Fix charset annotation in plan strings via saffron.properties

The getDefaultCharset() override (introduced to fix non-ASCII PPL string
literals) caused Calcite to annotate all VARCHAR/CHAR types and literals
with CHARACTER SET "UTF-8" and _UTF-8 prefix in plan strings, breaking
dozens of unit tests that compare logical plan representations.

Root cause: Calcite suppresses the charset annotation and _charset prefix
only when the charset matches CalciteSystemProperty.DEFAULT_CHARSET (which
defaults to ISO-8859-1). Overriding getDefaultCharset() to UTF-8 set the
charset on all types, but the suppression checks still compared against
ISO-8859-1, making the annotations appear everywhere.

Fix: add saffron.properties to core/src/main/resources with:
  calcite.default.charset=UTF-8
  calcite.default.collation.name=UTF-8$en_US

Calcite reads this file at CalciteSystemProperty class-load time (before
any DEFAULT_CHARSET or SqlCollation.IMPLICIT static field is initialized),
shifting the entire "default charset" universe to UTF-8. Both suppression
checks now compare against UTF-8, so plan strings are identical to before
while non-ASCII string literals continue to work correctly.

The now-redundant getDefaultCharset() override is removed; the inherited
SqlTypeFactoryImpl path already returns UTF-8 via Util.getDefaultCharset()
which reads from CalciteSystemProperty.DEFAULT_CHARSET = "UTF-8".

Signed-off-by: Radhakrishnan Pachyappan <gingeekrishna@gmail.com>

Author: gingeekrishna

core/src/main/java/org/opensearch/sql/calcite/CalciteRexNodeVisitor.java

@@ -131,9 +131,8 @@ public RexNode visitLiteral(Literal node, CalcitePlanContext context) {
       case NULL:
         return rexBuilder.makeNullLiteral(typeFactory.createSqlType(SqlTypeName.NULL));
       case STRING:
-        // UTF-8 charset is applied globally by OpenSearchTypeFactory.createSqlType() for
-        // VARCHAR/CHAR, so non-ASCII characters (e.g. Chinese, Arabic) are handled correctly
-        // and literal types stay compatible with column types for string operations.
+        // saffron.properties sets calcite.default.charset=UTF-8 so non-ASCII characters
+        // (e.g. Chinese, Arabic) are accepted and literal types stay compatible with column types.
         if (value.toString().length() == 1) {
           // To align Spark/PostgreSQL, Char(1) is useful, such as cast('1' to boolean) should
           // return true

core/src/main/java/org/opensearch/sql/calcite/utils/OpenSearchTypeFactory.java

@@ -32,7 +32,6 @@
 
 import java.lang.reflect.Type;
 import java.nio.charset.Charset;
-import java.nio.charset.StandardCharsets;
 import java.util.ArrayList;
 import java.util.LinkedHashMap;
 import java.util.List;
@@ -96,11 +95,6 @@ public RelDataType createTypeWithNullability(RelDataType type, boolean nullable)
     return super.createTypeWithNullability(type, nullable);
   }
 
-  @Override
-  public Charset getDefaultCharset() {
-    return StandardCharsets.UTF_8;
-  }
-
   @Override
   public RelDataType createTypeWithCharsetAndCollation(
       RelDataType type, Charset charset, SqlCollation collation) {

core/src/main/resources/saffron.properties

@@ -0,0 +1,7 @@
+# Shift Calcite's default charset from ISO-8859-1 to UTF-8 so that:
+# 1. Non-ASCII PPL string literals (Chinese, Arabic, etc.) are accepted without error.
+# 2. Plan-string representations (used in unit-test assertions) are unchanged,
+#    because Calcite suppresses the CHARACTER SET annotation and _charset prefix
+#    whenever the charset matches this "default" value.
+calcite.default.charset=UTF-8
+calcite.default.collation.name=UTF-8$en_US