Support date type in SubstraitToVeloxPlan and hash function (#96)

Author: PHILO-HE

velox/functions/sparksql/Hash.cpp

@@ -69,6 +69,7 @@ void applyWithType(
       CASE(VARBINARY, hash.hashBytes, StringView);
       CASE(REAL, hash.hashFloat, float);
       CASE(DOUBLE, hash.hashDouble, double);
+      CASE(DATE, hash.hashDate, int32_t);
 #undef CASE
       default:
         VELOX_NYI(
@@ -135,6 +136,10 @@ class Murmur3Hash final {
     return fmix(h1, input.size());
   }
 
+  uint32_t hashDate(Date input, uint32_t seed) {
+    return hashInt32(input.days(), seed);
+  }
+
  private:
   uint32_t mixK1(uint32_t k1) {
     k1 *= 0xcc9e2d51;
@@ -235,6 +240,10 @@ class XxHash64 final {
     return fmix(hash);
   }
 
+  uint32_t hashDate(Date input, uint32_t seed) {
+    return hashInt32(input.days(), seed);
+  }
+
  private:
   uint64_t fmix(uint64_t hash) {
     hash ^= hash >> 33;

velox/substrait/SubstraitToVeloxPlan.cpp

@@ -1615,6 +1615,13 @@ void SubstraitVeloxPlanConverter::setFilterMap(
       setColInfoMap<std::string>(
           functionName, colIdxVal, val, reverse, colInfoMap);
       break;
+    case TypeKind::DATE:
+      if (substraitLit) {
+        val = variant(Date(substraitLit.value().date()));
+      }
+      setColInfoMap<int>(
+          functionName, colIdxVal, val, reverse, colInfoMap);
+      break;
     default:
       VELOX_NYI(
           "Subfield filters creation not supported for input type '{}'",
@@ -1892,6 +1899,10 @@ connector::hive::SubfieldFilters SubstraitVeloxPlanConverter::mapToFilters(
         constructSubfieldFilters<TypeKind::VARCHAR, common::Filter>(
             colIdx, inputNameList[colIdx], colInfoMap[colIdx], filters);
         break;
+      case TypeKind::DATE:
+        constructSubfieldFilters<TypeKind::DATE, common::BigintRange>(
+            colIdx, inputNameList[colIdx], colInfoMap[colIdx], filters);
+        break;
       default:
         VELOX_NYI(
             "Subfield filters creation not supported for input type '{}'",

velox/substrait/SubstraitToVeloxPlanValidator.cpp

@@ -336,13 +336,8 @@ bool SubstraitToVeloxPlanValidator::validate(
     return false;
   }
 
-  // In velox, the supported hash type in projectNode is
-  // BOOLEAN, TINYINT, SMALLINT, INTEGER, BIGINT, VARCHAR, VARBINARY
-  // REAL, DOUBLE. Hash.cpp (L148 - L156)
   for (auto i = 0; i < types.size(); i++) {
     switch (types[i]->kind()) {
-      case TypeKind::DATE:
-        return false;
       case TypeKind::ARRAY:
         return false;
       default:;

velox/substrait/TypeUtils.h

@@ -60,6 +60,13 @@ struct RangeTraits<TypeKind::VARCHAR> {
   using NativeType = std::string;
 };
 
+template <>
+struct RangeTraits<TypeKind::DATE> {
+  using RangeType = common::BigintRange;
+  using MultiRangeType = common::BigintMultiRange;
+  using NativeType = int32_t;
+};
+
 #endif /* RANGETRAITS_H */
 
 } // namespace facebook::velox::substrait