LogicLogic in Coq
In previous chapters, we have seen many examples of factual
claims (propositions) and ways of presenting evidence of their
truth (proofs). In particular, we have worked extensively with
equality propositions of the form e1 = e2, with
implications (P → Q), and with quantified propositions (∀
x, P). In this chapter, we will see how Coq can be used to carry
out other familiar forms of logical reasoning.
Before diving into details, let's talk a bit about the status of
mathematical statements in Coq. Recall that Coq is a typed
language, which means that every sensible expression in its world
has an associated type. Logical claims are no exception: any
statement we might try to prove in Coq has a type, namely Prop,
the type of propositions. We can see this with the Check
command:
Note that all well-formed propositions have type Prop in Coq,
regardless of whether they are true or not. Simply being a
proposition is one thing; being provable is something else!
Indeed, propositions don't just have types: they are first-class
objects that can be manipulated in the same ways as the other
entities in Coq's world. So far, we've seen one primary place
that propositions can appear: in Theorem (and Lemma and
Example) declarations.
But propositions can be used in many other ways. For example, we
can give a name to a proposition using a Definition, just as we
have given names to expressions of other sorts.
We can later use this name in any situation where a proposition is
expected — for example, as the claim in a Theorem declaration.
We can also write parameterized propositions — that is,
functions that take arguments of some type and return a
proposition. For instance, the following function takes a number
and returns a proposition asserting that this number is equal to
three:
In Coq, functions that return propositions are said to define
properties of their arguments. For instance, here's a
polymorphic property defining the familiar notion of an injective
function.
Definition injective {A B} (f : A → B) :=
∀x y : A, f x = f y → x = y.
Lemma succ_inj : injective S.
Proof.
intros n m H. inversion H. reflexivity.
Qed.
The equality operator = that we have been using so far is also
just a function that returns a Prop. The expression n = m is
just syntactic sugar for eq n m, defined using Coq's Notation
mechanism. Because = can be used with elements of any type, it
is also polymorphic:
(Notice that we wrote @eq instead of eq: The type argument A
to eq is declared as implicit, so we need to turn off implicit
arguments to see the full type of eq.)
Logical Connectives
Conjunction
To prove a conjunction, use the split tactic. Its effect is to
generate two subgoals, one for each part of the statement:
Proof.
split.
- (* 3 + 4 = 7 *) reflexivity.
- (* 2 + 2 = 4 *) reflexivity.
Qed.
More generally, the following principle works for any two
propositions A and B:
Lemma and_intro : ∀A B : Prop, A → B → A ∧ B.
Proof.
intros A B HA HB. split.
- apply HA.
- apply HB.
Qed.
A logical statement with multiple arrows is just a theorem with
has several hypotheses. Here, and_intro says that, for any
propositions A and B, if we assume that A is true and we
assume that B is true, then A ∧ B is also true.
Since applying a theorem with hypotheses to some goal has the
effect of generating as many subgoals as there are hypotheses for
that theorem, we can, apply and_intro to achieve the same effect
as split.
Example and_example' : 3 + 4 = 7 ∧ 2 × 2 = 4.
Proof.
apply and_intro.
- (* 3 + 4 = 7 *) reflexivity.
- (* 2 + 2 = 4 *) reflexivity.
Qed.
☐
So much for proving conjunctive statements. To go in the other
direction — i.e., to use a conjunctive hypothesis to prove
something else — we employ the destruct tactic.
If the proof context contains a hypothesis H of the form A ∧
B, writing destruct H as [HA HB] will remove H from the
context and add two new hypotheses: HA, stating that A is
true, and HB, stating that B is true. For instance:
Lemma and_example2 :
∀n m : nat, n = 0 ∧ m = 0 → n + m = 0.
Proof.
intros n m H.
destruct H as [Hn Hm].
rewrite Hn. rewrite Hm.
reflexivity.
Qed.
As usual, we can also destruct H when we introduce it instead of
introducing and then destructing it:
Lemma and_example2' :
∀n m : nat, n = 0 ∧ m = 0 → n + m = 0.
Proof.
intros n m [Hn Hm].
rewrite Hn. rewrite Hm.
reflexivity.
Qed.
You may wonder why we bothered packing the two hypotheses n = 0
and m = 0 into a single conjunction, since we could have also
stated the theorem with two separate premises:
Lemma and_example2'' :
∀n m : nat, n = 0 → m = 0 → n + m = 0.
Proof.
intros n m Hn Hm.
rewrite Hn. rewrite Hm.
reflexivity.
Qed.
In this case, there is not much difference between the two
theorems. But it is often necessary to explicitly decompose
conjunctions that arise from intermediate steps in proofs,
especially in bigger developments. Here's a simplified
example:
Lemma and_example3 :
∀n m : nat, n + m = 0 → n × m = 0.
