Actually no need for Vitter if you can sort

1 files changed, 2 insertions, 39 deletions

diff --git a/test/fuzz.bqn b/test/fuzz.bqn
index 999aeb79..1a3e95fc 100644
--- a/test/fuzz.bqn
+++ b/test/fuzz.bqn
@@ -53,45 +53,8 @@ RandRank ← 4 _randUnbounded
 
 Sigmoid ← (40≤|)◶⟨1(-÷+)˜⋆,×⟩
 
-# Generate 𝕨 positive (not zero) integers summing to 𝕩
-RandPart ← {k←𝕨 ⋄ n←𝕩
-  RU ← Range∘0  # Random float in [0,1]
-  # Vitter, "An Efficient Algorithm for Sequential Random Sampling"
-  vPrime ← @
-
-  # Add or reject each possibility individually
-  MethodA ← {k←𝕩
-    v ← RU@
-    1 { (𝕨×-⟜k⊸÷n-𝕩) (⊣𝕊1+⊢)⍟(v<⊣) 𝕩 } 0
-  }
-
-  # Compute number of skipped possibilities
-  MethodD ← {k←𝕩
-    RvP ← k√RU  # Generate new vPrime
-    kn ← k-1    # k for next iteration
-    qu1 ← n-kn
-    Sk ← {vP←𝕩
-      s ← ⌊ x ← { 𝕊∘RvP⍟(qu1⊸≤) n׬𝕩 } vP
-      y1 ← kn √ (n ÷ qu1) × RU@
-      vPrime ↩ y1 × (1-x÷n) × qu1 (⊣÷-) s
-      Cont ← {
-        y2 ← ×´ kn (((n-1)-↕∘⌊) (⊣÷-) ⌈) 𝕩
-        (n(⊣÷-)x)  <  y1 × kn√y2
-      }
-      Cont◶⟨{ vPrime↩@ ⋄ 𝕩 }, 𝕊 RvP⟩⍟(1<vPrime) s
-    }
-    Sk RvP⍟(@⊸≡) vPrime
-  }
-
-  Gen ← {𝕤
-    alphaInv ← 13  # Used to choose A or D
-    M ← {n≤alphaInv×𝕩}◶MethodD‿{vPrime↩@ ⋄ (M↩MethodA)𝕩}
-    seq ← { n-↩1+s←M𝕩 ⋄ s }¨ k-↕k-1
-    last ← ⌊n×RU⍟(@⊸≡)vPrime
-    1 + seq ∾ last
-  }
-  0⊸<◶↕‿Gen k
-}
+# 𝕨 positive (not zero) integers summing to 𝕩
+RandPart ← { ¬⟜» (𝕨-1) (∧∘Rand∾1-˜⊢) 𝕩¬𝕨 }
 
 # 𝕩 is maximum bound plus 1 for both functions
 ⟨RandBound,RandShape⟩ ← {