Refactor AnalyzeJoinTest to align USING semantics with SQL standard

Co-authored-by: 936797922 <936797922@qq.com>

fe/fe-core/src/main/java/com/starrocks/sql/analyzer/QueryAnalyzer.java

@@ -906,6 +906,31 @@ public class QueryAnalyzer {
                 scope = new Scope(RelationId.of(join),
                         leftScope.getRelationFields().joinWith(rightScope.getRelationFields()));
             }
+            // When USING clause is present, SQL standard/MySQL expose only one column for each
+            // using-name in the output schema. Deduplicate here by keeping the first occurrence
+            // (left-first ordering in our construction) and dropping subsequent duplicates.
+            if (join.getUsingColNames() != null && !join.getUsingColNames().isEmpty()) {
+                Set<String> usingNamesLower = join.getUsingColNames().stream()
+                        .map(String::toLowerCase)
+                        .collect(Collectors.toSet());
+                Set<String> seen = new HashSet<>();
+                List<Field> deduplicated = new ArrayList<>();
+                for (Field field : scope.getRelationFields().getAllFields()) {
+                    String fieldName = field.getName();
+                    if (fieldName != null) {
+                        String lower = fieldName.toLowerCase();
+                        if (usingNamesLower.contains(lower)) {
+                            if (seen.contains(lower)) {
+                                // skip duplicate USING column from the right side
+                                continue;
+                            }
+                            seen.add(lower);
+                        }
+                    }
+                    deduplicated.add(field);
+                }
+                scope = new Scope(scope.getRelationId(), new RelationFields(deduplicated));
+            }
             join.setScope(scope);
 
             GeneratedColumnExprMappingCollector collector = new GeneratedColumnExprMappingCollector();

fe/fe-core/src/test/java/com/starrocks/sql/analyzer/AnalyzeJoinTest.java

@@ -91,7 +91,12 @@ public class AnalyzeJoinTest {
         analyzeFail("select * from (t0 join tnotnull using(v1)) t , t1",
                 "Getting syntax error at line 1, column 43. Detail message: Unexpected input 't', " +
                         "the most similar input is {<EOF>, ';'}.");
-        analyzeFail("select v1 from (t0 join tnotnull using(v1)), t1", "Column 'v1' is ambiguous");
+        // Using columns should be coalesced and appear only once in output names
+        QueryRelation query1 = ((QueryStatement) analyzeSuccess(
+                "select * from t0 a join t0 b using(v1)")).getQueryRelation();
+        Assertions.assertEquals("v1,v2,v3,v2,v3", String.join(",", query1.getColumnOutputNames()));
+        // Unqualified reference of using column should be unambiguous
+        analyzeSuccess("select v1 from (t0 join tnotnull using(v1)), t1");
         analyzeSuccess("select a.v1 from (t0 a join tnotnull b using(v1)), t1");
     }