace \texttt{x} with \texttt{y} in
\texttt{div x n = div y n} opens the additional goal \texttt{eq0 n n} that
is equivalent to \texttt{n=n $\land$ n$\neq$0}.
\end{cscexample}

\asection{Rewriting and non symmetric relations}
When the user works up to relations that are not symmetric, it is no longer
the case that any covariant morphism argument is also contravariant. As a
result it is no longer possible to replace a term with a related one in
every context, since the obtained goal implies the previous one if and
only if the replacement has been performed in a contravariant position.
In a similar way, replacement in an hypothesis can be performed only if
the replaced term occurs in a covariant position.

\begin{cscexample}[Covariance and contravariance]
Suppose that division over real numbers has been defined as a
morphism of signature \texttt{Zdiv: Zlt ++> Zlt -{}-> Zlt} (i.e.
\texttt{Zdiv} is increasing in its first argument, but decreasing on the
second one). Let \texttt{<} denotes \texttt{Zlt}. 
Under the hypothesis \texttt{H: x < y} we have
\texttt{k < x / y -> k < x / x}, but not
\texttt{k < y / x -> k < x / x}.
Dually, under the same hypothesis \texttt{k < x / y -> k < y / y} holds,
but \texttt{k < y / x -> k < y / y} does not.
Thus, if the current goal is \texttt{k < x / x}, it is possible to replace
only the second occurrence of \texttt{x} (in contravariant position)
with \texttt{y} since the obtained goal must imply the current one.
On the contrary, if \texttt{k < x / x} is
an hypothesis, it is possible to replace only the first occurrence of
\texttt{x} (in covariant position) with \texttt{y} since
the current hypothesis must imply the obtained one.
\end{cscexample}

An error message will be raised by the \texttt{rewrite} and \texttt{replace}
tactics when the user is trying to replace a term that occurs in the
wrong position.

As expected, composing morphisms together propagates the variance annotations by
switching the variance every time a contravariant position is traversed.
\begin{cscexample}
Let us continue the previous example and let us consider the goal
\texttt{x / (x / x) < k}. The first and third occurrences of \texttt{x} are
in a contravariant position, while the second one is in covariant position.
More in detail, the second occurrence of \texttt{x} occurs
covariantly in \texttt{(x / x)} (since division is covariant in its first
argument), and thus contravariantly in \texttt{x / (x / x)} (since division
is contravariant in its second argument), and finally covariantly in
\texttt{x / (x / x) < k} (since \texttt{<}, as every transitive relation,
is contravariant in its first argument with respect to the relation itself).
\end{cscexample}

\asection{Rewriting in ambiguous setoid contexts}
One function can respect several different relations and thus it can be
declared as a morphism having multiple signatures.

\begin{cscexample}
Union over homogeneous lists can be given all the following signatures:
\texttt{eq ==> eq ==> eq} (\texttt{eq} being the equality over ordered lists)
\texttt{set\_eq ==> set\_eq ==> set\_eq} (\texttt{set\_eq} being the equality
over unordered lists up to duplicates),
\texttt{multiset\_eq ==> multiset\_eq ==> multiset\_eq} (\texttt{multiset\_eq}
being the equality over unordered lists).
\end{cscexample}

To declare multiple signatures for a morphism, repeat the \texttt{Add Morphism}
command.

When morphisms have multiple signatures it can be the case that a rewrite
request is ambiguous, since it is unclear what relations should be used to
perform the rewriting. When non reflexive relations are involved, different
choices lead to different sets of new goals to prove. In this case the
tactic automatically picks one choice, but raises a warning describing the
set of alternative new goals. To force one particular choice, the user
can switch to the following alternative syntax for rewriting:

