feat(isthmus): map REGEXP_EXTRACT to Substrait regexp_match_substring

isthmus/src/main/java/io/substrait/isthmus/expression/FunctionMappings.java

@@ -131,6 +131,11 @@ public class FunctionMappings {
               s(SqlLibraryOperators.STARTS_WITH, "starts_with"),
               s(SqlLibraryOperators.ENDS_WITH, "ends_with"),
               s(SqlLibraryOperators.CONTAINS_SUBSTR, "contains"),
+              // Two-argument REGEXP_EXTRACT(value, regexp) maps to the two-argument
+              // regexp_match_substring(input, pattern), which returns the substring matching the
+              // full pattern. Patterns containing a capture group are not handled specially here:
+              // the full match is returned rather than the captured group.
+              s(SqlLibraryOperators.REGEXP_EXTRACT, "regexp_match_substring"),
               s(SqlStdOperatorTable.POSITION, "strpos"),
               s(SqlLibraryOperators.LEFT, "left"),
               s(SqlLibraryOperators.RIGHT, "right"),

isthmus/src/test/java/io/substrait/isthmus/AutomaticDynamicFunctionMappingRoundtripTest.java

@@ -13,12 +13,12 @@
  * Roundtrip test for the AutomaticDynamicFunctionMappingConverterProvider feature.
  *
  * <p>This test verifies that: 1. Substrait plans using unmapped functions (like strftime or
- * regexp_match_substring from extensions) are successfully converted. 2. With
+ * regexp_count_substring from extensions) are successfully converted. 2. With
  * AutomaticDynamicFunctionMappingConverterProvider enabled, these unmapped functions are
  * dynamically mapped to Calcite operators. 3. The roundtrip conversion (Substrait → Calcite →
  * Substrait) is stable, including for SQL queries.
  *
- * <p>The test uses unmapped functions like strftime and regexp_match_substring that are defined in
+ * <p>The test uses unmapped functions like strftime and regexp_count_substring that are defined in
  * extension YAML but not in FunctionMappings.
  */
 class AutomaticDynamicFunctionMappingRoundtripTest extends PlanTestBase {
@@ -95,15 +95,15 @@ void testUnmappedStrftimeSqlRoundtrip() throws Exception {
    * Test roundtrip with SQL query using multiple unmapped functions.
    *
    * <p>This test verifies that SQL queries with multiple unmapped function calls (like
-   * regexp_match_substring) can be handled when AutomaticDynamicFunctionMappingConverterProvider is
+   * regexp_count_substring) can be handled when AutomaticDynamicFunctionMappingConverterProvider is
    * used. The operator table is populated with unmapped function signatures, allowing occurrences
    * of unmapped functions to be recognized during SQL parsing and conversion.
    */
   @Test
   void testMultipleUnmappedFunctionsSqlRoundtrip() throws Exception {
     String createStatements = "CREATE TABLE t (date_str VARCHAR, ts_str VARCHAR)";
     String query =
-        "SELECT regexp_match_substring(date_str, '^[0-9]{4}') AS parsed_date, regexp_match_substring(ts_str, '^[0-9]{4}') AS parsed_ts FROM t";
+        "SELECT regexp_count_substring(date_str, '[0-9]') AS date_digits, regexp_count_substring(ts_str, '[0-9]') AS ts_digits FROM t";
 
     // Perform roundtrip with multiple unmapped function calls
     assertSqlSubstraitRelRoundTripLoosePojoComparison(

isthmus/src/test/java/io/substrait/isthmus/StringFunctionTest.java

@@ -229,6 +229,13 @@ void testContainsWithLiteral(String left, String right) throws Exception {
     assertSqlRoundTrip(query);
   }
 
+  @ParameterizedTest
+  @ValueSource(strings = {"c16", "vc32", "vc"})
+  void testRegexpExtract(String column) throws Exception {
+    String query = String.format("SELECT REGEXP_EXTRACT(%s, '[0-9]+') FROM strings", column);
+    assertSqlRoundTrip(query);
+  }
+
   @ParameterizedTest
   @CsvSource({"c16, c16", "c16, vc32", "c16, vc", "vc32, vc32", "vc32, vc", "vc, vc"})
   void testPosition(String substring, String input) throws Exception {