lean2/examples/lean/abelian.lean
Leonardo de Moura 56ad1c1728 doc(examples/lean/abelian): proof of concept
Signed-off-by: Leonardo de Moura <leonardo@microsoft.com>
2014-02-10 18:17:15 -08:00

145 lines
7.3 KiB
Text

import macros
definition associative {A : (Type U)} (f : A → A → A) := ∀ x y z, f (f x y) z = f x (f y z)
definition is_identity {A : (Type U)} (f : A → A → A) (id : A) := ∀ x, f x id = x
definition inverse_ex {A : (Type U)} (f : A → A → A) (id : A) := ∀ x, ∃ y, f x y = id
universe s ≥ 1
definition group := sig A : (Type s), sig mul : A → A → A, sig one : A, (associative mul) # (is_identity mul one) # (inverse_ex mul one)
definition to_group (A : (Type s)) (mul : A → A → A) (one : A) (H1 : associative mul) (H2 : is_identity mul one) (H3 : inverse_ex mul one) : group
:= pair A (pair mul (pair one (pair H1 (pair H2 H3))))
-- The following definitions can be generated automatically.
definition carrier (g : group) := proj1 g
definition G_mul {g : group} : carrier g → carrier g → carrier g
:= proj1 (proj2 g)
infixl 70 * : G_mul
definition one {g : group} : carrier g
:= proj1 (proj2 (proj2 g))
theorem G_assoc {g : group} (x y z : carrier g) : (x * y) * z = x * (y * z)
:= (proj1 (proj2 (proj2 (proj2 g)))) x y z
theorem G_id {g : group} (x : carrier g) : x * one = x
:= (proj1 (proj2 (proj2 (proj2 (proj2 g))))) x
theorem G_inv {g : group} (x : carrier g) : ∃ y, x * y = one
:= (proj2 (proj2 (proj2 (proj2 (proj2 g))))) x
set_opaque group true
set_opaque carrier true
set_opaque G_mul true
set_opaque one true
-- First example: the pairwise product of two groups is a group
definition product (g1 g2 : group) : group
:= let S := carrier g1 # carrier g2,
-- It would be nice to be able to define local notation, and write _*_ instead of f
f := λ x y, pair (proj1 x * proj1 y) (proj2 x * proj2 y),
o := pair one one -- this is actually (pair (@one g1) (@one g2))
in have assoc : associative f,
-- The elaborator failed to infer the type of the pairs.
-- I had to annotate the pairs with their types.
from take x y z : S, -- We don't really need to provide S, but it will make the elaborator to work much harder
-- since * is an overloaded operator, we also have * as notation for Nat::mul in the context.
calc f (f x y) z = (pair ((proj1 x * proj1 y) * proj1 z) ((proj2 x * proj2 y) * proj2 z) : S) : refl (f (f x y) z)
... = (pair (proj1 x * (proj1 y * proj1 z)) ((proj2 x * proj2 y) * proj2 z) : S) : { G_assoc (proj1 x) (proj1 y) (proj1 z) }
... = (pair (proj1 x * (proj1 y * proj1 z)) (proj2 x * (proj2 y * proj2 z)) : S) : { G_assoc (proj2 x) (proj2 y) (proj2 z) }
... = f x (f y z) : refl (f x (f y z)),
have id : is_identity f o,
from take x : S,
calc f x o = (pair (proj1 x * one) (proj2 x * one) : S) : refl (f x o)
... = (pair (proj1 x) (proj2 x * one) : S) : { G_id (proj1 x) }
... = (pair (proj1 x) (proj2 x) : S) : { G_id (proj2 x) }
... = x : pair_proj_eq x,
have inv : inverse_ex f o,
from take x : S,
obtain (y1 : carrier g1) (Hy1 : proj1 x * y1 = one), from G_inv (proj1 x),
obtain (y2 : carrier g2) (Hy2 : proj2 x * y2 = one), from G_inv (proj2 x),
show ∃ y, f x y = o,
from exists_intro (pair y1 y2 : S)
(calc f x (pair y1 y2 : S) = (pair (proj1 x * y1) (proj2 x * y2) : S) : refl (f x (pair y1 y2 : S))
... = (pair one (proj2 x * y2) : S) : { Hy1 }
... = (pair one one : S) : { Hy2 }
... = o : refl o),
to_group S f o assoc id inv
set_opaque product true
-- It would be nice to be able to write x.first and x.second instead of (proj1 x) and (proj2 x)
-- Remark: * is overloaded since Lean loads Nat.lean by default.