Proof.
intros n m H.
assert (H' : n = 0 ∧ m = 0).
{ apply and_exercise. apply H. }
destruct H' as [Hn Hm].
rewrite Hn. reflexivity.
Qed.
Another common situation with conjunctions is that we know A ∧
B but in some context we need just A (or just B). The
following lemmas are useful in such cases:
☐
Finally, we sometimes need to rearrange the order of conjunctions
and/or the grouping of conjuncts in multi-way conjunctions. The
following commutativity and associativity theorems come in handy
in such cases.
Theorem and_commut : ∀P Q : Prop,
P ∧ Q → Q ∧ P.
Proof.
(* WORKED IN CLASS *)
intros P Q [HP HQ].
split.
- (* left *) apply HQ.
- (* right *) apply HP. Qed.
Exercise: 2 stars (and_assoc)
(In the following proof of associativity, notice how the nested intro pattern breaks the hypothesis H : P ∧ (Q ∧ R) down into HP : P, HQ : Q, and HR : R. Finish the proof from there.)Theorem and_assoc : ∀P Q R : Prop,
P ∧ (Q ∧ R) → (P ∧ Q) ∧ R.
Proof.
intros P Q R [HP [HQ HR]].
(* FILL IN HERE *) Admitted.
☐
By the way, the infix notation ∧ is actually just syntactic
sugar for and A B. That is, and is a Coq operator that takes
two propositions as arguments and yields a proposition.
Disjunction
Lemma or_example :
∀n m : nat, n = 0 ∨ m = 0 → n × m = 0.
Proof.
(* This pattern implicitly does case analysis on
n = 0 ∨ m = 0 *)
intros n m [Hn | Hm].
- (* Here, n = 0 *)
rewrite Hn. reflexivity.
- (* Here, m = 0 *)
rewrite Hm. rewrite ← mult_n_O.
reflexivity.
Qed.
We can see in this example that, when we perform case analysis on
a disjunction A ∨ B, we must satisfy two proof obligations,
each showing that the conclusion holds under a different
assumption — A in the first subgoal and B in the second.
Note that the case analysis pattern (Hn | Hm) allows us to name
the hypothesis that is generated in each subgoal.
Conversely, to show that a disjunction holds, we need to show that
one of its sides does. This is done via two tactics, left and
right. As their names imply, the first one requires proving the
left side of the disjunction, while the second requires proving
its right side. Here is a trivial use...
... and a slightly more interesting example requiring the use of
both left and right:
Lemma zero_or_succ :
∀n : nat, n = 0 ∨ n = S (pred n).
Proof.
intros [|n].
- left. reflexivity.
- right. reflexivity.
Qed.
☐
So far, we have mostly been concerned with proving that certain
things are true — addition is commutative, appending lists is
associative, etc. Of course, we may also be interested in
negative results, showing that certain propositions are not
true. In Coq, such negative statements are expressed with the
negation operator ¬.
To see how negation works, recall the discussion of the principle
of explosion from the Tactics chapter; it asserts that, if we
assume a contradiction, then any other proposition can be derived.
Following this intuition, we could define ¬ P ("not P") as
∀ Q, P → Q. Coq actually makes a slightly different
choice, defining ¬ P as P → False, where False is a
particular contradictory proposition defined in the standard
library.
Falsehood and Negation
Module MyNot.
Definition not (P:Prop) := P → False.
Notation "¬ x" := (not x) : type_scope.
Check not.
(* ===> Prop -> Prop *)
End MyNot.
Since False is a contradictory proposition, the principle of
explosion also applies to it. If we get False into the proof
context, we can destruct it to complete any goal:
Theorem ex_falso_quodlibet : ∀(P:Prop),
False → P.
Proof.
(* WORKED IN CLASS *)
intros P contra.
destruct contra. Qed.
The Latin ex falso quodlibet means, literally, "from falsehood
follows whatever you like"; this is another common name for the
principle of explosion.
Exercise: 2 stars, optional (not_implies_our_not)
Show that Coq's definition of negation implies the intuitive one mentioned above:
☐
This is how we use not to state that 0 and 1 are different
elements of nat:
Such inequality statements are frequent enough to warrant a
special notation, x ≠ y:
It takes a little practice to get used to working with negation in
Coq. Even though you can see perfectly well why a statement
involving negation is true, it can be a little tricky at first to
get things into the right configuration so that Coq can understand
it! Here are proofs of a few familiar facts to get you warmed
up.
Theorem not_False :
¬ False.
Proof.
unfold not. intros H. destruct H. Qed.
Theorem contradiction_implies_anything : ∀P Q : Prop,
(P ∧ ¬P) → Q.
Proof.
(* WORKED IN CLASS *)
intros P Q [HP HNA]. unfold not in HNA.
apply HNA in HP. destruct HP. Qed.