\comindex{setoid\_rewrite}
\begin{verse}
  \texttt{setoid\_rewrite} \zeroone{\textit{orientation}} \textit{term}
  \zeroone{\texttt{in} \textit{ident}}\\
  \texttt{~generate side conditions}
  \textit{term}$_1$ \ldots \textit{term}$_n$\\
\end{verse}
Up to the \texttt{generate side conditions} part, the syntax is 
equivalent to the
one of the \texttt{rewrite} tactic. Additionally, the user can specify a list
of new goals that the tactic must generate. The tactic will prune out from
the alternative choices those choices that do not open at least the user
proposed goals. Thus, providing enough side conditions, the user can restrict
the tactic to at most one choice.

\begin{cscexample}
Let \texttt{[=]+} and \texttt{[=]-} be the smaller partial equivalence
relations over positive (resp. negative) integers. Integer multiplication
can be declared as a morphism with the following signatures:
\texttt{Zmult: Zlt ++> [=]+ ==> Zlt} (multiplication with a positive number
is increasing) and
\texttt{Zmult: Zlt -{}-> [=]- ==> Zlt} (multiplication with a negative number
is decreasing).
Given the hypothesis \texttt{H: x < y} and the goal
\texttt{(x * n) * m < 0} the tactic \texttt{rewrite H} proposes
two alternative sets of goals that correspond to proving that \texttt{n}
and \texttt{m} are both positive or both negative.
\begin{itemize}
 \item \texttt{\ldots $\vdash$ (y * n) * m < 0}\\
       \texttt{\ldots $\vdash$ n [=]+ n}\\
       \texttt{\ldots $\vdash$ m [=]+ m}\\
 \item \texttt{\ldots $\vdash$ (y * n) * m < 0}\\
       \texttt{\ldots $\vdash$ n [=]- n} \\
       \texttt{\ldots $\vdash$ m [=]- m}
\end{itemize}
Remember that \texttt{n [=]+ n} is equivalent to \texttt{n=n $\land$ n > 0}.

To pick the second set of goals it is sufficient to use
\texttt{setoid\_rewrite H generate side conditions (m [=]- m)}
since the side condition \texttt{m [=]- m} is contained only in the second set
of goals.
\end{cscexample}

\asection{First class setoids and morphisms}
First class setoids and morphisms can also be handled by encoding them
as records. The projections of the setoid relation and of the morphism
function can be registered as parametric relations and morphisms, as
illustrated by the following example.
\begin{cscexample}[First class setoids]
\begin{verbatim}
Require Export Relation_Definitions.
Require Setoid.

Record Setoid: Type :=
{ car:Type;
  eq:car->car->Prop;
  refl: reflexive _ eq;
  sym: symmetric _ eq;
  trans: transitive _ eq
}.

Add Relation car eq
 reflexivity proved by refl
 symmetry proved by symm
 transitivity proved by trans
as eq_rel.

Record Morphism (S1 S2:Setoid): Type :=
{ f:car S1 ->car S2;
  compat: forall (x1 x2: car S1), eq S1 x1 x2 -> eq S2 (f x1) (f x2)
}.

Add Morphism f with signature eq ==> eq as apply_mor.
Proof.
 intros S1 S2 m.
 apply (compat S1 S2 m).
Qed.

Lemma test: forall (S1 S2:Setoid) (m: Morphism S1 S2)
 (x y: car S1), eq S1 x y -> eq S2 (f _ _ m x) (f _ _ m y).
Proof.
 intros.
 rewrite H.
 reflexivity.
Qed.
\end{verbatim}
\end{cscexample}

\asection{Tactics enabled on user provided relations}
\label{setoidtactics}
The following tactics, all prefixed by \texttt{setoid\_}, 
deal with arbitrary
registered relations and morphisms. Moreover, all the corresponding unprefixed
tactics (i.e. \texttt{reflexivity, symmetry, transitivity, replace, rewrite})
have been extended to fall back to their prefixed counterparts when
the relation involved is not Leibniz equality. Notice, however, that using
the prefixed tactics it is possible to pass additional arguments such as
\texttt{generate side conditions} or \texttt{using relation}.

