Estimator for Levenshtein distance.

The usual algorithm for computing Levenshtein distances provides successively better lower bounds for the Levenshtein distance as it runs, as proved in Mathlib/Data/List/EditDistance/Bounds.lean.

In this file we package that fact as an instance of

Estimator (Thunk.mk fun _ => levenshtein C xs ys) (LevenshteinEstimator C xs ys)

allowing us to use the Estimator framework for Levenshtein distances.

This is then used in the implementation of rewrite_search to avoid needing the entire edit distance calculation in unlikely search paths.

structure LevenshteinEstimator' {α : Type u_1} {β δ : Type} [AddCommMonoid δ] [LinearOrder δ] (C : Levenshtein.Cost α β δ) (xs : List α) (ys : List β) : Type

Data showing that the Levenshtein distance from xs to ys is bounded below by the minimum Levenshtein distance between some suffix of xs and a particular suffix of ys.

This bound is (non-strict) monotone as we take longer suffixes of ys.

This is an auxiliary definition for the later LevenshteinEstimator: this variant constructs a lower bound for the pair consisting of the Levenshtein distance from xs to ys, along with the length of ys.

• pre_rev : List β

  The prefix of ys that is not is not involved in the bound, in reverse order.

• suff : List β

  The suffix of ys, such that the distance from xs to ys is bounded below by the minimum distance from any suffix of xs to this suffix.

• split : self.pre_rev.reverse ++ self.suff = ys

  Witness that ys has been decomposed into a prefix and suffix.

• distances : { r : List δ // 0 < r.length }

  The distances from each suffix of xs to suff.

• distances_eq : suffixLevenshtein C xs self.suff = self.distances

  Witness that distances are correct.

• bound : δ × ℕ

  The current bound on the pair (distance from xs to ys, length of ys).

• bound_eq : (match self.pre_rev, ⋯ with | [], split => ((↑self.distances)[0], ys.length) | x, split => (List.minimum_of_length_pos ⋯, self.suff.length)) = self.bound

  Predicate describing the current bound.

instance instEstimatorDataProdNatMkMkLevenshteinLengthLevenshteinEstimator' {α : Type u_1} {β δ : Type} [AddCommMonoid δ] [LinearOrder δ] (C : Levenshtein.Cost α β δ) (xs : List α) (ys : List β) : EstimatorData { fn := fun (x : Unit) => (levenshtein C xs ys, ys.length) } (LevenshteinEstimator' C xs ys)

instance estimator' {α : Type u_1} {β δ : Type} [AddCommMonoid δ] [LinearOrder δ] [CanonicallyOrderedAdd δ] (C : Levenshtein.Cost α β δ) (xs : List α) (ys : List β) : Estimator { fn := fun (x : Unit) => (levenshtein C xs ys, ys.length) } (LevenshteinEstimator' C xs ys)

def LevenshteinEstimator {α : Type u_1} {β δ : Type} [AddCommMonoid δ] [LinearOrder δ] [CanonicallyOrderedAdd δ] (C : Levenshtein.Cost α β δ) (xs : List α) (ys : List β) : Type

An estimator for Levenshtein distances.

instance instEstimatorMkLevenshteinLevenshteinEstimatorOfWellFoundedGTSubtypeProdNatLe {α : Type u_1} {β δ : Type} [AddCommMonoid δ] [LinearOrder δ] [CanonicallyOrderedAdd δ] (C : Levenshtein.Cost α β δ) (xs : List α) (ys : List β) [∀ (a : δ × ℕ), WellFoundedGT { x : δ × ℕ // x ≤ a }] : Estimator { fn := fun (x : Unit) => levenshtein C xs ys } (LevenshteinEstimator C xs ys)

instance instBotLevenshteinEstimator {α : Type u_1} {β δ : Type} [AddCommMonoid δ] [LinearOrder δ] [CanonicallyOrderedAdd δ] (C : Levenshtein.Cost α β δ) (xs : List α) (ys : List β) : Bot (LevenshteinEstimator C xs ys)

The initial estimator for Levenshtein distances.