Theorem double_neg : ∀P : Prop,
P → ~~P.
Proof.
(* WORKED IN CLASS *)
intros P H. unfold not. intros G. apply G. apply H. Qed.
Exercise: 2 stars, advanced, recommended (double_neg_inf)
Write an informal proof of double_neg:☐
Exercise: 2 stars, recommended (contrapositive)
☐
Exercise: 1 star, advanced (informal_not_PNP)
Write an informal proof (in English) of the proposition ∀ P : Prop, ~(P ∧ ¬P).(* FILL IN HERE *)
☐
Similarly, since inequality involves a negation, it requires a
little practice to be able to work with it fluently. Here is one
useful trick. If you are trying to prove a goal that is
nonsensical (e.g., the goal state is false = true), apply
ex_falso_quodlibet to change the goal to False. This makes it
easier to use assumptions of the form ¬P that may be available
in the context — in particular, assumptions of the form
x≠y.
Theorem not_true_is_false : ∀b : bool,
b ≠ true → b = false.
Proof.
intros [] H.
- (* b = true *)
unfold not in H.
apply ex_falso_quodlibet.
apply H. reflexivity.
- (* b = false *)
reflexivity.
Qed.
Since reasoning with ex_falso_quodlibet is quite common, Coq
provides a built-in tactic, exfalso, for applying it.
Theorem not_true_is_false' : ∀b : bool,
b ≠ true → b = false.
Proof.
intros [] H.
- (* b = false *)
unfold not in H.
exfalso. (* <=== *)
apply H. reflexivity.
- (* b = true *) reflexivity.
Qed.
Truth
Unlike False, which is used extensively, True is used quite
rarely, since it is trivial (and therefore uninteresting) to prove
as a goal, and it carries no useful information as a hypothesis.
But it can be quite useful when defining complex Props using
conditionals or as a parameter to higher-order Props. We will
see some examples such uses of True later on.
The handy "if and only if" connective, which asserts that two
propositions have the same truth value, is just the conjunction of
two implications.
Logical Equivalence
Module MyIff.
Definition iff (P Q : Prop) := (P → Q) ∧ (Q → P).
Notation "P ↔ Q" := (iff P Q)
(at level 95, no associativity)
: type_scope.
End MyIff.
Theorem iff_sym : ∀P Q : Prop,
(P ↔ Q) → (Q ↔ P).
Proof.
(* WORKED IN CLASS *)
intros P Q [HAB HBA].
split.
- (* -> *) apply HBA.
- (* <- *) apply HAB. Qed.
Lemma not_true_iff_false : ∀b,
b ≠ true ↔ b = false.
Proof.
(* WORKED IN CLASS *)
intros b. split.
- (* -> *) apply not_true_is_false.
- (* <- *)
intros H. rewrite H. intros H'. inversion H'.
Qed.
Exercise: 1 star, optional (iff_properties)
Using the above proof that ↔ is symmetric (iff_sym) as a guide, prove that it is also reflexive and transitive.Theorem iff_refl : ∀P : Prop,
P ↔ P.
Proof.
(* FILL IN HERE *) Admitted.
Theorem iff_trans : ∀P Q R : Prop,
(P ↔ Q) → (Q ↔ R) → (P ↔ R).
Proof.
(* FILL IN HERE *) Admitted.
Theorem or_distributes_over_and : ∀P Q R : Prop,
P ∨ (Q ∧ R) ↔ (P ∨ Q) ∧ (P ∨ R).
Proof.
(* FILL IN HERE *) Admitted.
P ∨ (Q ∧ R) ↔ (P ∨ Q) ∧ (P ∨ R).
Proof.
(* FILL IN HERE *) Admitted.
☐
Some of Coq's tactics treat iff statements specially, avoiding
the need for some low-level proof-state manipulation. In
particular, rewrite and reflexivity can be used with iff
statements, not just equalities. To enable this behavior, we need
to import a special Coq library that allows rewriting with other
formulas besides equality:
Here is a simple example demonstrating how these tactics work with
iff. First, let's prove a couple of basic iff equivalences:
Lemma mult_0 : ∀n m, n × m = 0 ↔ n = 0 ∨ m = 0.
Proof.
split.
- apply mult_eq_0.
- apply or_example.
Qed.
Lemma or_assoc :
∀P Q R : Prop, P ∨ (Q ∨ R) ↔ (P ∨ Q) ∨ R.
Proof.
intros P Q R. split.
- intros [H | [H | H]].
+ left. left. apply H.
+ left. right. apply H.
+ right. apply H.
- intros [[H | H] | H].
+ left. apply H.
+ right. left. apply H.
+ right. right. apply H.
Qed.
