view_obstruction_ray_dist_abs
For every integer vector v in ℤ^n with no zero coordinate, the view-obstruction distance d_n(|v|) equals d_n(v).
Statement
theorem view_obstruction_ray_dist_abs (n : ℕ) (v : Fin n → ℤ) (hnz : ∀ i, v i ≠ 0) : view_obstruction_ray_dist n (fun i => |v i|) = view_obstruction_ray_dist n v
- Formalizes
- arXiv:2610.37711 Lemma 2.3(a)
- Verified
- 08 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 3
- view_obstruction_ray_distThe view-obstruction squared distance d_n(v) for an integer direction vector v, defined as the infimum over t ∈ ℝ of the squared distance…
- int_to_euclideanThe canonical embedding of an integer vector z ∈ ℤ^n into the Euclidean space ℝ^n.
- view_obstruction_centerThe center point c = (1/2, ..., 1/2) of the shifted lattice in Euclidean space ℝ^n.
Read back from the Lean
For every natural number n, every integer vector v : Fin n → ℤ, and assuming every component of v is nonzero (∀ i, v i ≠ 0), the view_obstruction_ray_dist of the componentwise absolute value vector fun i => |v i| is equal to the view_obstruction_ray_dist of v.
Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.
The proof, explained
To prove that the view-obstruction ray distance is invariant under coordinatewise absolute values |v|, the distance is expressed as an infimum over scaling parameters t ∈ ℝ and lattice points z ∈ ℤ^n of the squared Euclidean distance to the shifted cube center. The central observation is that reflection around 1/2 maps z ↦ -z - 1, which is an involution on ℤ satisfying (t v_i - (1/2 + (-z_i - 1)))^2 = (t |v_i| - (1/2 + z_i))^2 whenever v_i < 0. Applying this reflection coordinatewise where v_i < 0 defines an involution (hence a bijection) on the integer lattice ℤ^n. Re-indexing the infimum over z ∈ ℤ^n by this bijection preserves the infimum and equates the squared coordinate sums term by term.
The coordinate-flip map is an involution on ℤ.
For a boolean flag neg ∈ {false, true}, the map sending z ∈ ℤ to -z - 1 if neg is true and to z otherwise is its own inverse, verified by splitting into the two boolean cases.
def view_obstruction_ray_dist_abs_flip_coord (neg : Bool) (z : ℤ) : ℤ := if neg then -z - 1 else z lemma view_obstruction_ray_dist_abs_flip_coord_involutive (neg : Bool) (z : ℤ) : view_obstruction_ray_dist_abs_flip_coord neg (view_obstruction_ray_dist_abs_flip_coord neg z) = z := by cases neg <;> simp [view_obstruction_ray_dist_abs_flip_coord]
Coordinate-wise reflection defines a bijection (equivalence) of the integer lattice ℤ^n.
Applying the coordinate flip independently to each coordinate according to a vector of boolean flags produces an equivalence on ℤ^n, since the map is an involution in each coordinate.
uses
view_obstruction_ray_dist_abs_flip_coord_involutivedef view_obstruction_ray_dist_abs_flip_equiv {n : ℕ} (neg : Fin n → Bool) : (Fin n → ℤ) ≃ (Fin n → ℤ) where toFun z i := view_obstruction_ray_dist_abs_flip_coord (neg i) (z i) invFun z i := view_obstruction_ray_dist_abs_flip_coord (neg i) (z i) left_inv z := by ext i; apply view_obstruction_ray_dist_abs_flip_coord_involutive right_inv z := by ext i; apply view_obstruction_ray_dist_abs_flip_coord_involutive
When v < 0, the squared coordinate distance for v using the reflected integer -z - 1 equals that for |v| using z.
When v < 0, we have |v| = -v. Unfolding the coordinate flip, t * v - (1/2 + (-z - 1)) equals -(t * (-v) - (1/2 + z)). Squaring both sides eliminates the negative sign, yielding the equality.
uses
decide_eq_true_iff,abs_of_neglemma view_obstruction_ray_dist_abs_coord_sq_eq (t : ℝ) (v z : ℤ) : (t * (v : ℝ) - (1/2 + (view_obstruction_ray_dist_abs_flip_coord (decide (v < 0)) z : ℝ)))^2 = (t * ((|v| : ℤ) : ℝ) - (1/2 + (z : ℝ)))^2 := by dsimp [view_obstruction_ray_dist_abs_flip_coord] split_ifs with h · rw [decide_eq_true_iff] at h have hv : ((|v| : ℤ) : ℝ) = - (v : ℝ) := by push_cast exact_mod_cast abs_of_neg h rw [hv] have : t * (v : ℝ) - (1/2 + ((-z - 1 : ℤ) : ℝ)) = - (t * - (v : ℝ) - (1/2 + (z : ℝ))) := by push_cast; ring rw [this] ring
When v ≥ 0, the squared coordinate distance for v using z equals that for |v| using z.
