feat: v1.15.0 dialect gaps — Oracle, ClickHouse 100%, WINDOW, AST roundtrip

Taskfile.yml

@@ -79,7 +79,15 @@ tasks:
     desc: Run tests with race detection (CRITICAL for production)
     cmds:
       - echo "Running tests with race detection..."
-      - go test -race -timeout 60s ./...
+      - go test -race -timeout 60s $(go list ./... | grep -v /cbinding)
+
+  test:cbinding:
+    desc: Test C binding package (requires CGO)
+    env:
+      CGO_ENABLED: '1'
+    cmds:
+      - echo "Running cbinding tests with CGO enabled..."
+      - go test -race -timeout 60s ./pkg/cbinding/...
 
   test:pkg:
     desc: Run tests for a specific package (use PKG=./pkg/sql/parser)

pkg/sql/ast/ast.go

@@ -128,6 +128,7 @@ type CommonTableExpr struct {
 	Name         string
 	Columns      []string
 	Statement    Statement
+	ScalarExpr   Expression      // ClickHouse: WITH <expr> AS <name> (scalar CTE, no subquery)
 	Materialized *bool           // nil = default, true = MATERIALIZED, false = NOT MATERIALIZED
 	Pos          models.Location // Source position of the CTE name (1-based line and column)
 }
@@ -429,8 +430,9 @@ type SelectStatement struct {
 	From              []TableReference
 	TableName         string // Added for pool operations
 	Joins             []JoinClause
-	PrewhereClause    Expression    // ClickHouse PREWHERE clause (applied before WHERE, before reading data)
-	Sample            *SampleClause // ClickHouse SAMPLE clause (comes after FROM/FINAL, before PREWHERE)
+	ArrayJoin         *ArrayJoinClause // ClickHouse ARRAY JOIN / LEFT ARRAY JOIN clause
+	PrewhereClause    Expression       // ClickHouse PREWHERE clause (applied before WHERE, before reading data)
+	Sample            *SampleClause    // ClickHouse SAMPLE clause (comes after FROM/FINAL, before PREWHERE)
 	Where             Expression
 	GroupBy           []Expression
 	Having            Expression
@@ -2294,6 +2296,20 @@ func (c ConnectByClause) Children() []Node {
 // via TABLESAMPLE, but this implementation targets SAMPLE).
 // Value is stored as a raw string to preserve the original representation
 // (e.g., "0.1", "1000", "1/10").
+// ArrayJoinClause represents a ClickHouse ARRAY JOIN or LEFT ARRAY JOIN clause.
+// Syntax: [LEFT] ARRAY JOIN expr [AS alias], expr [AS alias], ...
+type ArrayJoinClause struct {
+	Left     bool               // true for LEFT ARRAY JOIN
+	Elements []ArrayJoinElement // One or more join elements
+	Pos      models.Location
+}
+
+// ArrayJoinElement is a single expression in an ARRAY JOIN clause with an optional alias.
+type ArrayJoinElement struct {
+	Expr  Expression
+	Alias string
+}
+
 type SampleClause struct {
 	// Value is the sampling size/ratio as a raw token string (e.g., "0.1", "1000", "1/10").
 	Value string

pkg/sql/ast/sql.go

@@ -562,6 +562,23 @@ func (s *SelectStatement) SQL() string {
 		sb.WriteString(joinSQL(&j))
 	}
 
+	if s.ArrayJoin != nil {
+		if s.ArrayJoin.Left {
+			sb.WriteString(" LEFT ARRAY JOIN ")
+		} else {
+			sb.WriteString(" ARRAY JOIN ")
+		}
+		elems := make([]string, len(s.ArrayJoin.Elements))
+		for i, e := range s.ArrayJoin.Elements {
+			elemStr := exprSQL(e.Expr)
+			if e.Alias != "" {
+				elemStr += " AS " + e.Alias
+			}
+			elems[i] = elemStr
+		}
+		sb.WriteString(strings.Join(elems, ", "))
+	}
+
 	if s.PrewhereClause != nil {
 		sb.WriteString(" PREWHERE ")
 		sb.WriteString(exprSQL(s.PrewhereClause))
@@ -1406,6 +1423,13 @@ func forSQL(f *ForClause) string {
 func cteSQL(cte *CommonTableExpr) string {
 	sb := getBuilder()
 	defer putBuilder(sb)
+	// ClickHouse scalar CTE: WITH <expr> AS <name>
+	if cte.ScalarExpr != nil {
+		sb.WriteString(exprSQL(cte.ScalarExpr))
+		sb.WriteString(" AS ")
+		sb.WriteString(cte.Name)
+		return sb.String()
+	}
 	sb.WriteString(cte.Name)
 	if len(cte.Columns) > 0 {
 		sb.WriteString(" (")
@@ -1740,6 +1764,83 @@ func (p *PeriodDefinition) SQL() string {
 	return b.String()
 }
 
