Enable corr, covar_pop and covar_samp (#147)

* Enable corr covar_pop and covar_samp in aggregate. * Adjust the accuracy to align with Spark.
oap-project · Mar 30, 2023 · 799c054 · 799c054
1 parent 462979e
commit 799c054
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 4 deletions.
diff --git a/velox/functions/prestosql/aggregates/CovarianceAggregates.cpp b/velox/functions/prestosql/aggregates/CovarianceAggregates.cpp
@@ -229,9 +229,8 @@ struct CorrResultAccessor {
   }
 
   static double result(const CorrAccumulator& accumulator) {
-    double stddevX = std::sqrt(accumulator.m2X());
-    double stddevY = std::sqrt(accumulator.m2Y());
-    return accumulator.c2() / stddevX / stddevY;
+    // Need to modify the calculation order to maintain the same accuracy as spark
+    return accumulator.c2() / std::sqrt(accumulator.m2X() * accumulator.m2Y());
   }
 };
 

diff --git a/velox/substrait/SubstraitToVeloxPlanValidator.cpp b/velox/substrait/SubstraitToVeloxPlanValidator.cpp
@@ -789,7 +789,10 @@ bool SubstraitToVeloxPlanValidator::validate(
       "var_samp",
       "var_pop",
       "bitwise_and_agg",
-      "bitwise_or_agg"};
+      "bitwise_or_agg",
+      "corr",
+      "covar_pop",
+      "covar_samp"};
   for (const auto& funcSpec : funcSpecs) {
     auto funcName = subParser_->getSubFunctionName(funcSpec);
     if (supportedFuncs.find(funcName) == supportedFuncs.end()) {