Port the code for supporting md5/sha1/sha2 from apache apache#10464 (apache#135)

* Port the code and resolve conflicts

* Remove excessively ported code

* Fix issue in casting float to string

* Support sha2

* Fix compile error

* Add a ut

* Follow the naming convention

Author: philo-he

cpp/src/gandiva/function_registry_common.h

@@ -228,6 +228,17 @@ typedef std::unordered_map<const FunctionSignature*, const NativeFunction*, KeyH
                  utf8(), kResultNullNever, ARROW_STRINGIFY(gdv_fn_sha1_##TYPE),   \
                  NativeFunction::kNeedsContext | NativeFunction::kCanReturnErrors)
 
+// HashSHA2 functions that :
+// - Needs to specify bits_length
+// - NULL handling is of type NULL_NEVER
+// - can return errors
+//
+// The function name includes the base name & input type name. gdv_fn_sha256_float64
+#define HASH_SHA2_NULL_NEVER(NAME, ALIASES, TYPE)                                          \
+  NativeFunction(#NAME, std::vector<std::string> ALIASES, DataTypeVector{TYPE(), int32()}, \
+                 utf8(), kResultNullNever, ARROW_STRINGIFY(gdv_fn_sha2_##TYPE##_int32),    \
+                 NativeFunction::kNeedsContext | NativeFunction::kCanReturnErrors)
+
 // HashSHA256 functions that :
 // - NULL handling is of type NULL_NEVER
 // - can return errors
@@ -238,6 +249,16 @@ typedef std::unordered_map<const FunctionSignature*, const NativeFunction*, KeyH
                  utf8(), kResultNullNever, ARROW_STRINGIFY(gdv_fn_sha256_##TYPE), \
                  NativeFunction::kNeedsContext | NativeFunction::kCanReturnErrors)
 
+// HashMD5 functions that :
+// - NULL handling is of type NULL_NEVER
+// - can return errors
+//
+// The function name includes the base name & input type name. gdv_fn_md5_float64
+#define HASH_MD5_NULL_NEVER(NAME, ALIASES, TYPE)                                          \
+  NativeFunction(#NAME, std::vector<std::string> ALIASES, DataTypeVector{TYPE()}, utf8(), \
+                 kResultNullNever, ARROW_STRINGIFY(gdv_fn_md5_##TYPE),                    \
+                 NativeFunction::kNeedsContext | NativeFunction::kCanReturnErrors)
+
 // Iterate the inner macro over all numeric types
 #define NUMERIC_TYPES(INNER, NAME, ALIASES)                                             \
   INNER(NAME, ALIASES, int8), INNER(NAME, ALIASES, int16), INNER(NAME, ALIASES, int32), \

cpp/src/gandiva/function_registry_hash.cc

@@ -35,9 +35,15 @@ namespace gandiva {
 #define HASH_SHA1_NULL_NEVER_FN(name, ALIASES) \
   NUMERIC_BOOL_DATE_VAR_LEN_TYPES(HASH_SHA1_NULL_NEVER, name, ALIASES)
 
+#define HASH_SHA2_NULL_NEVER_FN(name, ALIASES) \
+  NUMERIC_BOOL_DATE_VAR_LEN_TYPES(HASH_SHA2_NULL_NEVER, name, ALIASES)
+
 #define HASH_SHA256_NULL_NEVER_FN(name, ALIASES) \
   NUMERIC_BOOL_DATE_VAR_LEN_TYPES(HASH_SHA256_NULL_NEVER, name, ALIASES)
 
+#define HASH_MD5_NULL_NEVER_FN(name, ALIASES) \
+  NUMERIC_BOOL_DATE_VAR_LEN_TYPES(HASH_MD5_NULL_NEVER, name, ALIASES)
+
 std::vector<NativeFunction> GetHashFunctionRegistry() {
   static std::vector<NativeFunction> hash_fn_registry_ = {
       NativeFunction("hash32_spark", {}, DataTypeVector{boolean(), int32()}, int32(),
@@ -72,9 +78,13 @@ std::vector<NativeFunction> GetHashFunctionRegistry() {
       HASH64_SEED_SAFE_NULL_NEVER_FN(hash64, {}),
       HASH64_SEED_SAFE_NULL_NEVER_FN(hash64AsDouble, {}),
 
-      HASH_SHA1_NULL_NEVER_FN(hashSHA1, {}),
+      HASH_SHA1_NULL_NEVER_FN(hashSHA1, {"sha1"}),
+
+      HASH_SHA2_NULL_NEVER_FN(hashSHA2, {"sha2"}),
+
+      HASH_SHA256_NULL_NEVER_FN(hashSHA256, {"sha256"}),
 
-      HASH_SHA256_NULL_NEVER_FN(hashSHA256, {})};
+      HASH_MD5_NULL_NEVER_FN(hashMD5, {"md5"})};
 
   return hash_fn_registry_;
 }

cpp/src/gandiva/function_registry_string.cc

@@ -112,21 +112,25 @@ std::vector<NativeFunction> GetStringFunctionRegistry() {
                      kResultNullIfNull, "castVARCHAR_int16_int64",
                      NativeFunction::kNeedsContext),
 
-      NativeFunction("castVARCHAR", {}, DataTypeVector{int32(), int64()}, utf8(),
-                     kResultNullIfNull, "castVARCHAR_int32_int64",
-                     NativeFunction::kNeedsContext),
-
-      NativeFunction("castVARCHAR", {}, DataTypeVector{int64(), int64()}, utf8(),
-                     kResultNullIfNull, "castVARCHAR_int64_int64",
-                     NativeFunction::kNeedsContext),
-
-      NativeFunction("castVARCHAR", {}, DataTypeVector{float32(), int64()}, utf8(),
-                     kResultNullIfNull, "castVARCHAR_float32_int64",
-                     NativeFunction::kNeedsContext),
-
-      NativeFunction("castVARCHAR", {}, DataTypeVector{float64(), int64()}, utf8(),
-                     kResultNullIfNull, "castVARCHAR_float64_int64",
-                     NativeFunction::kNeedsContext),
+      // There are other implementations that can replace the below functions, e.g.,
+      // gdv_fn_castVARCHAR_int32_int64, etc., which can get the same result as spark
+      // in handling cases like cast(2.0 as string) without extra zeros appended after
+      // the casting.
+      // NativeFunction("castVARCHAR", {}, DataTypeVector{int32(), int64()}, utf8(),
+      //                kResultNullIfNull, "castVARCHAR_int32_int64",
+      //                NativeFunction::kNeedsContext),
+
+      // NativeFunction("castVARCHAR", {}, DataTypeVector{int64(), int64()}, utf8(),
+      //                kResultNullIfNull, "castVARCHAR_int64_int64",
+      //                NativeFunction::kNeedsContext),
+
+      // NativeFunction("castVARCHAR", {}, DataTypeVector{float32(), int64()}, utf8(),
+      //                kResultNullIfNull, "castVARCHAR_float32_int64",
+      //                NativeFunction::kNeedsContext),
+
+      // NativeFunction("castVARCHAR", {}, DataTypeVector{float64(), int64()}, utf8(),
+      //                kResultNullIfNull, "castVARCHAR_float64_int64",
+      //                NativeFunction::kNeedsContext),
 
       NativeFunction("castVARCHAR", {}, DataTypeVector{boolean(), int64()}, utf8(),
                      kResultNullIfNull, "castVARCHAR_bool_int64",