Fix event mixing sliding for spin correlation

Author: skundu692

PWGLF/Tasks/Strangeness/lambdaspincorrderived.cxx

@@ -1869,9 +1869,10 @@ struct lambdaspincorrderived {
       }
     }
 
-    // Neighbor policy (continuous mixing)
-    constexpr int nN_pt = 1;  // ±1 pt-bin
-    constexpr int nN_eta = 1; // ±1 eta/y-bin (can make configurable later)
+    // Neighbor policy from configurables
+    const int nN_pt = std::max(0, cfgV5NeighborPt.value);
+    const int nN_eta = std::max(0, cfgV5NeighborEta.value);
+    const int nN_phi = std::max(0, cfgV5NeighborPhi.value);
 
     std::vector<int> ptBins, etaBins, phiBins;
     std::vector<MatchRef> matches;
@@ -1897,6 +1898,7 @@ struct lambdaspincorrderived {
           continue; // same-event ordering
         }
 
+        // no shared daughters (same-event)
         if (t1.protonIndex() == t2.protonIndex())
           continue;
         if (t1.pionIndex() == t2.pionIndex())
@@ -1928,7 +1930,7 @@ struct lambdaspincorrderived {
 
         collectNeighborBinsClamp(ptB, nPt, nN_pt, ptBins);
         collectNeighborBinsClamp(etaB, nEta, nN_eta, etaBins);
-        collectPhiBinsWithEdgeWrap(phiB, nPhi, phiBins);
+        collectNeighborBinsPhi(phiB, nPhi, nN_phi, phiBins);
 
         matches.clear();
 
@@ -1949,12 +1951,10 @@ struct lambdaspincorrderived {
                   continue;
                 }
 
-                // extra strict kinematic check (uses eta or rapidity based on userapidity)
                 if (!checkKinematics(t1, tX)) {
                   continue;
                 }
 
-                // safety (should be redundant because different event)
                 if (tX.globalIndex() == t1.globalIndex())
                   continue;
                 if (tX.globalIndex() == t2.globalIndex())
@@ -1970,7 +1970,6 @@ struct lambdaspincorrderived {
           continue;
         }
 
-        // dedupe
         std::sort(matches.begin(), matches.end(),
                   [](auto const& a, auto const& b) {
                     return std::tie(a.collisionIdx, a.rowIndex) < std::tie(b.collisionIdx, b.rowIndex);
@@ -1984,9 +1983,8 @@ struct lambdaspincorrderived {
           continue;
         }
 
-        // unbiased cap (cfgV5MaxMatches==1 => pick-one mode)
         if (cfgV5MaxMatches.value > 0 && (int)matches.size() > cfgV5MaxMatches.value) {
-          uint64_t seed = 0;
+          uint64_t seed = cfgMixSeed.value;
           seed ^= splitmix64((uint64_t)t1.globalIndex());
           seed ^= splitmix64((uint64_t)t2.globalIndex() + 0x1234567ULL);
           seed ^= splitmix64((uint64_t)curColIdx + 0x9abcULL);
@@ -2029,8 +2027,6 @@ struct lambdaspincorrderived {
 
   void processMCMEV5(EventCandidatesMC const& collisions, AllTrackCandidatesMC const& V0sMC)
   {
-    // Buffer binning: v0etaMixBuffer = max(|eta|) or max(|y|) if userapidity
-    //                etaMix         = step for that axis (and also your matching window inside checkKinematicsMC)
     MixBinner mb{
       ptMin.value, ptMax.value, ptMix.value,
       v0etaMixBuffer.value, etaMix.value,
@@ -2039,7 +2035,7 @@ struct lambdaspincorrderived {
     const int nCol = colBinning.getAllBinsCount();
     const int nStat = N_STATUS;
     const int nPt = mb.nPt();
-    const int nEta = mb.nEta(); // logical "nY" if userapidity=true
+    const int nEta = mb.nEta();
     const int nPhi = mb.nPhi();
 
     const size_t nKeys = static_cast<size_t>(nCol) * nStat * nPt * nEta * nPhi;
@@ -2098,8 +2094,9 @@ struct lambdaspincorrderived {
       }
     }
 
-    constexpr int nN_pt = 1;
-    constexpr int nN_eta = 1;
+    const int nN_pt = std::max(0, cfgV5NeighborPt.value);
+    const int nN_eta = std::max(0, cfgV5NeighborEta.value);
+    const int nN_phi = std::max(0, cfgV5NeighborPhi.value);
 
     std::vector<int> ptBins, etaBins, phiBins;
     std::vector<MatchRef> matches;
@@ -2125,7 +2122,6 @@ struct lambdaspincorrderived {
           continue;
         }
 
-        // no shared daughters
         if (mcacc::prIdx(t1) == mcacc::prIdx(t2))
           continue;
         if (mcacc::piIdx(t1) == mcacc::piIdx(t2))
@@ -2136,7 +2132,6 @@ struct lambdaspincorrderived {
           continue;
 
         const int status = mcacc::v0Status(t1);
-
         if (status < 0 || status >= nStat) {
           continue;
         }
@@ -2157,7 +2152,7 @@ struct lambdaspincorrderived {
 
         collectNeighborBinsClamp(ptB, nPt, nN_pt, ptBins);
         collectNeighborBinsClamp(etaB, nEta, nN_eta, etaBins);
-        collectPhiBinsWithEdgeWrap(phiB, nPhi, phiBins);
+        collectNeighborBinsPhi(phiB, nPhi, nN_phi, phiBins);
 
         matches.clear();
 
@@ -2168,7 +2163,7 @@ struct lambdaspincorrderived {
 
               for (auto const& bc : vec) {
                 if (bc.collisionIdx == curColIdx) {
-                  continue; // different event
+                  continue;
                 }
 
                 auto tX = V0sMC.iteratorAt(static_cast<uint64_t>(bc.rowIndex));
@@ -2179,7 +2174,6 @@ struct lambdaspincorrderived {
                   continue;
                 }
 
-                // safety (should be redundant due to different event)
                 if (tX.globalIndex() == t1.globalIndex())
                   continue;
                 if (tX.globalIndex() == t2.globalIndex())
@@ -2195,7 +2189,6 @@ struct lambdaspincorrderived {
           continue;
         }
 
-        // dedupe
         std::sort(matches.begin(), matches.end(),
                   [](auto const& a, auto const& b) {
                     return std::tie(a.collisionIdx, a.rowIndex) < std::tie(b.collisionIdx, b.rowIndex);
@@ -2209,9 +2202,8 @@ struct lambdaspincorrderived {
           continue;
         }
 
-        // unbiased cap (cfgV5MaxMatches==1 => pick-one mode)
         if (cfgV5MaxMatches.value > 0 && (int)matches.size() > cfgV5MaxMatches.value) {
-          uint64_t seed = 0;
+          uint64_t seed = cfgMixSeed.value;
           seed ^= splitmix64((uint64_t)t1.globalIndex());
           seed ^= splitmix64((uint64_t)t2.globalIndex() + 0x1234567ULL);
           seed ^= splitmix64((uint64_t)curColIdx + 0x9abcULL);