We can now use these facts with rewrite and reflexivity to
give smooth proofs of statements involving equivalences. Here is
a ternary version of the previous mult_0 result:
Lemma mult_0_3 :
∀n m p, n × m × p = 0 ↔ n = 0 ∨ m = 0 ∨ p = 0.
Proof.
intros n m p.
rewrite mult_0. rewrite mult_0. rewrite or_assoc.
reflexivity.
Qed.
The apply tactic can also be used with ↔. When given an
equivalence as its argument, apply tries to guess which side of
the equivalence to use.
Lemma apply_iff_example :
∀n m : nat, n × m = 0 → n = 0 ∨ m = 0.
Proof.
intros n m H. apply mult_0. apply H.
Qed.
Existential Quantification
Conversely, if we have an existential hypothesis ∃ x, P in
the context, we can destruct it to obtain a witness x and a
hypothesis stating that P holds of x.
Theorem exists_example_2 : ∀n,
(∃m, n = 4 + m) →
(∃o, n = 2 + o).
Proof.
intros n [m Hm].
∃(2 + m).
apply Hm. Qed.
Exercise: 1 star (dist_not_exists)
Prove that "P holds for all x" implies "there is no x for which P does not hold."Theorem dist_not_exists : ∀(X:Type) (P : X → Prop),
(∀x, P x) → ¬ (∃x, ¬ P x).
Proof.
(* FILL IN HERE *) Admitted.
☐
Exercise: 2 stars (dist_exists_or)
Prove that existential quantification distributes over disjunction.Theorem dist_exists_or : ∀(X:Type) (P Q : X → Prop),
(∃x, P x ∨ Q x) ↔ (∃x, P x) ∨ (∃x, Q x).
Proof.
(* FILL IN HERE *) Admitted.
☐
Programming with Propositions
- If l is the empty list, then x cannot occur on it, so the
property "x appears in l" is simply false.
- Otherwise, l has the form x' :: l'. In this case, x occurs in l if either it is equal to x' or it occurs in l'.
Fixpoint In {A : Type} (x : A) (l : list A) : Prop :=
match l with
| [] ⇒ False
| x' :: l' ⇒ x' = x ∨ In x l'
end.
When In is applied to a concrete list, it expands into a
concrete sequence of nested conjunctions.
Example In_example_1 : In 4 [3; 4; 5].
Proof.
simpl. right. left. reflexivity.
Qed.
Example In_example_2 :
∀n, In n [2; 4] →
∃n', n = 2 × n'.
Proof.
simpl.
intros n [H | [H | []]].
- ∃1. rewrite ← H. reflexivity.
- ∃2. rewrite ← H. reflexivity.
Qed.
(Notice the use of the empty pattern to discharge the last case
en passant.)
We can also prove more generic, higher-level lemmas about In.
Note, in the next, how In starts out applied to a variable and
only gets expanded when we do case analysis on this variable:
Lemma In_map :
∀(A B : Type) (f : A → B) (l : list A) (x : A),
In x l →
In (f x) (map f l).
Proof.
intros A B f l x.
induction l as [|x' l' IHl'].
- (* l = nil, contradiction *)
simpl. intros [].
- (* l = x' :: l' *)
simpl. intros [H | H].
+ rewrite H. left. reflexivity.
+ right. apply IHl'. apply H.
Qed.
This way of defining propositions, though convenient in some
cases, also has some drawbacks. In particular, it is subject to
Coq's usual restrictions regarding the definition of recursive
functions, e.g., the requirement that they be "obviously
terminating." In the next chapter, we will see how to define
propositions inductively, a different technique with its own set
of strengths and limitations.
Exercise: 2 stars (In_map_iff)
Lemma In_map_iff :
∀(A B : Type) (f : A → B) (l : list A) (y : B),
In y (map f l) ↔
∃x, f x = y ∧ In x l.
Proof.
(* FILL IN HERE *) Admitted.
∀(A B : Type) (f : A → B) (l : list A) (y : B),
In y (map f l) ↔
∃x, f x = y ∧ In x l.
Proof.
(* FILL IN HERE *) Admitted.
Lemma in_app_iff : ∀A l l' (a:A),
In a (l++l') ↔ In a l ∨ In a l'.
Proof.
(* FILL IN HERE *) Admitted.
In a (l++l') ↔ In a l ∨ In a l'.
Proof.
(* FILL IN HERE *) Admitted.
☐
Drawing inspiration from In, write a recursive function All
stating that some property P holds of all elements of a list
l. To make sure your definition is correct, prove the All_In
lemma below. (Of course, your definition should not just
restate the left-hand side of All_In.)
Exercise: 3 stars (All)
Recall that functions returning propositions can be seen as properties of their arguments. For instance, if P has type nat → Prop, then P n states that property P holds of n.Fixpoint All {T} (P : T → Prop) (l : list T) : Prop :=
(* FILL IN HERE *) admit.
