Fix chr SparkSQL function (#4909)

Summary:
Fix chr to make the function behavior consistent with SparkSQL.

> SparkSQL chr:
> (https://github.com/apache/spark/blob/87a5442f7ed96b11051d8a9333476d080054e5a0/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/stringExpressions.scala#L2253）
>
> ```scala
>   protected override def nullSafeEval(lon: Any): Any = {
>     val longVal = lon.asInstanceOf[Long]
>     if (longVal < 0) {
>       UTF8String.EMPTY_UTF8
>     } else if ((longVal & 0xFF) == 0) {
>       UTF8String.fromString(Character.MIN_VALUE.toString)
>     } else {
>       UTF8String.fromString((longVal & 0xFF).toChar.toString)
>     }
>   }
> ```

Pull Request resolved: #4909

Reviewed By: Yuhta

Differential Revision: D47724533

Pulled By: mbasmanova

fbshipit-source-id: a037bfaef3e6f32133c3b9e5152d96535c9c94dd

velox/docs/functions/spark/string.rst

@@ -11,6 +11,8 @@ Unless specified otherwise, all functions return NULL if at least one of the arg
 .. spark:function:: chr(n) -> varchar
 
     Returns the Unicode code point ``n`` as a single character string.
+    If ``n < 0``, the result is an empty string.
+    If ``n >= 256``, the result is equivalent to chr(``n % 256``).
 
 .. spark:function:: contains(left, right) -> boolean
 

velox/functions/sparksql/String.h

@@ -81,18 +81,29 @@ struct AsciiFunction {
   }
 };
 
+/// chr function
+/// chr(n) -> string
+/// Returns the Unicode code point ``n`` as a single character string.
+/// If ``n < 0``, the result is an empty string.
+/// If ``n >= 256``, the result is equivalent to chr(``n % 256``).
 template <typename T>
 struct ChrFunction {
   VELOX_DEFINE_FUNCTION_TYPES(T);
 
-  FOLLY_ALWAYS_INLINE bool call(out_type<Varchar>& result, int64_t ord) {
-    if (ord < 0) {
+  FOLLY_ALWAYS_INLINE void call(out_type<Varchar>& result, int64_t n) {
+    if (n < 0) {
       result.resize(0);
     } else {
-      result.resize(1);
-      *result.data() = ord;
+      n = n & 0xFF;
+      if (n < 0x80) {
+        result.resize(1);
+        result.data()[0] = n;
+      } else {
+        result.resize(2);
+        result.data()[0] = 0xC0 + (n >> 6);
+        result.data()[1] = 0x80 + (n & 0x3F);
+      }
     }
-    return true;
   }
 };
 

velox/functions/sparksql/tests/StringTest.cpp

@@ -201,11 +201,16 @@ TEST_F(StringTest, Ascii) {
 }
 
 TEST_F(StringTest, Chr) {
-  EXPECT_EQ(chr(0), std::string("\0", 1));
-  EXPECT_EQ(chr(32), " ");
   EXPECT_EQ(chr(-16), "");
-  EXPECT_EQ(chr(256), std::string("\0", 1));
-  EXPECT_EQ(chr(256 + 32), std::string(" ", 1));
+  EXPECT_EQ(chr(0), std::string("\0", 1));
+  EXPECT_EQ(chr(0x100), std::string("\0", 1));
+  EXPECT_EQ(chr(0x1100), std::string("\0", 1));
+  EXPECT_EQ(chr(0x20), "\x20");
+  EXPECT_EQ(chr(0x100 + 0x20), "\x20");
+  EXPECT_EQ(chr(0x80), "\xC2\x80");
+  EXPECT_EQ(chr(0x100 + 0x80), "\xC2\x80");
+  EXPECT_EQ(chr(0xFF), "\xC3\xBF");
+  EXPECT_EQ(chr(0x100 + 0xFF), "\xC3\xBF");
   EXPECT_EQ(chr(std::nullopt), std::nullopt);
 }