\comindex{setoid\_reflexivity}
\begin{verse}
  \texttt{setoid\_reflexivity}
\end{verse}

\comindex{setoid\_symmetry}
\begin{verse}
  \texttt{setoid\_symmetry}
  \zeroone{\texttt{in} \textit{ident}}\\
\end{verse}

\comindex{setoid\_transitivity}
\begin{verse}
  \texttt{setoid\_transitivity}
\end{verse}

\comindex{setoid\_rewrite}
\begin{verse}
  \texttt{setoid\_rewrite} \zeroone{\textit{orientation}} \textit{term}\\
  ~\zeroone{\texttt{in} \textit{ident}}\\
  ~\zeroone{\texttt{generate side conditions}
  \textit{term}$_1$ \ldots \textit{term}$_n$}\\
\end{verse}

The \texttt{generate side conditions} argument cannot be passed to the
unprefixed form.

\comindex{setoid\_replace}
\begin{verse}
  \texttt{setoid\_replace} \textit{term} \texttt{with} \textit{term}
  ~\zeroone{\texttt{in} \textit{ident}}\\
  ~\zeroone{\texttt{using relation} \textit{term}}\\
  ~\zeroone{\texttt{generate side conditions}
  \textit{term}$_1$ \ldots \textit{term}$_n$}\\
\end{verse}

The \texttt{generate side conditions} and \texttt{using relation} arguments cannot be
passed to the unprefixed form. The latter argument tells the tactic what
parametric relation should be used to replace the first tactic argument
with the second one. If omitted, it defaults to Leibniz equality.

Every derived tactic that is based on the unprefixed forms of the tactics
considered above will also work up to user defined relations. For instance,
it is possible to register hints for \texttt{autorewrite} that are
not proof of Leibniz equalities. In particular it is possible to exploit
\texttt{autorewrite} to simulate normalization in a term rewriting system
up to user defined equalities.

\asection{Printing relations and morphisms}
The \texttt{Print Setoids} command shows the list of currently registered
parametric relations and morphisms. For each morphism its signature is also
given. When the rewriting tactics refuse to replace a term in a context
because the latter is not a composition of morphisms, the \texttt{Print Setoids}
command is useful to understand what additional morphisms should be registered.

\asection{Deprecated syntax and backward incompatibilities}
Due to backward compatibility reasons, the following syntax for the
declaration of setoids and morphisms is also accepted.

\comindex{Add Setoid}
\begin{verse}
  \texttt{Add Setoid} \textit{A Aeq ST} \texttt{as} \textit{ident}
\end{verse}
where \textit{Aeq} is a congruence relation without parameters,
\textit{A} is its carrier and \textit{ST} is an object of type
\verb|(Setoid_Theory A Aeq)| (i.e. a record packing together the reflexivity,
symmetry and transitivity lemmas). Notice that the syntax is not completely
backward compatible since the identifier was not required.

\comindex{Add Morphism}
\begin{verse}
  \texttt{Add Morphism} \textit{ f }:\textit{ ident}.\\
  Proof.\\
  \ldots\\
  Qed.
\end{verse}

The latter command is restricted to the declaration of morphisms without
parameters. It is not fully backward compatible since the property the user
is asked to prove is slightly different: for $n$-ary morphisms the hypotheses
of the property are permuted; moreover, when the morphism returns a
proposition, the property is now stated using a bi-implication in place of
a simple implication. In practice, porting an old development to the new
semantics is usually quite simple.

Notice that several limitations of the old implementation have been lifted.
In particular, it is now possible to declare several relations with the
same carrier and several signatures for the same morphism. Moreover, it is
now also possible to declare several morphisms having the same signature.
Finally, the replace and rewrite tactics can be used to replace terms in
contexts that were refused by the old implementation.

%%% Local Variables: 
%%% mode: latex
%%% TeX-master: "Reference-Manual"
%%% End: