module plfa.part1.Connectives where

This chapter introduces the basic logical connectives, by observing a correspondence between connectives of logic and data types, a principle known as Propositions as Types:

  • conjunction is product,
  • disjunction is sum,
  • true is unit type,
  • false is empty type,
  • implication is function space.

Imports

import Relation.Binary.PropositionalEquality as Eq
open Eq using (_≡_; refl)
open Eq.≡-Reasoning
open import Data.Nat using ()
open import Function using (_∘_)
open import plfa.part1.Isomorphism using (_≃_; _≲_; extensionality; _⇔_)
open plfa.part1.Isomorphism.≃-Reasoning

Conjunction is product

Given two propositions A and B, the conjunction A × B holds if both A holds and B holds. We formalise this idea by declaring a suitable datatype:
data _×_ (A B : Set) : Set where

  ⟨_,_⟩ :
      A
     B
      -----
     A × B

Evidence that A × B holds is of the form ⟨ M , N ⟩, where M provides evidence that A holds and N provides evidence that B holds.

Given evidence that A × B holds, we can conclude that both A holds and B holds:
proj₁ :  {A B : Set}
   A × B
    -----
   A
proj₁  x , y  = x

proj₂ :  {A B : Set}
   A × B
    -----
   B
proj₂  x , y  = y

If L provides evidence that A × B holds, then proj₁ L provides evidence that A holds, and proj₂ L provides evidence that B holds.

When ⟨_,_⟩ appears in a term on the right-hand side of an equation we refer to it as a constructor, and when it appears in a pattern on the left-hand side of an equation we refer to it as a destructor. We may also refer to proj₁ and proj₂ as destructors, since they play a similar role.

Other terminology refers to ⟨_,_⟩ as introducing a conjunction, and to proj₁ and proj₂ as eliminating a conjunction; indeed, the former is sometimes given the name ×-I and the latter two the names ×-E₁ and ×-E₂. As we read the rules from top to bottom, introduction and elimination do what they say on the tin: the first introduces a formula for the connective, which appears in the conclusion but not in the hypotheses; the second eliminates a formula for the connective, which appears in a hypothesis but not in the conclusion. An introduction rule describes under what conditions we say the connective holds—how to define the connective. An elimination rule describes what we may conclude when the connective holds—how to use the connective.1

In this case, applying each destructor and reassembling the results with the constructor is the identity over products:
η-× :  {A B : Set} (w : A × B)   proj₁ w , proj₂ w   w
η-×  x , y  = refl

The pattern matching on the left-hand side is essential, since replacing w by ⟨ x , y ⟩ allows both sides of the propositional equality to simplify to the same term.

We set the precedence of conjunction so that it binds less tightly than anything save disjunction:
infixr 2 _×_

Thus, m ≤ n × n ≤ p parses as (m ≤ n) × (n ≤ p).

Alternatively, we can declare conjunction as a record type:
record _×′_ (A B : Set) : Set where
  constructor ⟨_,_⟩′
  field
    proj₁′ : A
    proj₂′ : B
open _×′_

The record construction record { proj₁′ = M ; proj₂′ = N } corresponds to the term ⟨ M , N ⟩ where M is a term of type A and N is a term of type B. The constructor declaration allows us to write ⟨ M , N ⟩′ in place of the record construction.

The data type _×_ and the record type _×′_ behave similarly. One difference is that for data types we have to prove η-equality, but for record types, η-equality holds by definition. While proving η-×′, we do not have to pattern match on w to know that η-equality holds:
η-×′ :  {A B : Set} (w : A ×′ B)   proj₁′ w , proj₂′ w ⟩′  w
η-×′ w = refl

It can be very convenient to have η-equality definitionally, and so the standard library defines _×_ as a record type. We use the definition from the standard library in later chapters.

Given two types A and B, we refer to A × B as the product of A and B. In set theory, it is also sometimes called the Cartesian product, and in computing it corresponds to a record type. Among other reasons for calling it the product, note that if type A has m distinct members, and type B has n distinct members, then the type A × B has m * n distinct members. For instance, consider a type Bool with two members, and a type Tri with three members:
data Bool : Set where
  true  : Bool
  false : Bool

data Tri : Set where
  aa : Tri
  bb : Tri
  cc : Tri

Then the type Bool × Tri has six members:

⟨ true  , aa ⟩    ⟨ true  , bb ⟩    ⟨ true ,  cc ⟩
⟨ false , aa ⟩    ⟨ false , bb ⟩    ⟨ false , cc ⟩
For example, the following function enumerates all possible arguments of type Bool × Tri:
×-count : Bool × Tri  
×-count  true  , aa   =  1
×-count  true  , bb   =  2
×-count  true  , cc   =  3
×-count  false , aa   =  4
×-count  false , bb   =  5
×-count  false , cc   =  6

Product on types also shares a property with product on numbers in that there is a sense in which it is commutative and associative. In particular, product is commutative and associative up to isomorphism.

For commutativity, the to function swaps a pair, taking ⟨ x , y ⟩ to ⟨ y , x ⟩, and the from function does the same (up to renaming). Instantiating the patterns correctly in from∘to and to∘from is essential. Replacing the definition of from∘to by λ w → refl will not work; and similarly for to∘from:
×-comm :  {A B : Set}  A × B  B × A
×-comm =
  record
    { to       =  λ{  x , y    y , x  }
    ; from     =  λ{  y , x    x , y  }
    ; from∘to  =  λ{  x , y   refl }
    ; to∘from  =  λ{  y , x   refl }
    }

Being commutative is different from being commutative up to isomorphism. Compare the two statements:

m * n ≡ n * m
A × B ≃ B × A

In the first case, we might have that m is 2 and n is 3, and both m * n and n * m are equal to 6. In the second case, we might have that A is Bool and B is Tri, and Bool × Tri is not the same as Tri × Bool. But there is an isomorphism between the two types. For instance, ⟨ true , aa ⟩, which is a member of the former, corresponds to ⟨ aa , true ⟩, which is a member of the latter.

For associativity, the to function reassociates two uses of pairing, taking ⟨ ⟨ x , y ⟩ , z ⟩ to ⟨ x , ⟨ y , z ⟩ ⟩, and the from function does the inverse. Again, the evidence of left and right inverse requires matching against a suitable pattern to enable simplification:
×-assoc :  {A B C : Set}  (A × B) × C  A × (B × C)
×-assoc =
  record
    { to      = λ{   x , y  , z    x ,  y , z   }
    ; from    = λ{  x ,  y , z      x , y  , z  }
    ; from∘to = λ{   x , y  , z   refl }
    ; to∘from = λ{  x ,  y , z    refl }
    }

Being associative is not the same as being associative up to isomorphism. Compare the two statements:

(m * n) * p ≡ m * (n * p)
(A × B) × C ≃ A × (B × C)

For example, the type (ℕ × Bool) × Tri is not the same as ℕ × (Bool × Tri). But there is an isomorphism between the two types. For instance ⟨ ⟨ 1 , true ⟩ , aa ⟩, which is a member of the former, corresponds to ⟨ 1 , ⟨ true , aa ⟩ ⟩, which is a member of the latter.

Exercise ⇔≃× (practice)

Show that A ⇔ B as defined earlier is isomorphic to (A → B) × (B → A).

-- Your code goes here

Truth is unit

Truth always holds. We formalise this idea by declaring a suitable datatype:
data  : Set where

  tt :
    --
    

Evidence that holds is of the form tt.

There is an introduction rule, but no elimination rule. Given evidence that holds, there is nothing more of interest we can conclude. Since truth always holds, knowing that it holds tells us nothing new.

The nullary case of η-× is η-⊤, which asserts that any value of type must be equal to tt:
η-⊤ :  (w : )  tt  w
η-⊤ tt = refl

The pattern matching on the left-hand side is essential. Replacing w by tt allows both sides of the propositional equality to simplify to the same term.

Alternatively, we can declare truth as an empty record:
record ⊤′ : Set where
  constructor tt′

The record construction record {} corresponds to the term tt. The constructor declaration allows us to write tt′.

As with the product, the data type and the record type ⊤′ behave similarly, but η-equality holds by definition for the record type. While proving η-⊤′, we do not have to pattern match on w—Agda knows it is equal to tt′:
η-⊤′ :  (w : ⊤′)  tt′  w
η-⊤′ w = refl
Agda knows that any value of type ⊤′ must be tt′, so any time we need a value of type ⊤′, we can tell Agda to figure it out:
truth′ : ⊤′
truth′ = _
We refer to as the unit type. And, indeed, type has exactly one member, tt. For example, the following function enumerates all possible arguments of type :
⊤-count :   
⊤-count tt = 1
For numbers, one is the identity of multiplication. Correspondingly, unit is the identity of product up to isomorphism. For left identity, the to function takes ⟨ tt , x ⟩ to x, and the from function does the inverse. The evidence of left inverse requires matching against a suitable pattern to enable simplification:
⊤-identityˡ :  {A : Set}   × A  A
⊤-identityˡ =
  record
    { to      = λ{  tt , x   x }
    ; from    = λ{ x   tt , x  }
    ; from∘to = λ{  tt , x   refl }
    ; to∘from = λ{ x  refl }
    }

Having an identity is different from having an identity up to isomorphism. Compare the two statements:

1 * m ≡ m
⊤ × A ≃ A

In the first case, we might have that m is 2, and both 1 * m and m are equal to 2. In the second case, we might have that A is Bool, and ⊤ × Bool is not the same as Bool. But there is an isomorphism between the two types. For instance, ⟨ tt , true ⟩, which is a member of the former, corresponds to true, which is a member of the latter.

Right identity follows from commutativity of product and left identity:
⊤-identityʳ :  {A : Set}  (A × )  A
⊤-identityʳ {A} =
  ≃-begin
    (A × )
  ≃⟨ ×-comm 
    ( × A)
  ≃⟨ ⊤-identityˡ 
    A
  ≃-∎

Here we have used a chain of isomorphisms, analogous to that used for equality.

Disjunction is sum

Given two propositions A and B, the disjunction A ⊎ B holds if either A holds or B holds. We formalise this idea by declaring a suitable inductive type:
data _⊎_ (A B : Set) : Set where

  inj₁ :
      A
      -----
     A  B

  inj₂ :
      B
      -----
     A  B

Evidence that A ⊎ B holds is either of the form inj₁ M, where M provides evidence that A holds, or inj₂ N, where N provides evidence that B holds.

Given evidence that A → C and B → C both hold, then given evidence that A ⊎ B holds we can conclude that C holds:
case-⊎ :  {A B C : Set}
   (A  C)
   (B  C)
   A  B
    -----------
   C
case-⊎ f g (inj₁ x) = f x
case-⊎ f g (inj₂ y) = g y

Pattern matching against inj₁ and inj₂ is typical of how we exploit evidence that a disjunction holds.

When inj₁ and inj₂ appear on the right-hand side of an equation we refer to them as constructors, and when they appear on the left-hand side we refer to them as destructors. We also refer to case-⊎ as a destructor, since it plays a similar role. Other terminology refers to inj₁ and inj₂ as introducing a disjunction, and to case-⊎ as eliminating a disjunction; indeed the former are sometimes given the names ⊎-I₁ and ⊎-I₂ and the latter the name ⊎-E.

Applying the destructor to each of the constructors is the identity:
η-⊎ :  {A B : Set} (w : A  B)  case-⊎ inj₁ inj₂ w  w
η-⊎ (inj₁ x) = refl
η-⊎ (inj₂ y) = refl
More generally, we can also throw in an arbitrary function from a disjunction:
uniq-⊎ :  {A B C : Set} (h : A  B  C) (w : A  B) 
  case-⊎ (h  inj₁) (h  inj₂) w  h w
uniq-⊎ h (inj₁ x) = refl
uniq-⊎ h (inj₂ y) = refl

The pattern matching on the left-hand side is essential. Replacing w by inj₁ x allows both sides of the propositional equality to simplify to the same term, and similarly for inj₂ y.

We set the precedence of disjunction so that it binds less tightly than any other declared operator:
infixr 1 _⊎_

Thus, A × C ⊎ B × C parses as (A × C) ⊎ (B × C).

Given two types A and B, we refer to A ⊎ B as the sum of A and B. In set theory, it is also sometimes called the disjoint union, and in computing it corresponds to a variant record type. Among other reasons for calling it the sum, note that if type A has m distinct members, and type B has n distinct members, then the type A ⊎ B has m + n distinct members. For instance, consider a type Bool with two members, and a type Tri with three members, as defined earlier. Then the type Bool ⊎ Tri has five members:

inj₁ true     inj₂ aa
inj₁ false    inj₂ bb
              inj₂ cc
For example, the following function enumerates all possible arguments of type Bool ⊎ Tri:
⊎-count : Bool  Tri  
⊎-count (inj₁ true)   =  1
⊎-count (inj₁ false)  =  2
⊎-count (inj₂ aa)     =  3
⊎-count (inj₂ bb)     =  4
⊎-count (inj₂ cc)     =  5

Sum on types also shares a property with sum on numbers in that it is commutative and associative up to isomorphism.

Show sum is commutative up to isomorphism.

-- Your code goes here

Exercise ⊎-assoc (practice)

Show sum is associative up to isomorphism.

-- Your code goes here

False is empty

False never holds. We formalise this idea by declaring a suitable inductive type:
data  : Set where
  -- no clauses!

There is no possible evidence that holds.

Dual to , for there is no introduction rule but an elimination rule. Since false never holds, knowing that it holds tells us we are in a paradoxical situation. Given evidence that holds, we might conclude anything! This is a basic principle of logic, known in medieval times by the Latin phrase ex falso, and known to children through phrases such as “if pigs had wings, then I’d be the Queen of Sheba”. We formalise it as follows:
⊥-elim :  {A : Set}
   
    --
   A
⊥-elim ()

This is our first use of the absurd pattern (). Here since is a type with no members, we indicate that it is never possible to match against a value of this type by using the pattern ().

The nullary case of case-⊎ is ⊥-elim. By analogy, we might have called it case-⊥, but chose to stick with the name in the standard library.

The nullary case of uniq-⊎ is uniq-⊥, which asserts that ⊥-elim is equal to any arbitrary function from :
uniq-⊥ :  {C : Set} (h :   C) (w : )  ⊥-elim w  h w
uniq-⊥ h ()

Using the absurd pattern asserts there are no possible values for w, so the equation holds trivially.

We refer to as the empty type. And, indeed, type has no members. For example, the following function enumerates all possible arguments of type :
⊥-count :   
⊥-count ()

Here again the absurd pattern () indicates that no value can match type .

For numbers, zero is the identity of addition. Correspondingly, empty is the identity of sums up to isomorphism.

Show empty is the left identity of sums up to isomorphism.

-- Your code goes here

Exercise ⊥-identityʳ (practice)

Show empty is the right identity of sums up to isomorphism.

-- Your code goes here

Implication is function

Given two propositions A and B, the implication A → B holds if whenever A holds then B must also hold. We formalise implication using the function type, which has appeared throughout this book.

Evidence that A → B holds is of the form

λ (x : A) → N

where N is a term of type B containing as a free variable x of type A. Given a term L providing evidence that A → B holds, and a term M providing evidence that A holds, the term L M provides evidence that B holds. In other words, evidence that A → B holds is a function that converts evidence that A holds into evidence that B holds.

Put another way, if we know that A → B and A both hold, then we may conclude that B holds:
→-elim :  {A B : Set}
   (A  B)
   A
    -------
   B
→-elim L M = L M

In medieval times, this rule was known by the name modus ponens. It corresponds to function application.

Defining a function, with a named definition or a lambda abstraction, is referred to as introducing a function, while applying a function is referred to as eliminating the function.

Elimination followed by introduction is the identity:
η-→ :  {A B : Set} (f : A  B)   (x : A)  f x)  f
η-→ f = refl

Implication binds less tightly than any other operator. Thus, A ⊎ B → B ⊎ A parses as (A ⊎ B) → (B ⊎ A).

Given two types A and B, we refer to A → B as the function space from A to B. It is also sometimes called the exponential, with B raised to the A power. Among other reasons for calling it the exponential, note that if type A has m distinct members, and type B has n distinct members, then the type A → B has nᵐ distinct members. For instance, consider a type Bool with two members and a type Tri with three members, as defined earlier. Then the type Bool → Tri has nine (that is, three squared) members:

λ{true → aa; false → aa}  λ{true → aa; false → bb}  λ{true → aa; false → cc}
λ{true → bb; false → aa}  λ{true → bb; false → bb}  λ{true → bb; false → cc}
λ{true → cc; false → aa}  λ{true → cc; false → bb}  λ{true → cc; false → cc}
For example, the following function enumerates all possible arguments of the type Bool → Tri:
→-count : (Bool  Tri)  
→-count f with f true | f false
...          | aa     | aa      =   1
...          | aa     | bb      =   2
...          | aa     | cc      =   3
...          | bb     | aa      =   4
...          | bb     | bb      =   5
...          | bb     | cc      =   6
...          | cc     | aa      =   7
...          | cc     | bb      =   8
...          | cc     | cc      =   9

Exponential on types also share a property with exponential on numbers in that many of the standard identities for numbers carry over to the types.

Corresponding to the law

(p ^ n) ^ m  ≡  p ^ (n * m)

we have the isomorphism

A → (B → C)  ≃  (A × B) → C
Both types can be viewed as functions that given evidence that A holds and evidence that B holds can return evidence that C holds. This isomorphism sometimes goes by the name currying. The proof of the right inverse requires extensionality:
currying :  {A B C : Set}  (A  B  C)  (A × B  C)
currying =
  record
    { to      =  λ{ f  λ{  x , y   f x y }}
    ; from    =  λ{ g  λ{ x  λ{ y  g  x , y  }}}
    ; from∘to =  λ{ f  refl }
    ; to∘from =  λ{ g  extensionality λ{  x , y   refl }}
    }

Currying tells us that instead of a function that takes a pair of arguments, we can have a function that takes the first argument and returns a function that expects the second argument. Thus, for instance, our way of writing addition

_+_ : ℕ → ℕ → ℕ

is isomorphic to a function that accepts a pair of arguments:

_+′_ : (ℕ × ℕ) → ℕ

Agda is optimised for currying, so 2 + 3 abbreviates _+_ 2 3. In a language optimised for pairing, we would instead take 2 +′ 3 as an abbreviation for _+′_ ⟨ 2 , 3 ⟩.

Corresponding to the law

p ^ (n + m) = (p ^ n) * (p ^ m)

we have the isomorphism:

(A ⊎ B) → C  ≃  (A → C) × (B → C)
That is, the assertion that if either A holds or B holds then C holds is the same as the assertion that if A holds then C holds and if B holds then C holds. The proof of the left inverse requires extensionality:
→-distrib-⊎ :  {A B C : Set}  (A  B  C)  ((A  C) × (B  C))
→-distrib-⊎ =
  record
    { to      = λ{ f   f  inj₁ , f  inj₂  }
    ; from    = λ{  g , h   λ{ (inj₁ x)  g x ; (inj₂ y)  h y } }
    ; from∘to = λ{ f  extensionality λ{ (inj₁ x)  refl ; (inj₂ y)  refl } }
    ; to∘from = λ{  g , h   refl }
    }

Corresponding to the law

(p * n) ^ m = (p ^ m) * (n ^ m)

we have the isomorphism:

A → B × C  ≃  (A → B) × (A → C)
That is, the assertion that if A holds then B holds and C holds is the same as the assertion that if A holds then B holds and if A holds then C holds. The proof of left inverse requires both extensionality and the rule η-× for products:
→-distrib-× :  {A B C : Set}  (A  B × C)  (A  B) × (A  C)
→-distrib-× =
  record
    { to      = λ{ f   proj₁  f , proj₂  f  }
    ; from    = λ{  g , h   λ x   g x , h x  }
    ; from∘to = λ{ f  extensionality λ{ x  η-× (f x) } }
    ; to∘from = λ{  g , h   refl }
    }

Distribution

Products distribute over sum, up to isomorphism. The code to validate this fact is similar in structure to our previous results:
×-distrib-⊎ :  {A B C : Set}  (A  B) × C  (A × C)  (B × C)
×-distrib-⊎ =
  record
    { to      = λ{  inj₁ x , z   (inj₁  x , z )
                 ;  inj₂ y , z   (inj₂  y , z )
                 }
    ; from    = λ{ (inj₁  x , z )   inj₁ x , z 
                 ; (inj₂  y , z )   inj₂ y , z 
                 }
    ; from∘to = λ{  inj₁ x , z   refl
                 ;  inj₂ y , z   refl
                 }
    ; to∘from = λ{ (inj₁  x , z )  refl
                 ; (inj₂  y , z )  refl
                 }
    }
Sums do not distribute over products up to isomorphism, but it is an embedding:
⊎-distrib-× :  {A B C : Set}  (A × B)  C  (A  C) × (B  C)
⊎-distrib-× =
  record
    { to      = λ{ (inj₁  x , y )   inj₁ x , inj₁ y 
                 ; (inj₂ z)           inj₂ z , inj₂ z 
                 }
    ; from    = λ{  inj₁ x , inj₁ y   (inj₁  x , y )
                 ;  inj₁ x , inj₂ z   (inj₂ z)
                 ;  inj₂ z , _        (inj₂ z)
                 }
    ; from∘to = λ{ (inj₁  x , y )  refl
                 ; (inj₂ z)          refl
                 }
    }

Note that there is a choice in how we write the from function. As given, it takes ⟨ inj₂ z , inj₂ z′ ⟩ to inj₂ z, but it is easy to write a variant that instead returns inj₂ z′. We have an embedding rather than an isomorphism because the from function must discard either z or z′ in this case.

In the usual approach to logic, both of the distribution laws are given as equivalences, where each side implies the other:

A × (B ⊎ C) ⇔ (A × B) ⊎ (A × C)
A ⊎ (B × C) ⇔ (A ⊎ B) × (A ⊎ C)

But when we consider the functions that provide evidence for these implications, then the first corresponds to an isomorphism while the second only corresponds to an embedding, revealing a sense in which one of these laws is “more true” than the other.

Show that the following property holds:
postulate
  ⊎-weak-× :  {A B C : Set}  (A  B) × C  A  (B × C)

This is called a weak distributive law. Give the corresponding distributive law, and explain how it relates to the weak version.

-- Your code goes here

Exercise ⊎×-implies-×⊎ (practice)

Show that a disjunct of conjuncts implies a conjunct of disjuncts:
postulate
  ⊎×-implies-×⊎ :  {A B C D : Set}  (A × B)  (C × D)  (A  C) × (B  D)

Does the converse hold? If so, prove; if not, give a counterexample.

-- Your code goes here

Standard library

Definitions similar to those in this chapter can be found in the standard library:
import Data.Product using (_×_; proj₁; proj₂) renaming (_,_ to ⟨_,_⟩)
import Data.Unit using (; tt)
import Data.Sum using (_⊎_; inj₁; inj₂) renaming ([_,_] to case-⊎)
import Data.Empty using (; ⊥-elim)
import Function.Equivalence using (_⇔_)

The standard library constructs pairs with _,_ whereas we use ⟨_,_⟩. The former makes it convenient to build triples or larger tuples from pairs, permitting a , b , c to stand for (a , (b , c)). But it conflicts with other useful notations, such as [_,_] to construct a list of two elements in Chapter Lists and Γ , A to extend environments in Chapter DeBruijn. The standard library _⇔_ is similar to ours, but the one in the standard library is less convenient, since it is parameterised with respect to an arbitrary notion of equivalence.

Unicode

This chapter uses the following unicode:

×  U+00D7  MULTIPLICATION SIGN (\x)
⊎  U+228E  MULTISET UNION (\u+)
⊤  U+22A4  DOWN TACK (\top)
⊥  U+22A5  UP TACK (\bot)
η  U+03B7  GREEK SMALL LETTER ETA (\eta)
₁  U+2081  SUBSCRIPT ONE (\_1)
₂  U+2082  SUBSCRIPT TWO (\_2)
⇔  U+21D4  LEFT RIGHT DOUBLE ARROW (\<=>)

  1. This paragraph was adopted from “Propositions as Types”, Philip Wadler, Communications of the ACM, December 2015.↩︎