Fix chr SparkSQL function (#278)

Author: izchen

velox/docs/functions/spark/string.rst

@@ -10,7 +10,9 @@ 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.
+    Returns a utf8 string of single ASCII character. The ASCII character has the binary
+    equivalent of ``n``. 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

@@ -38,6 +38,11 @@ struct AsciiFunction {
   }
 };
 
+/// chr function
+/// chr(n) -> string
+/// Returns a utf8 string of single ASCII character. The ASCII character has
+/// the binary equivalent of n. 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);
@@ -46,8 +51,15 @@ struct ChrFunction {
     if (ord < 0) {
       result.resize(0);
     } else {
-      result.resize(1);
-      *result.data() = ord;
+      ord = ord & 0xFF;
+      if (ord < 0x80) {
+        result.resize(1);
+        result.data()[0] = ord;
+      } else {
+        result.resize(2);
+        result.data()[0] = 0xC0 + (ord >> 6);
+        result.data()[1] = 0x80 + (ord & 0x3F);
+      }
     }
     return true;
   }

velox/functions/sparksql/tests/StringTest.cpp

@@ -162,11 +162,16 @@ 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);
 }