Merge pull request #9 from jingchen2222/feat/support_last_join

feat: support last join

zetasql/parser/bison_parser.y

@@ -703,6 +703,7 @@ using zetasql::ASTDropStatement;
 %token KW_INTO "INTO"
 %token KW_IS "IS"
 %token KW_JOIN "JOIN"
+%token KW_LAST "LAST"
 %token KW_LEFT "LEFT"
 %token KW_LIKE "LIKE"
 %token KW_LIMIT "LIMIT"
@@ -842,7 +843,6 @@ using zetasql::ASTDropStatement;
 %token KW_JSON "JSON"
 %token KW_KEY "KEY"
 %token KW_LANGUAGE "LANGUAGE"
-%token KW_LAST "LAST"
 %token KW_LEAVE "LEAVE"
 %token KW_LEVEL "LEVEL"
 %token KW_LOOP "LOOP"
@@ -4657,7 +4657,20 @@ join:
         zetasql::parser::ErrorInfo error_info;
         auto node = zetasql::parser::JoinRuleAction(
             FirstNonEmptyLocation({@2, @3, @4, @5}), @$,
-            $1, $2, $3, $4, $6, $7, $8, parser, &error_info);
+            $1, $2, $3, $4, $6, $7, nullptr, $8, parser, &error_info);
+        if (node == nullptr) {
+          YYERROR_AND_ABORT_AT(error_info.location, error_info.message);
+        }
+
+        $$ = node;
+      }
+    | join_input opt_natural "LAST" join_hint "JOIN" opt_hint table_primary
+    opt_order_by_clause opt_on_or_using_clause_list
+      {
+        zetasql::parser::ErrorInfo error_info;
+        auto node = zetasql::parser::JoinRuleAction(
+            FirstNonEmptyLocation({@2, @3, @4, @5}), @$,
+            $1, $2, zetasql::ASTJoin::LAST, $4, $6, $7, $8, $9, parser, &error_info);
         if (node == nullptr) {
           YYERROR_AND_ABORT_AT(error_info.location, error_info.message);
         }
@@ -4715,15 +4728,33 @@ from_clause_contents:
         zetasql::parser::ErrorInfo error_info;
         auto node = zetasql::parser::JoinRuleAction(
             FirstNonEmptyLocation({@2, @3, @4, @5}), @$,
-            $1, $2, $3, $4, $6, $7, $8,
+            $1, $2, $3, $4, $6, $7, nullptr, $8,
             parser, &error_info);
         if (node == nullptr) {
           YYERROR_AND_ABORT_AT(error_info.location, error_info.message);
         }
 
         $$ = node;
       }
-    | "@"
+    | from_clause_contents opt_natural "LAST" join_hint "JOIN" opt_hint
+      table_primary opt_order_by_clause on_or_using_clause_list
+      {
+        // Give an error if we have a RIGHT or FULL JOIN following a comma
+        // join since our left-to-right binding would violate the standard.
+        // See (broken link).
+        zetasql::parser::ErrorInfo error_info;
+        auto node = zetasql::parser::JoinRuleAction(
+            FirstNonEmptyLocation({@2, @3, @4, @5}), @$,
+            $1, $2, zetasql::ASTJoin::LAST, $4, $6, $7, $8, $9,
+            parser, &error_info);
+        if (node == nullptr) {
+          YYERROR_AND_ABORT_AT(error_info.location, error_info.message);
+        }
+
+        $$ = node;
+      }
+    |
+    "@"
       {
         YYERROR_AND_ABORT_AT(
             @1, "Query parameters cannot be used in place of table names");
@@ -7036,6 +7067,7 @@ reserved_keyword_rule:
     | "INTO"
     | "IS"
     | "JOIN"
+    | "LAST"
     | "LEFT"
     | "LIKE"
     | "LIMIT"
@@ -7171,7 +7203,6 @@ keyword_as_identifier:
     | "JSON"
     | "KEY"
     | "LANGUAGE"
-    | "LAST"
     | "LEAVE"
     | "LEVEL"
     | "LOOP"

zetasql/parser/join_proccessor.cc

@@ -400,7 +400,8 @@ ASTNode* JoinRuleAction(
     const zetasql_bison_parser::location& start_location,
     const zetasql_bison_parser::location& end_location, ASTNode* lhs,
     bool natural, ASTJoin::JoinType join_type, ASTJoin::JoinHint join_hint,
-    ASTNode* hint, ASTNode* table_primary, ASTNode* on_or_using_clause_list,
+    ASTNode* hint, ASTNode* table_primary, ASTNode* order_by,
+    ASTNode* on_or_using_clause_list,
     BisonParser* parser,
     ErrorInfo* error_info) {
   auto clause_list =
@@ -440,11 +441,11 @@ ASTNode* JoinRuleAction(
     }
     join = parser->CreateASTNode<ASTJoin>(
         start_location, end_location,
-        {lhs, hint, table_primary, on_or_using_clause});
+        {lhs, hint, table_primary, order_by, on_or_using_clause});
     join->set_transformation_needed(IsTransformationNeeded(lhs));
   } else {
     join = parser->CreateASTNode<ASTJoin>(
-        start_location, end_location, {lhs, hint, table_primary, clause_list});
+        start_location, end_location, {lhs, hint, table_primary, order_by, clause_list});
     join->set_transformation_needed(true);
   }
 