+// SQL returns the SQL string for a PRAGMA statement (SQLite).
+func (p *PragmaStatement) SQL() string {
+	if p == nil {
+		return ""
+	}
+	sb := getBuilder()
+	defer putBuilder(sb)
+	sb.WriteString("PRAGMA ")
+	sb.WriteString(p.Name)
+	if p.Arg != "" {
+		sb.WriteString("(")
+		sb.WriteString(p.Arg)
+		sb.WriteString(")")
+	} else if p.Value != "" {
+		sb.WriteString(" = ")
+		sb.WriteString(p.Value)
+	}
+	return sb.String()
+}
+
+// SQL returns the SQL string for a SHOW statement (MySQL).
+func (s *ShowStatement) SQL() string {
+	if s == nil {
+		return ""
+	}
+	sb := getBuilder()
+	defer putBuilder(sb)
+	sb.WriteString("SHOW ")
+	sb.WriteString(s.ShowType)
+	if s.ObjectName != "" {
+		sb.WriteString(" ")
+		sb.WriteString(s.ObjectName)
+	}
+	if s.From != "" {
+		sb.WriteString(" FROM ")
+		sb.WriteString(s.From)
+	}
+	return sb.String()
+}
+
+// SQL returns the SQL string for a DESCRIBE statement (MySQL).
+func (d *DescribeStatement) SQL() string {
+	if d == nil {
+		return ""
+	}
+	return "DESCRIBE " + d.TableName
+}
+
+// SQL returns the SQL string for a REPLACE statement (MySQL).
+func (r *ReplaceStatement) SQL() string {
+	if r == nil {
+		return ""
+	}
+	sb := getBuilder()
+	defer putBuilder(sb)
+	sb.WriteString("REPLACE INTO ")
+	sb.WriteString(r.TableName)
+	if len(r.Columns) > 0 {
+		sb.WriteString(" (")
+		sb.WriteString(exprListSQL(r.Columns))
+		sb.WriteString(")")
+	}
+	if len(r.Values) > 0 {
+		sb.WriteString(" VALUES ")
+		rows := make([]string, len(r.Values))
+		for idx, row := range r.Values {
+			vals := make([]string, len(row))
+			for j, v := range row {
+				vals[j] = exprSQL(v)
+			}
+			rows[idx] = "(" + strings.Join(vals, ", ") + ")"
+		}
+		sb.WriteString(strings.Join(rows, ", "))
+	}
+	return sb.String()
+}
+
 // ToSQL returns the SQL string for a CONNECT BY clause (MariaDB 10.2+).
 func (c *ConnectByClause) ToSQL() string {
 	var b strings.Builder

pkg/sql/parser/cte.go

@@ -23,6 +23,7 @@ import (
 	goerrors "github.com/ajitpratap0/GoSQLX/pkg/errors"
 	"github.com/ajitpratap0/GoSQLX/pkg/models"
 	"github.com/ajitpratap0/GoSQLX/pkg/sql/ast"
+	"github.com/ajitpratap0/GoSQLX/pkg/sql/keywords"
 )
 
 // WITH summary(region, total) AS (SELECT region, SUM(amount) FROM sales GROUP BY region) SELECT * FROM summary
@@ -124,6 +125,31 @@ func (p *Parser) parseCommonTableExpr() (*ast.CommonTableExpr, error) {
 		)
 	}
 
+	// ClickHouse scalar CTE: WITH <expr> AS <name>, ...
+	// Detected when the token after WITH is not an identifier, or is an
+	// identifier not followed by AS/( (which would be a standard CTE).
+	if p.dialect == string(keywords.DialectClickHouse) && !p.isIdentifier() {
+		scalarExpr, err := p.parseExpression()
+		if err != nil {
+			return nil, err
+		}
+		if !p.isType(models.TokenTypeAs) {
+			return nil, p.expectedError("AS after scalar CTE expression")
+		}
+		p.advance() // Consume AS
+		if !p.isIdentifier() {
+			return nil, p.expectedError("name after AS in scalar CTE")
+		}
+		scalarName := p.currentToken.Token.Value
+		scalarPos := p.currentLocation()
+		p.advance()
+		return &ast.CommonTableExpr{
+			Name:       scalarName,
+			ScalarExpr: scalarExpr,
+			Pos:        scalarPos,
+		}, nil
+	}
+
 	// Parse CTE name (supports double-quoted identifiers)
 	if !p.isIdentifier() {
 		return nil, p.expectedError("CTE name")

pkg/sql/parser/ddl.go

@@ -199,6 +199,7 @@ func (p *Parser) parseCreateTable(temporary bool) (*ast.CreateTableStatement, er
 	}
 
