SLIDE 1 1
redtt
university of minnesota
favonia
joint work with Carlo Angiuli, Evan Cavallo, Robert Harper, Anders Mörtberg and Jonathan Sterling
cartesian cubical proof assistant
SLIDE 2
2
type theory
Γ ⊦ A type Γ ⊦ A = B type Γ ⊦ M : A Γ ⊦ M = N : A
SLIDE 3
3
cubical type theory
formal intervals 𝕁 Γ ⊦ 0:𝕁 Γ ⊦ 1:𝕁 Γ ⊦ x:𝕁 x:𝕁 ∈ Γ Γ ⊦ r∧s:𝕁 Γ ⊦ r:𝕁 Γ ⊦ s:𝕁 Γ ⊦ r∨s:𝕁 Γ ⊦ r:𝕁 Γ ⊦ s:𝕁 Γ ⊦ ¬r:𝕁 Γ ⊦ r:𝕁
SLIDE 4
4
cubical type theory
formal intervals 𝕁
⬄ M is an n-cube in A x1:𝕁, x2:𝕁, ..., xn:𝕁 ⊦ M : A
M⟨0/x⟩ M⟨1/x⟩ M⟨y/x⟩ x y
SLIDE 5 5
cubical type theory
formal intervals 𝕁
- rdinary typing rules hold uniformly
Γ ⊦ λa.M : (a:A) → B Γ, a:A ⊦ M : B with any number of 𝕁 in the Γ
SLIDE 6 5
cubical type theory
formal intervals 𝕁
- rdinary typing rules hold uniformly
Γ ⊦ λa.M : (a:A) → B Γ, a:A ⊦ M : B with any number of 𝕁 in the Γ function extensionality due to dimensions commuting with function application
SLIDE 7
6
cubical type theory
formal intervals 𝕁 canonicity any closed term of ℕ is equal to some numeral type-theory tango: internalization of judgmental structure, harmony
SLIDE 8
7
cubical type theories
base category most developed: cartesian, de morgan structural rules + operators {0,1,∧,∨,¬,...}
SLIDE 9
7
cubical type theories
base category most developed: cartesian, de morgan structural rules + operators {0,1,∧,∨,¬,...} kan structure cofibrations, fiberwise fibrant replacement
SLIDE 10
7
cubical type theories
base category most developed: cartesian, de morgan structural rules + operators {0,1,∧,∨,¬,...} kan structure cofibrations, fiberwise fibrant replacement mythos proofs or realizers?
SLIDE 11 8 RedPRL yacctt redtt Agda cubicaltt
cartesian r⇝s, r=s de morgan 0⇝1, i=0/1 realizers proofs
SLIDE 12 8 RedPRL yacctt redtt Agda cubicaltt
cartesian r⇝s, r=s de morgan 0⇝1, i=0/1 cmu chalmers gothenburg realizers proofs spartan fancy
RedPRL yacctt redtt Agda cubicaltt
SLIDE 13
9
redtt specialities
higher inductive types two-level type theory nbe-like algorithm (conjectured correct) extension types judgmental refinements holes, tactics, unification
SLIDE 14 9
redtt specialities
higher inductive types two-level type theory nbe-like algorithm (conjectured correct) extension types judgmental refinements holes, tactics, unification
see demo
SLIDE 15
10
redtt specialities
higher inductive types a general schema; indexed ones on the way see chtt part 4 [Cavallo & Harper]
SLIDE 16
11
redtt specialities
two-level type theory
type pretype (no equality types yet)
SLIDE 17
12
redtt specialities
todo: many-level type theory
type pretype discrete discrete: paths equal to equality consistent with (strict) UIP
SLIDE 18
13
redtt specialities
nbe algorithm cubicaltt adopts a similar one
SLIDE 19
13
redtt specialities
nbe algorithm cubicaltt adopts a similar one difficulty 1: value re-evaluation: loopx[0/x] difficulty 2: constraints: r=s
SLIDE 20
13
redtt specialities
nbe algorithm cubicaltt adopts a similar one difficulty 1: value re-evaluation: loopx[0/x] difficulty 2: constraints: r=s decidable: Φ ⊧ r = s
SLIDE 21
14
todo
correctness of nbe
SLIDE 22
14
todo
correctness of nbe equality types
SLIDE 23
14
todo
correctness of nbe equality types user-defined tactic, pattern matching, etc
SLIDE 24
14
todo
correctness of nbe improved kan operations of universes equality types user-defined tactic, pattern matching, etc
SLIDE 25
14
todo
correctness of nbe improved kan operations of universes synthetic homotopy theory (!) equality types user-defined tactic, pattern matching, etc
SLIDE 26 15
synthetic homotopy theory
redtt RedPRL standard homotopy theory
"obvious" todo