File size: 7,275 Bytes
4a56cd6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
import OAI.Algorithms.SubexponentialSAT.SqrtLogSpace.Counters
import OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutRankSignature
import Mathlib.Data.Vector.Basic
import Mathlib.Data.Fin.Tuple.Basic
import Mathlib.SetTheory.Cardinal.Finite
import Mathlib.Algebra.Order.BigOperators.Group.Finset

/-!
# First occurrences of attained keys

The subtype below is a specification of emissions, not a stored family.
The executable scan checks a candidate against its predecessors one at a time.
The key may be an entire integer row-signature; no coordinate projection occurs.
-/

namespace OAI.Algorithms.SubexponentialSAT.SqrtLogSpace.AttainedSignatures

variable {N : ℕ} {κ : Type*} (key : Fin N → κ)

abbrev Representative := {i : Fin N // ∀ j, j < i → key j ≠ key i}

def toAttained (i : Representative key) : Set.range key := ⟨key i, i, rfl⟩

theorem toAttained_injective : Function.Injective (toAttained key) := by
  intro i j h
  have he : key i.val = key j.val := congrArg Subtype.val h
  apply Subtype.ext
  rcases lt_trichotomy i.val j.val with hij | hij | hij
  · exact (j.property i.val hij he).elim
  · exact hij
  · exact (i.property j.val hij he.symm).elim

theorem toAttained_surjective : Function.Surjective (toAttained key) := by
  classical
  intro y
  obtain ⟨j, hj⟩ := y.property
  have hex : ∃ i, key i = y.val := ⟨j, hj⟩
  let i := Fin.find (fun i => key i = y.val) hex
  have hkey := Fin.find_spec hex
  refine ⟨⟨i, ?_⟩, Subtype.ext hkey⟩
  intro j hj hji
  exact Fin.find_min hex hj (hji.trans hkey)

theorem toAttained_bijective : Function.Bijective (toAttained key) :=
  ⟨toAttained_injective key, toAttained_surjective key⟩

noncomputable def representativeEquiv : Representative key ≃ Set.range key :=
  Equiv.ofBijective (toAttained key) (toAttained_bijective key)

@[simp] theorem representativeEquiv_value (i : Representative key) :
    (representativeEquiv key i).val = key i := by
  rw [representativeEquiv, Equiv.ofBijective_apply]
  rfl

theorem card_representative : Nat.card (Representative key) = Nat.card (Set.range key) :=
  Nat.card_congr (representativeEquiv key)

theorem representative_exists (hN : 0 < N) : Nonempty (Representative key) :=
  ⟨⟨⟨0, hN⟩, by intro j hj; exact (Nat.not_lt_zero _ hj).elim⟩⟩

theorem card_representative_pos (hN : 0 < N) : 0 < Nat.card (Representative key) := by
  let _ := representative_exists key hN
  exact Nat.card_pos

theorem card_representative_one (key : Fin 1 → κ) : Nat.card (Representative key) = 1 := by
  have h : ∀ i : Fin 1, ∀ j : Fin 1, ¬j < i := by
    intro i j hij
    have hi := i.isLt
    have hj := j.isLt
    omega
  let e : Representative key ≃ Fin 1 :=
    { toFun := Subtype.val
      invFun := fun i => ⟨i, fun j hij => (h i j hij).elim⟩
      left_inv := fun _ => rfl
      right_inv := fun _ => rfl }
  exact (Nat.card_congr e).trans (by simp)

/-- Increasing numeric order of little-endian words, equivalently lexicographic
order after reversing the words. The empty head has its unique empty word. -/
def assignmentEquiv (h : ℕ) : Fin (2 ^ h) ≃ (Fin h → Bool) :=
  (Counters.wordEquiv h).trans (Equiv.vectorEquivFin Bool h)

theorem assignment_word (h : ℕ) (i : Fin (2 ^ h)) :
    List.ofFn (assignmentEquiv h i) =
      ExecCNFTupleEnumeration.fixedBits h i := by
  have hv := congrArg List.Vector.toList (List.Vector.ofFn_get (Counters.wordEquiv h i))
  change List.ofFn (List.Vector.get (Counters.wordEquiv h i)) =
    List.Vector.toList (Counters.wordEquiv h i)
  simpa only [List.Vector.toList_ofFn] using hv