zetasql/parser/join_proccessor.h

@@ -394,6 +394,7 @@ ASTNode* JoinRuleAction(
     const zetasql_bison_parser::location& end_location, ASTNode* lhs,
     bool opt_natural, ASTJoin::JoinType join_type, ASTJoin::JoinHint join_hint,
     ASTNode* opt_hint, ASTNode* table_primary,
+    ASTNode* opt_order_by,
     ASTNode* opt_on_or_using_clause_list,
     BisonParser* parser, ErrorInfo* error_info);
 

zetasql/parser/keywords.cc

@@ -173,7 +173,7 @@ constexpr KeywordInfoPOD kAllKeywords[] = {
     {"json", KW_JSON},
     {"key", KW_KEY},
     {"language", KW_LANGUAGE},
-    {"last", KW_LAST},
+    {"last", KW_LAST, KeywordInfo::kReserved},
     {"lateral", KW_LATERAL, KeywordInfo::kReserved},
     {"leave", KW_LEAVE},
     {"left", KW_LEFT, KeywordInfo::kReserved},

zetasql/parser/keywords_test.cc

@@ -237,7 +237,7 @@ TEST(ParserTest, DontAddNewReservedKeywords) {
   // allows new queries to work that will not work on older code.
   // Before changing this, co-ordinate with all engines to make sure the change
   // is done safely.
-  EXPECT_EQ(97 /* CAUTION */, num_reserved);
+  EXPECT_EQ(98 /* CAUTION */, num_reserved);
 }
 
 }  // namespace

zetasql/parser/parse_tree.cc

@@ -649,6 +649,8 @@ std::string ASTJoin::GetSQLForJoinType() const {
       return "LEFT";
     case RIGHT:
       return "RIGHT";
+    case LAST:
+      return "LAST";
   }
 }
 

zetasql/parser/parse_tree_manual.h

@@ -1286,7 +1286,7 @@ class ASTJoin final : public ASTTableExpression {
   std::string GetSQLForJoinType() const;
   std::string GetSQLForJoinHint() const;
 
-  enum JoinType { DEFAULT_JOIN_TYPE, COMMA, CROSS, FULL, INNER, LEFT, RIGHT };
+  enum JoinType { DEFAULT_JOIN_TYPE, COMMA, CROSS, FULL, INNER, LEFT, RIGHT, LAST};
 
   JoinType join_type() const { return join_type_; }
   void set_join_type(JoinType join_type) { join_type_ = join_type; }
@@ -1302,6 +1302,8 @@ class ASTJoin final : public ASTTableExpression {
   const ASTHint* hint() const { return hint_; }
   const ASTTableExpression* lhs() const { return lhs_; }
   const ASTTableExpression* rhs() const { return rhs_; }
+  // If present, order rhs before apply using clause
+  const ASTOrderBy* order_by() const { return order_by_; }
   const ASTOnClause* on_clause() const { return on_clause_; }
   const ASTUsingClause* using_clause() const { return using_clause_; }
 
@@ -1331,6 +1333,7 @@ class ASTJoin final : public ASTTableExpression {
     fl.AddRequired(&lhs_);
     fl.AddOptional(&hint_, AST_HINT);
     fl.AddRequired(&rhs_);
+    fl.AddOptional(&order_by_, AST_ORDER_BY);
     fl.AddOptional(&on_clause_, AST_ON_CLAUSE);
     fl.AddOptional(&using_clause_, AST_USING_CLAUSE);
 
@@ -1345,6 +1348,7 @@ class ASTJoin final : public ASTTableExpression {
   const ASTHint* hint_ = nullptr;
   const ASTTableExpression* lhs_ = nullptr;
   const ASTTableExpression* rhs_ = nullptr;
+  const ASTOrderBy* order_by_ = nullptr;
   const ASTOnClause* on_clause_ = nullptr;
   const ASTUsingClause* using_clause_ = nullptr;
   const ASTOnOrUsingClauseList* clause_list_ = nullptr;
@@ -1405,10 +1409,10 @@ class ASTOnClause final : public ASTNode {
     FieldLoader fl(this);
     fl.AddRequired(&expression_);
   }
-
   const ASTExpression* expression_ = nullptr;
 };
 
+
 class ASTUsingClause final : public ASTNode {
  public:
   static constexpr ASTNodeKind kConcreteNodeKind = AST_USING_CLAUSE;
@@ -1418,16 +1422,20 @@ class ASTUsingClause final : public ASTNode {
   zetasql_base::StatusOr<VisitResult> Accept(
       NonRecursiveParseTreeVisitor* visitor) const override;
 
+  // If present, order dataset before apply using clause
+  const ASTOrderBy* order_by() const { return order_by_; }
+
   const absl::Span<const ASTIdentifier* const>& keys() const {
     return keys_;
   }
 
  private:
   void InitFields() final {
     FieldLoader fl(this);
+    fl.AddRequired(&order_by_);
     fl.AddRestAsRepeated(&keys_);
   }
-
+  const ASTOrderBy* order_by_ = nullptr;
   absl::Span<const ASTIdentifier* const> keys_;
 };