 	// CREATE TABLE ... AS SELECT — no column list, just a query.
+	// ClickHouse also: CREATE TABLE t AS source_table ENGINE = ...
 	if p.isType(models.TokenTypeAs) {
 		p.advance() // AS
 		if p.isType(models.TokenTypeSelect) || p.isType(models.TokenTypeWith) {
@@ -210,6 +211,16 @@ func (p *Parser) parseCreateTable(temporary bool) (*ast.CreateTableStatement, er
 			_ = query // CTAS query not modeled on CreateTableStatement yet
 			return stmt, nil
 		}
+		// ClickHouse: CREATE TABLE t AS <source_table> ENGINE = ...
+		// The identifier is the source table; consume remaining clauses.
+		if p.dialect == string(keywords.DialectClickHouse) && p.isIdentifier() {
+			p.advance() // Consume source table name
+			// Consume ENGINE and trailing clauses
+			for !p.isType(models.TokenTypeEOF) && !p.isType(models.TokenTypeSemicolon) {
+				p.advance()
+			}
+			return stmt, nil
+		}
 		return nil, p.expectedError("SELECT after AS")
 	}
 

pkg/sql/parser/ddl_view.go

@@ -21,6 +21,7 @@ import (
 	goerrors "github.com/ajitpratap0/GoSQLX/pkg/errors"
 	"github.com/ajitpratap0/GoSQLX/pkg/models"
 	"github.com/ajitpratap0/GoSQLX/pkg/sql/ast"
+	"github.com/ajitpratap0/GoSQLX/pkg/sql/keywords"
 )
 
 // parseCreateView parses CREATE [OR REPLACE] [TEMPORARY] VIEW statement
@@ -187,6 +188,26 @@ func (p *Parser) parseCreateMaterializedView() (*ast.CreateMaterializedViewState
 		p.advance()
 	}
 
+	// ClickHouse: optional TO <table> before ENGINE/AS
+	if p.dialect == string(keywords.DialectClickHouse) && p.isType(models.TokenTypeTo) {
+		p.advance() // Consume TO
+		toName, toErr := p.parseQualifiedName()
+		if toErr != nil {
+			return nil, p.expectedError("target table after TO")
+		}
+		stmt.Tablespace = toName // reuse Tablespace for ClickHouse TO
+	}
+
+	// ClickHouse: optional ENGINE = ... ORDER BY ... before AS SELECT
+	if p.dialect == string(keywords.DialectClickHouse) {
+		for p.isTokenMatch("ENGINE") || p.isType(models.TokenTypeOrder) || p.isTokenMatch("PRIMARY") || p.isTokenMatch("PARTITION") || p.isTokenMatch("SETTINGS") {
+			// Consume all engine clauses token-by-token until AS
+			for !p.isType(models.TokenTypeAs) && !p.isType(models.TokenTypeEOF) && !p.isType(models.TokenTypeSemicolon) {
+				p.advance()
+			}
+		}
+	}
+
 	// Expect AS
 	if !p.isType(models.TokenTypeAs) {
 		return nil, p.expectedError("AS")

pkg/sql/parser/dml_insert.go

@@ -68,9 +68,11 @@ func (p *Parser) parseInsertStatement() (ast.Statement, error) {
 		p.advance() // Consume )
 	}
 
-	// Parse SQL Server OUTPUT clause (between column list and VALUES)
+	// Parse SQL Server OUTPUT clause (between column list and VALUES).
+	// Accept OUTPUT regardless of dialect — the keyword is unambiguous here
+	// and allows dialect-agnostic parsing of T-SQL INSERT statements.
 	var outputCols []ast.Expression
-	if p.dialect == string(keywords.DialectSQLServer) && strings.ToUpper(p.currentToken.Token.Value) == "OUTPUT" {
+	if strings.ToUpper(p.currentToken.Token.Value) == "OUTPUT" {
 		p.advance() // Consume OUTPUT
 		var err error
 		outputCols, err = p.parseOutputColumns()

pkg/sql/parser/expressions_literal.go

@@ -128,6 +128,29 @@ func (p *Parser) parsePrimaryExpression() (ast.Expression, error) {
 		return p.parseExtractExpression()
 	}
 
+	// Oracle/MariaDB pseudo-columns: ROWNUM, ROWID, LEVEL, SYSDATE, SYSTIMESTAMP.
+	// These are tokenized as keywords but act as column-like expressions.
+	// We return them as zero-argument FunctionCall nodes so that implicit
+	// aliasing works naturally (SELECT ROWNUM rn → AliasedExpression) and
+	// they don't collide with the bare-Identifier alias guard.
+	if p.isType(models.TokenTypeKeyword) && p.isOraclePseudoColumn() {
+		identPos := p.currentLocation()
+		identName := p.currentToken.Token.Value
+		p.advance()
+		// SYSDATE() / SYSTIMESTAMP() — some drivers allow parens
+		if p.isType(models.TokenTypeLParen) {
+			funcCall, err := p.parseFunctionCall(identName)
+			if err != nil {
+				return nil, err
+			}
+			if funcCall.Pos.IsZero() {
+				funcCall.Pos = identPos
+			}
+			return funcCall, nil
+		}
+		return &ast.FunctionCall{Name: identName, Pos: identPos}, nil
+	}
+
 	if p.isType(models.TokenTypeIdentifier) || p.isType(models.TokenTypeDoubleQuotedString) || ((p.dialect == string(keywords.DialectSQLServer) || p.dialect == string(keywords.DialectClickHouse)) && p.isNonReservedKeyword()) {
 		// Handle identifiers and function calls
 		// Double-quoted strings are treated as identifiers in SQL (e.g., "column_name")