/-- All stored constraint rows occur in this key, with integer arithmetic. -/
def integerKey {d h : ℕ} (A : Fin d → Fin h → ℤ) (i : Fin (2 ^ h)) : Fin d → ℤ :=
  OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutSignatureBound.boolSignature A (assignmentEquiv h i)

theorem card_integer_representative {d h : ℕ} (A : Fin d → Fin h → ℤ) :
    Nat.card (Representative (integerKey A)) =
      Nat.card (Set.range (OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutSignatureBound.boolSignature A)) := by
  rw [card_representative]
  have hr : Set.range (integerKey A) =
      Set.range (OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutSignatureBound.boolSignature A) :=
    (assignmentEquiv h).surjective.range_comp _
  rw [hr]

theorem card_integer_representative_le_rank {d h : ℕ} (A : Fin d → Fin h → ℤ)
    (W : ℕ) (hA : ∀ j i, |A j i| ≤ (W : ℤ)) :
    Nat.card (Representative (integerKey A)) ≤
      (2 * h * W + 1) ^ OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutRank.rowRank A := by
  rw [card_integer_representative]
  simpa using OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutSignatureBound.card_range_boolSignature_le_rank A W hA

theorem card_integer_representative_pos {d h : ℕ} (A : Fin d → Fin h → ℤ) :
    0 < Nat.card (Representative (integerKey A)) :=
  card_representative_pos _ (pow_pos (by omega) _)

theorem card_empty_head {d : ℕ} (A : Fin d → Fin 0 → ℤ) :
    Nat.card (Representative (integerKey A)) = 1 :=
  card_representative_one _

/-- The mutable duplicate latch has this proof-level recurrence. The actual
program implements it with a bounded word and one Boolean register. -/
def duplicateScan [DecidableEq κ] (key : ℕ → κ) (candidate : ℕ) : ℕ → Bool
  | 0 => false
  | bound + 1 => duplicateScan key candidate bound || decide (key bound = key candidate)

theorem duplicateScan_eq_true_iff [DecidableEq κ] (key : ℕ → κ) (candidate bound : ℕ) :
    duplicateScan key candidate bound = true ↔ ∃ j < bound, key j = key candidate := by
  induction bound with
  | zero => simp [duplicateScan]
  | succ bound ih =>
    simp only [duplicateScan, Bool.or_eq_true, decide_eq_true_eq, ih]
    constructor
    · rintro (⟨j, hj, he⟩ | he)
      · exact ⟨j, Nat.lt_succ_of_lt hj, he⟩
      · exact ⟨bound, Nat.lt_succ_self _, he⟩
    · rintro ⟨j, hj, he⟩
      rcases Nat.lt_or_eq_of_le (Nat.le_of_lt_succ hj) with hj | rfl
      · exact Or.inl ⟨j, hj, he⟩
      · exact Or.inr he

theorem duplicateScan_eq_false_iff [DecidableEq κ] (key : ℕ → κ) (candidate bound : ℕ) :
    duplicateScan key candidate bound = false ↔ ∀ j < bound, key j ≠ key candidate := by
  rw [← Bool.not_eq_true, duplicateScan_eq_true_iff]
  simp

theorem rescan_emits_iff [DecidableEq κ] (key : ℕ → κ) (i : Fin N) :
    duplicateScan key i i = false ↔ ∀ j : Fin N, j < i → key j ≠ key i := by
  rw [duplicateScan_eq_false_iff]
  constructor
  · intro h j hj
    exact h j hj
  · intro h j hj
    exact h ⟨j, hj.trans i.isLt⟩ hj

/-- Both candidate and predecessor scans are charged even when a key loses.
This is a bound on scan counts, not an implementation cost theorem. -/
theorem rescan_comparisons_le (N : ℕ) : (∑ i : Fin N, i.val) ≤ N * N := by
  calc
    _ ≤ ∑ _i : Fin N, N := Finset.sum_le_sum (fun i _ => Nat.le_of_lt i.isLt)
    _ = N * N := by simp

theorem rescan_boolean_comparisons_le (h : ℕ) :
    (∑ i : Fin (2 ^ h), i.val) ≤ 4 ^ h := by
  have he : 2 ^ h * 2 ^ h = 4 ^ h := by rw [← mul_pow]; rfl
  simpa only [he] using rescan_comparisons_le (2 ^ h)

end OAI.Algorithms.SubexponentialSAT.SqrtLogSpace.AttainedSignatures