Lemma All_In :
∀T (P : T → Prop) (l : list T),
(∀x, In x l → P x) ↔
All P l.
Proof.
(* FILL IN HERE *) Admitted.
☐
Exercise: 3 stars (combine_odd_even)
Complete the definition of the combine_odd_even function below. It takes as arguments two properties of numbers, Podd and Peven, and it should return a property P such that P n is equivalent to Podd n when n is odd and equivalent to Peven n otherwise.
To test your definition, prove the following facts:
Theorem combine_odd_even_intro :
∀(Podd Peven : nat → Prop) (n : nat),
(oddb n = true → Podd n) →
(oddb n = false → Peven n) →
combine_odd_even Podd Peven n.
Proof.
(* FILL IN HERE *) Admitted.
Theorem combine_odd_even_elim_odd :
∀(Podd Peven : nat → Prop) (n : nat),
combine_odd_even Podd Peven n →
oddb n = true →
Podd n.
Proof.
(* FILL IN HERE *) Admitted.
Theorem combine_odd_even_elim_even :
∀(Podd Peven : nat → Prop) (n : nat),
combine_odd_even Podd Peven n →
oddb n = false →
Peven n.
Proof.
(* FILL IN HERE *) Admitted.
☐
Applying Theorems to Arguments
Coq prints the statement of the plus_comm theorem in the same
way that it prints the type of any term that we ask it to
Check. Why?
The reason is that the identifier plus_comm actually refers to a
proof object — a data structure that represents a logical
derivation establishing of the truth of the statement ∀ n m
: nat, n + m = m + n. The type of this object is the statement
of the theorem that it is a proof of.
Intuitively, this makes sense because the statement of a theorem
tells us what we can use that theorem for, just as the type of a
computational object tells us what we can do with that object —
e.g., if we have a term of type nat → nat → nat, we can give
it two nats as arguments and get a nat back. Similarly, if we
have an object of type n = m → n + n = m + m and we provide it
an "argument" of type n = m, we can derive n + n = m + m.
Operationally, this analogy goes even further: by applying a
theorem, as if it were a function, to hypotheses with matching
types, we can specialize its result without having to resort to
intermediate assertions. For example, suppose we wanted to prove
the following result:
It appears at first sight that we ought to be able to prove this
by rewriting with plus_comm twice to make the two sides match.
The problem, however, is that the second rewrite will undo the
effect of the first.
One simple way of fixing this problem, using only tools that we
already know, is to use assert to derive a specialized version
of plus_comm that can be used to rewrite exactly where we
want.
rewrite plus_comm.
assert (H : m + p = p + m).
{ rewrite plus_comm. reflexivity. }
rewrite H.
reflexivity.
Qed.
A more elegant alternative is to apply plus_comm directly to the
arguments we want to instantiate it with, in much the same way as
we apply a polymorphic function to a type argument.
Lemma plus_comm3_take2 :
∀n m p, n + (m + p) = (p + m) + n.
Proof.
intros n m p.
rewrite plus_comm.
rewrite (plus_comm m).
reflexivity.
Qed.
You can "use theorems as functions" in this way with almost all
tactics that take a theorem name as an argument. Note also that
theorem application uses the same inference mechanisms as function
application; thus, it is possible, for example, to supply
wildcards as arguments to be inferred, or to declare some
hypotheses to a theorem as implicit by default. These features
are illustrated in the proof below.
Example lemma_application_ex :
∀{n : nat} {ns : list nat},
In n (map (fun m ⇒ m × 0) ns) →
n = 0.
Proof.
intros n ns H.
destruct (proj1 _ _ (In_map_iff _ _ _ _ _) H)
as [m [Hm _]].
rewrite mult_0_r in Hm. rewrite ← Hm. reflexivity.
Qed.
We will see many more examples of the idioms from this section in
later chapters.
Coq vs. Set Theory
Functional Extensionality
In common mathematical practice, two functions f and g are
considered equal if they produce the same outputs:
Informally speaking, an "extensional property" is one that
pertains to an object's observable behavior. Thus, functional
extensionality simply means that a function's identity is
completely determined by what we can observe from it — i.e., in
Coq terms, the results we obtain after applying it.
Functional extensionality is not part of Coq's basic axioms: the
only way to show that two functions are equal is by
simplification (as we did in the proof of function_equality_ex).
But we can add it to Coq's core logic using the Axiom
command.
(∀x, f x = g x) → f = g
This is known as the principle of functional extensionality.
Using Axiom has the same effect as stating a theorem and
skipping its proof using Admitted, but it alerts the reader that
this isn't just something we're going to come back and fill in
later!
We can now invoke functional extensionality in proofs:
Lemma plus_comm_ext : plus = fun n m ⇒ m + n.