When v is not strictly negative, 0 ≤ v, so |v| = v and the coordinate reflection is the identity, making the two squared terms trivially identical.
uses
decide_eq_true_iff,not_lt.mp,abs_of_nonneg· rw [decide_eq_true_iff] at h have hv : ((|v| : ℤ) : ℝ) = (v : ℝ) := by have : 0 ≤ v := not_lt.mp h push_cast exact_mod_cast abs_of_nonneg this rw [hv]
The squared Euclidean distance from t • v to the shifted center c + z is the sum of coordinate squared differences.
By definition of the Euclidean norm on ℝ^n, the squared norm ‖t • v - (c + z)‖^2 expands into the sum over all coordinates i of (t v_i - (1/2 + z_i))^2.
uses
EuclideanSpace.real_norm_sq_eqlemma view_obstruction_ray_dist_abs_norm_sq_eq_sum (n : ℕ) (t : ℝ) (v z : Fin n → ℤ) : ‖t • int_to_euclidean n v - (view_obstruction_center n + int_to_euclidean n z)‖ ^ 2 = ∑ i, (t * (v i : ℝ) - (1/2 + (z i : ℝ))) ^ 2 := by rw [EuclideanSpace.real_norm_sq_eq] rfl
It suffices to show that for every scaling parameter t ∈ ℝ, the infima over z ∈ ℤ^n for |v| and v coincide.
Unfolding the definition of the view-obstruction ray distance as an infimum over t ∈ ℝ followed by an infimum over z ∈ ℤ^n, stripping the outer infimum reduces the goal to proving equality of the inner infima for an arbitrary fixed t.
/-- Lemma 2.3(a): View-obstruction distance is invariant under coordinatewise absolute values. -/ theorem view_obstruction_ray_dist_abs (n : ℕ) (v : Fin n → ℤ) (hnz : ∀ i, v i ≠ 0) : view_obstruction_ray_dist n (fun i => |v i|) = view_obstruction_ray_dist n v := by have _ := hnz dsimp [view_obstruction_ray_dist] congr 1 ext t
Re-indexing by the coordinate-flip bijection reduces the equality of infima over z ∈ ℤ^n to pointwise equality at each lattice point z.
Let e be the equivalence on ℤ^n that flips coordinate i whenever v_i < 0. Because e is a bijection, precomposing by e does not change the infimum over ℤ^n. Stripping the infimum reduces the goal to equality of the squared norms for |v| at z and for v at e(z).
uses
Equiv.iInf_complet e := view_obstruction_ray_dist_abs_flip_equiv (n := n) (fun i => decide (v i < 0)) conv_rhs => rw [← Equiv.iInf_comp e] congr 1 ext z
The squared Euclidean distance for |v| at z equals that for v at e(z).
Rewriting both squared Euclidean norms as coordinate sums using view_obstruction_ray_dist_abs_norm_sq_eq_sum, the goal reduces to coordinate-wise equality. At each index i, the terms match by view_obstruction_ray_dist_abs_coord_sq_eq.
uses
view_obstruction_ray_dist_abs_norm_sq_eq_sum,view_obstruction_ray_dist_abs_coord_sq_eqrw [view_obstruction_ray_dist_abs_norm_sq_eq_sum, view_obstruction_ray_dist_abs_norm_sq_eq_sum] congr 1 ext i dsimp [e, view_obstruction_ray_dist_abs_flip_equiv] exact (view_obstruction_ray_dist_abs_coord_sq_eq t (v i) (z i)).symm
view_obstruction_ray_dist_abs_flip_coord_involutive- The coordinate-flip map z ↦ -z - 1 is an involution on ℤ, meaning applying it twice yields z.
decide_eq_true_iff- A decidable proposition p evaluates to true under decide if and only if p holds.
abs_of_neg- If an element a of an ordered additive group satisfies a < 0, then its absolute value |a| equals -a.
not_lt.mp- In a linear order, if a is not strictly less than b, then b ≤ a.
abs_of_nonneg- If an element a of an ordered additive group satisfies 0 ≤ a, then its absolute value |a| equals a.
