CAST(VARCHAR as JSON) should escape Unicode characters

[mbasmanova]

Bug description

Velox currently doesn't escape Unicode characters when casting to JSON.

presto> select U&'\+01F64F';
 _col0
-------
 🙏
(1 row)

presto> select cast(U&'\+01F64F' as json);
     _col0
----------------
 "\uD83D\uDE4F"
(1 row)

Velox:

testCastToJson<StringView>(VARCHAR(), {"\U0001F64F"}, {"\"\\ud83d\\ude4f\""});

at 0: expected "\ud83d\ude4f", but got "🙏"

Velox uses folly::json::escapeString to case string to json. This function allows to specify configuration options including

  // If true, non-ASCII utf8 characters would be encoded as \uXXXX:
  // - if the code point is in [U+0000..U+FFFF] => encode as a single \uXXXX
  // - if the code point is > U+FFFF => encode as 2 UTF-16 surrogate pairs.
  bool encode_non_ascii{false};

The only difference with Presto is that folly::json::escapeString uses lowercase hex digits, while Presto uses uppercase.

CC: @zacw7 @aditi-pandit @amitkdutta @kagamiori @kevinwilfong

Related: FasterXML/jackson-core#717

System information

n/a

Relevant logs

No response