Proof.
apply functional_extensionality. intros n.
apply functional_extensionality. intros m.
apply plus_comm.
Qed.
Naturally, we must be careful when adding new axioms into Coq's
logic, as they may render it inconsistent — that is, it may
become possible to prove every proposition, including False!
Unfortunately, there is no simple way of telling whether an axiom
is safe: hard work is generally required to establish the
consistency of any particular combination of axioms. Fortunately,
it is known that adding functional extensionality, in particular,
is consistent.
Note that it is possible to check whether a particular proof
relies on any additional axioms, using the Print Assumptions
command. For instance, if we run it on plus_comm_ext, we see
that it uses functional_extensionality:
Print Assumptions plus_comm_ext.
(* ===>
Axioms:
functional_extensionality :
forall (X Y : Type) (f g : X -> Y),
(forall x : X, f x = g x) -> f = g *)
Exercise: 5 stars (tr_rev)
One problem with the definition of the list-reversing function rev that we have is that it performs a call to app on each step; running app takes time asymptotically linear in the size of the list, which means that rev has quadratic running time. We can improve this with the following definition:Fixpoint rev_append {X} (l1 l2 : list X) : list X :=
match l1 with
| [] ⇒ l2
| x :: l1' ⇒ rev_append l1' (x :: l2)
end.
Definition tr_rev {X} (l : list X) : list X :=
rev_append l [].
This version is said to be tail-recursive, because the recursive
call to the function is the last operation that needs to be
performed (i.e., we don't have to execute ++ after the recursive
call); a decent compiler will generate very efficient code in this
case. Prove that both definitions are indeed equivalent.
☐
We've seen that Coq has two different ways of encoding logical
facts: with booleans (of type bool), and with
propositions (of type Prop). For instance, to claim that a
number n is even, we can say either (1) that evenb n returns
true or (2) that there exists some k such that n = double k.
Indeed, these two notions of evenness are equivalent, as can
easily be shown with a couple of auxiliary lemmas (one of which is
left as an exercise).
We often say that the boolean evenb n reflects the proposition
∃ k, n = double k.
Propositions and Booleans
Theorem evenb_double : ∀k, evenb (double k) = true.
Proof.
intros k. induction k as [|k' IHk'].
- reflexivity.
- simpl. apply IHk'.
Qed.
Theorem evenb_double_conv : ∀n,
∃k, n = if evenb n then double k
else S (double k).
Proof.
(* Hint: Use the evenb_S lemma from Induction.v. *)
(* FILL IN HERE *) Admitted.
∃k, n = if evenb n then double k
else S (double k).
Proof.
(* Hint: Use the evenb_S lemma from Induction.v. *)
(* FILL IN HERE *) Admitted.
☐
Theorem even_bool_prop : ∀n,
evenb n = true ↔ ∃k, n = double k.
Proof.
intros n. split.
- intros H. destruct (evenb_double_conv n) as [k Hk].
rewrite Hk. rewrite H. ∃k. reflexivity.
- intros [k Hk]. rewrite Hk. apply evenb_double.
Qed.
Similarly, to state that two numbers n and m are equal, we can
say either (1) that beq_nat n m returns true or (2) that n =
m. These two notions are equivalent.
Theorem beq_nat_true_iff : ∀n1 n2 : nat,
beq_nat n1 n2 = true ↔ n1 = n2.
Proof.
intros n1 n2. split.
- apply beq_nat_true.
- intros H. rewrite H. rewrite ← beq_nat_refl. reflexivity.
Qed.
However, while the boolean and propositional formulations of a
claim are equivalent from a purely logical perspective, we have
also seen that they need not be equivalent operationally.
Equality provides an extreme example: knowing that beq_nat n m =
true is generally of little help in the middle of a proof
involving n and m; however, if we convert the statement to the
equivalent form n = m, we can rewrite with it.
The case of even numbers is also interesting. Recall that, when
proving the backwards direction of
even_bool_prop (evenb_double, going from the propositional to
the boolean claim), we used a simple induction on k). On the
other hand, the converse (the evenb_double_conv exercise)
required a clever generalization, since we can't directly prove
(∃ k, n = double k) → evenb n = true.
For these examples, the propositional claims were more useful than
their boolean counterparts, but this is not always the case. For
instance, we cannot test whether a general proposition is true or
not in a function definition; as a consequence, the following code
fragment is rejected:
Coq complains that n = 2 has type Prop, while it expects an
elements of bool (or some other inductive type with two
elements). The reason for this error message has to do with the
computational nature of Coq's core language, which is designed
so that every function that it can express is computable and
total. One reason for this is to allow the extraction of
executable programs from Coq developments. As a consequence,
Prop in Coq does not have a universal case analysis operation
telling whether any given proposition is true or false, since such
an operation would allow us to write non-computable functions.