EuclideanSpace.real_norm_sq_eq- The squared Euclidean norm of a vector in Euclidean space is the sum of the squares of its coordinates.
Equiv.iInf_comp- Precomposing an indexed family with an equivalence (bijection) of the index set preserves the infimum.
view_obstruction_ray_dist_abs_norm_sq_eq_sum- The squared Euclidean distance ‖t v - (c + z)‖^2 equals the sum over coordinates i of (t v_i - (1/2 + z_i))^2.
view_obstruction_ray_dist_abs_coord_sq_eq- For any coordinate, (t v - (1/2 + flip(v < 0, z)))^2 = (t |v| - (1/2 + z))^2.
- Although the theorem statement assumes that no coordinate of v is zero (hnz : ∀ i, v i ≠ 0), this hypothesis is not actually used in the proof because the algebraic coordinate identity holds for all integers v, including zero (as retained by `have _ := hnz` on line 45).
Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
def view_obstruction_ray_dist_abs_flip_coord (neg : Bool) (z : ℤ) : ℤ := if neg then -z - 1 else zlemma view_obstruction_ray_dist_abs_flip_coord_involutive (neg : Bool) (z : ℤ) : view_obstruction_ray_dist_abs_flip_coord neg (view_obstruction_ray_dist_abs_flip_coord neg z) = z := by cases neg <;> simp [view_obstruction_ray_dist_abs_flip_coord]def view_obstruction_ray_dist_abs_flip_equiv {n : ℕ} (neg : Fin n → Bool) : (Fin n → ℤ) ≃ (Fin n → ℤ) where toFun z i := view_obstruction_ray_dist_abs_flip_coord (neg i) (z i) invFun z i := view_obstruction_ray_dist_abs_flip_coord (neg i) (z i) left_inv z := by ext i; apply view_obstruction_ray_dist_abs_flip_coord_involutive right_inv z := by ext i; apply view_obstruction_ray_dist_abs_flip_coord_involutivelemma view_obstruction_ray_dist_abs_coord_sq_eq (t : ℝ) (v z : ℤ) : (t * (v : ℝ) - (1/2 + (view_obstruction_ray_dist_abs_flip_coord (decide (v < 0)) z : ℝ)))^2 = (t * ((|v| : ℤ) : ℝ) - (1/2 + (z : ℝ)))^2 := by dsimp [view_obstruction_ray_dist_abs_flip_coord] split_ifs with h · rw [decide_eq_true_iff] at h have hv : ((|v| : ℤ) : ℝ) = - (v : ℝ) := by push_cast exact_mod_cast abs_of_neg h rw [hv] have : t * (v : ℝ) - (1/2 + ((-z - 1 : ℤ) : ℝ)) = - (t * - (v : ℝ) - (1/2 + (z : ℝ))) := by push_cast; ring rw [this] ring · rw [decide_eq_true_iff] at h have hv : ((|v| : ℤ) : ℝ) = (v : ℝ) := by have : 0 ≤ v := not_lt.mp h push_cast exact_mod_cast abs_of_nonneg this rw [hv]lemma view_obstruction_ray_dist_abs_norm_sq_eq_sum (n : ℕ) (t : ℝ) (v z : Fin n → ℤ) : ‖t • int_to_euclidean n v - (view_obstruction_center n + int_to_euclidean n z)‖ ^ 2 = ∑ i, (t * (v i : ℝ) - (1/2 + (z i : ℝ))) ^ 2 := by rw [EuclideanSpace.real_norm_sq_eq] rfl/-- Lemma 2.3(a): View-obstruction distance is invariant under coordinatewise absolute values. -/theorem view_obstruction_ray_dist_abs (n : ℕ) (v : Fin n → ℤ) (hnz : ∀ i, v i ≠ 0) : view_obstruction_ray_dist n (fun i => |v i|) = view_obstruction_ray_dist n v := by have _ := hnz dsimp [view_obstruction_ray_dist] congr 1 ext t let e := view_obstruction_ray_dist_abs_flip_equiv (n := n) (fun i => decide (v i < 0)) conv_rhs => rw [← Equiv.iInf_comp e] congr 1 ext z rw [view_obstruction_ray_dist_abs_norm_sq_eq_sum, view_obstruction_ray_dist_abs_norm_sq_eq_sum] congr 1 ext i dsimp [e, view_obstruction_ray_dist_abs_flip_equiv] exact (view_obstruction_ray_dist_abs_coord_sq_eq t (v i) (z i)).symm