-- The type errors related to * are quite cryptic because of that
-- Use 'star' for creating products
infixr 50 ⋆ : product
-- It would be nice to be able to write (p1 p2 : g1 ⋆ g2 ⋆ g3)
check λ (g1 g2 g3 : group) (p1 p2 : carrier (g1 ⋆ g2 ⋆ g3)), p1 * p2 = p2 * p1
theorem group_inhab (g : group) : inhabited (carrier g)
:= inhabited_intro (@one g)
definition inv {g : group} (a : carrier g) : carrier g
:= ε (group_inhab g) (λ x : carrier g, a * x = one)
theorem G_idl {g : group} (x : carrier g) : x * one = x
:= G_id x
theorem G_invl {g : group} (x : carrier g) : x * inv x = one
:= obtain (y : carrier g) (Hy : x * y = one), from G_inv x,
eps_ax (group_inhab g) y Hy
set_opaque inv true
theorem G_inv_aux {g : group} (x : carrier g) : inv x = (inv x * x) * inv x
:= symm (calc (inv x * x) * inv x = inv x * (x * inv x) : G_assoc (inv x) x (inv x)
... = inv x * one : { G_invl x }
... = inv x : G_idl (inv x))
theorem G_invr {g : group} (x : carrier g) : inv x * x = one
:= calc inv x * x = (inv x * x) * one : symm (G_idl (inv x * x))
... = (inv x * x) * (inv x * inv (inv x)) : { symm (G_invl (inv x)) }
... = ((inv x * x) * inv x) * inv (inv x) : symm (G_assoc (inv x * x) (inv x) (inv (inv x)))
... = (inv x * (x * inv x)) * inv (inv x) : { G_assoc (inv x) x (inv x) }
... = (inv x * one) * inv (inv x) : { G_invl x }
... = (inv x) * inv (inv x) : { G_idl (inv x) }
... = one : G_invl (inv x)
theorem G_idr {g : group} (x : carrier g) : one * x = x
:= calc one * x = (x * inv x) * x : { symm (G_invl x) }
... = x * (inv x * x) : G_assoc x (inv x) x
... = x * one : { G_invr x }
... = x : G_idl x
theorem G_inv_inv {g : group} (x : carrier g) : inv (inv x) = x
:= calc inv (inv x) = inv (inv x) * one : symm (G_idl (inv (inv x)))
... = inv (inv x) * (inv x * x) : { symm (G_invr x) }
... = (inv (inv x) * inv x) * x : symm (G_assoc (inv (inv x)) (inv x) x)
... = one * x : { G_invr (inv x) }
... = x : G_idr x
definition commutative {A : (Type U)} (f : A → A → A) := ∀ x y, f x y = f y x
definition abelian_group := sig g : group, commutative (@G_mul g)
definition ab_to_g (ag : abelian_group) : group
:= proj1 ag
-- Coercions currently only work with opaque types
-- We must first define "extract" the information we want, and then
-- mark abelian_group as opaque
definition AG_comm {g : abelian_group} (x y : carrier (ab_to_g g)) : x * y = y * x
:= (proj2 g) x y
set_opaque abelian_group true
set_opaque ab_to_g true
set_opaque AG_comm true
coercion ab_to_g
-- Now, we can use abelian groups where groups are expected.
theorem AG_left_comm {g : abelian_group} (x y z : carrier g) : x * (y * z) = y * (x * z)
:= calc x * (y * z) = (x * y) * z : symm (G_assoc x y z)
... = (y * x) * z : { AG_comm x y }
... = y * (x * z) : G_assoc y x z