Although general non-computable properties cannot be phrased as
boolean computations, it is worth noting that even many
computable properties are easier to express using Prop than
bool, since recursive function definitions are subject to
significant restrictions in Coq. For instance, the next chapter
shows how to define the property that a regular expression matches
a given string using Prop. Doing the same with bool would
amount to writing a regular expression matcher, which would be
more complicated, harder to understand, and harder to reason
about.
Conversely, an important side benefit of stating facts using
booleans is enabling some proof automation through computation
with Coq terms, a technique known as proof by
reflection. Consider the following statement:
The most direct proof of this fact is to give the value of k
explicitly.
Proof. ∃500. reflexivity. Qed.
On the other hand, the proof of the corresponding boolean
statement is even simpler:
What is interesting is that, since the two notions are equivalent,
we can use the boolean formulation to prove the other one without
mentioning 500 explicitly:
Although we haven't gained much in terms of proof size in this
case, larger proofs can often be made considerably simpler by the
use of reflection. As an extreme example, the Coq proof of the
famous 4-color theorem uses reflection to reduce the analysis of
hundreds of different cases to a boolean computation. We won't
cover reflection in great detail, but it serves as a good example
showing the complementary strengths of booleans and general
propositions.
Exercise: 2 stars (logical_connectives)
The following lemmas relate the propositional connectives studied in this chapter to the corresponding boolean operations.Lemma andb_true_iff : ∀b1 b2:bool,
b1 && b2 = true ↔ b1 = true ∧ b2 = true.
Proof.
(* FILL IN HERE *) Admitted.
Lemma orb_true_iff : ∀b1 b2,
b1 || b2 = true ↔ b1 = true ∨ b2 = true.
Proof.
(* FILL IN HERE *) Admitted.
☐
Exercise: 1 star (beq_nat_false_iff)
The following theorem is an alternate "negative" formulation of beq_nat_true_iff that is more convenient in certain situations (we'll see examples in later chapters).Theorem beq_nat_false_iff : ∀x y : nat,
beq_nat x y = false ↔ x ≠ y.
Proof.
(* FILL IN HERE *) Admitted.
☐
Exercise: 3 stars (beq_list)
Given a boolean operator beq for testing equality of elements of some type A, we can define a function beq_list beq for testing equality of lists with elements in A. Complete the definition of the beq_list function below. To make sure that your definition is correct, prove the lemma beq_list_true_iff.Fixpoint beq_list {A} (beq : A → A → bool)
(l1 l2 : list A) : bool :=
(* FILL IN HERE *) admit.
Lemma beq_list_true_iff :
∀A (beq : A → A → bool),
(∀a1 a2, beq a1 a2 = true ↔ a1 = a2) →
∀l1 l2, beq_list beq l1 l2 = true ↔ l1 = l2.
Proof.
(* FILL IN HERE *) Admitted.
☐
Exercise: 2 stars, recommended (All_forallb)
Recall the function forallb, from the exercise forall_exists_challenge in chapter Tactics:Fixpoint forallb {X : Type} (test : X → bool) (l : list X) : bool :=
match l with
| [] ⇒ true
| x :: l' ⇒ andb (test x) (forallb test l')
end.
Prove the theorem below, which relates forallb to the All
property of the above exercise.
Theorem forallb_true_iff : ∀X test (l : list X),
forallb test l = true ↔ All (fun x ⇒ test x = true) l.
Proof.
(* FILL IN HERE *) Admitted.
Are there any important properties of the function forallb which
are not captured by your specification?
(* FILL IN HERE *)
☐
We have seen that it is not possible to test whether or not a
proposition P holds while defining a Coq function. You may be
surprised to learn that a similar restriction applies to proofs!
In other words, the following intuitive reasoning principle is not
derivable in Coq:
Classical vs. Constructive Logic
To understand operationally why this is the case, recall that, to
prove a statement of the form P ∨ Q, we use the left and
right tactics, which effectively require knowing which side of
the disjunction holds. However, the universally quantified P in
excluded_middle is an arbitrary proposition, which we know
nothing about. We don't have enough information to choose which
of left or right to apply, just as Coq doesn't have enough
information to mechanically decide whether P holds or not inside
a function. On the other hand, if we happen to know that P is
reflected in some boolean term b, then knowing whether it holds
or not is trivial: we just have to check the value of b. This
leads to the following theorem:
Theorem restricted_excluded_middle : ∀P b,
(P ↔ b = true) → P ∨ ¬ P.
Proof.
intros P [] H.
- left. rewrite H. reflexivity.
- right. rewrite H. intros contra. inversion contra.
Qed.
In particular, the excluded middle is valid for equations n = m,
between natural numbers n and m.
