We have seen how to formalize (1) and (2). To express we need to formalize condition (3). First we define the induced relation
(this is how it appears in the libraries). This language is a function of a given language
, but that parameter is not always displayed although it is implicit.
We establish straight forwardly that is an equivalence relation.
The formulation of (3) as in Hopcroft and Ullman would be that is
of finite index. But we will see that to prove
constructively we need to be explicit that
is a decidable
language. So we take (3) to be:
is decidable and
is
of finite index.
The proof of (3) from (2) appears to be the simplest of the
implications. (From (2) we know immediately that is decidable.)
We show that if
refines
, then the index of
is no larger
than that of
, that is,
So the only nonroutine step is to show
This follows directly from the fact that is extension invariant
since
, but then
This seems to be the whole story until we look at the details of the lemma
It requires that we prove that
the relation is decidable (see
). This
complication suggests another more elegant proof which we outline
after stating the theorem. This second proof is the one we formalize.
Corresp
Corresp
We can use the same devices as before to render this theorem more readable. Here is Stuart Allen's version.
Proof
The key idea in this proof is to show that where
is like
but is
defined on the quotient type using the boolean valued function
.
This function
characterizes
in a simple way and is easier to
work with than
itself. This leads us to work with an equivalence
relation
instead of
. The proof is essentially establishing
two isomorphisms,
Qed
Here is the main line of the Web proof as it appears after
applying Allen's technique to the full Web proof. Here will be displayed as
to reveal the dependence on
. Notice that there
are two side proofs as well as a number of lemma references. These
can be expanded in the on-line version just by clicking on the names.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
The real work for us in proving this theorem was actually spent on
building general facts about quotients and in defining and
showing that it is an equivalence relation on
.
This required a long sequence of lemmas. All of this is left implicit in Hopcroft and Ullman who need at least the properties of
@ on quotient sets and facts about equivalence relations on quotient
sets.
EquivRel(