You may find it strange that the general excluded middle is not
available by default in Coq; after all, any given claim must be
either true or false. Nonetheless, there is an advantage in not
assuming the excluded middle: statements in Coq can make stronger
claims than the analogous statements in standard mathematics.
Notably, if there is a Coq proof of ∃ x, P x, it is
possible to explicitly exhibit a value of x for which we can
prove P x — in other words, every proof of existence is
necessarily constructive. Because of this, logics like Coq's,
which do not assume the excluded middle, are referred to as
constructive logics. More conventional logical systems such as
ZFC, in which the excluded middle does hold for arbitrary
propositions, are referred to as classical.
The following example illustrates why assuming the excluded middle
may lead to non-constructive proofs:
Claim: There exist irrational numbers a and b such that a ^
b is rational.
Proof: It is not difficult to show that sqrt 2 is irrational.
If sqrt 2 ^ sqrt 2 is rational, it suffices to take a = b =
sqrt 2 and we are done. Otherwise, sqrt 2 ^ sqrt 2 is
irrational. In this case, we can take a = sqrt 2 ^ sqrt 2 and
b = sqrt 2, since a ^ b = sqrt 2 ^ (sqrt 2 × sqrt 2) = sqrt 2 ^
2 = 2. ☐
Do you see what happened here? We used the excluded middle to
consider separately the cases where sqrt 2 ^ sqrt 2 is rational
and where it is not, without knowing which one actually holds!
Because of that, we wind up knowing that such a and b exist
but we cannot determine what their actual values are (at least,
using this line of argument).
As useful as constructive logic is, it does have its limitations:
There are many statements that can easily be proven in classical
logic but that have much more complicated constructive proofs, and
there are some that are known to have no constructive proof at
all! Fortunately, like functional extensionality, the excluded
middle is known to be compatible with Coq's logic, allowing us to
add it safely as an axiom. However, we will not need to do so in
this book: the results that we cover can be developed entirely
within constructive logic at negligible extra cost.
It takes some practice to understand which proof techniques must
be avoided in constructive reasoning, but arguments by
contradiction, in particular, are infamous for leading to
non-constructive proofs. Here's a typical example: suppose that
we want to show that there exists x with some property P,
i.e., such that P x. We start by assuming that our conclusion
is false; that is, ¬ ∃ x, P x. From this premise, it is not
hard to derive ∀ x, ¬ P x. If we manage to show that this
intermediate fact results in a contradiction, we arrive at an
existence proof without ever exhibiting a value of x for which
P x holds!
The technical flaw here, from a constructive standpoint, is that
we claimed to prove ∃ x, P x using a proof of ¬ ¬ ∃
x, P x. However, allowing ourselves to remove double negations
from arbitrary statements is equivalent to assuming the excluded
middle, as shown in one of the exercises below. Thus, this line
of reasoning cannot be encoded in Coq without assuming additional
axioms.
Exercise: 3 stars (excluded_middle_irrefutable)
The consistency of Coq with the general excluded middle axiom requires complicated reasoning that cannot be carried out within Coq itself. However, the following theorem implies that it is always safe to assume a decidability axiom (i.e., an instance of excluded middle) for any particular Prop P. Why? Because we cannot prove the negation of such an axiom; if we could, we would have both ¬ (P ∨ ¬P) and ¬ ¬ (P ∨ ¬P), a contradiction.
☐
Exercise: 3 stars, optional (not_exists_dist)
It is a theorem of classical logic that the following two assertions are equivalent:
¬ (∃x, ¬ P x)
∀x, P x
The dist_not_exists theorem above proves one side of this
equivalence. Interestingly, the other direction cannot be proved
in constructive logic. Your job is to show that it is implied by
the excluded middle.
∀x, P x
Theorem not_exists_dist :
excluded_middle →
∀(X:Type) (P : X → Prop),
¬ (∃x, ¬ P x) → (∀x, P x).
Proof.
(* FILL IN HERE *) Admitted.
☐
Prove that all five propositions (these four plus
excluded_middle) are equivalent.
Exercise: 5 stars, advanced, optional (classical_axioms)
For those who like a challenge, here is an exercise taken from the Coq'Art book by Bertot and Casteran (p. 123). Each of the following four statements, together with excluded_middle, can be considered as characterizing classical logic. We can't prove any of them in Coq, but we can consistently add any one of them as an axiom if we wish to work in classical logic.Definition peirce := ∀P Q: Prop,
((P→Q)→P)→P.
Definition double_negation_elimination := ∀P:Prop,
~~P → P.
Definition de_morgan_not_and_not := ∀P Q:Prop,
~(~P ∧ ¬Q) → P∨Q.
Definition implies_to_or := ∀P Q:Prop,
(P→Q) → (¬P∨Q).
(* FILL IN HERE *)
☐