Add WITH clause column schema support for BULK OPENROWSET

Add WithColumns field to BulkOpenRowset for schema specification in
OPENROWSET BULK statements. Supports COLLATE clause and column ordinal
or JSON path specifications.

Author: claude

ast/bulk_insert_statement.go

@@ -66,11 +66,12 @@ func (o *OrderBulkInsertOption) bulkInsertOption() {}
 
 // BulkOpenRowset represents an OPENROWSET (BULK ...) table reference.
 type BulkOpenRowset struct {
-	DataFiles []ScalarExpression `json:"DataFiles,omitempty"`
-	Options   []BulkInsertOption `json:"Options,omitempty"`
-	Columns   []*Identifier      `json:"Columns,omitempty"`
-	Alias     *Identifier        `json:"Alias,omitempty"`
-	ForPath   bool               `json:"ForPath"`
+	DataFiles   []ScalarExpression            `json:"DataFiles,omitempty"`
+	Options     []BulkInsertOption            `json:"Options,omitempty"`
+	WithColumns []*OpenRowsetColumnDefinition `json:"WithColumns,omitempty"`
+	Columns     []*Identifier                 `json:"Columns,omitempty"`
+	Alias       *Identifier                   `json:"Alias,omitempty"`
+	ForPath     bool                          `json:"ForPath"`
 }
 
 func (b *BulkOpenRowset) node()           {}

ast/openrowset.go

@@ -39,6 +39,8 @@ func (o *OpenRowsetTableReference) tableReference() {}
 
 // OpenRowsetColumnDefinition represents a column definition in WITH clause.
 type OpenRowsetColumnDefinition struct {
+	ColumnOrdinal    ScalarExpression  `json:"ColumnOrdinal,omitempty"`
 	ColumnIdentifier *Identifier       `json:"ColumnIdentifier,omitempty"`
 	DataType         DataTypeReference `json:"DataType,omitempty"`
+	Collation        *Identifier       `json:"Collation,omitempty"`
 }

parser/marshal.go

@@ -2389,6 +2389,13 @@ func tableReferenceToJSON(ref ast.TableReference) jsonNode {
 			}
 			node["Options"] = opts
 		}
+		if len(r.WithColumns) > 0 {
+			cols := make([]jsonNode, len(r.WithColumns))
+			for i, c := range r.WithColumns {
+				cols[i] = openRowsetColumnDefinitionToJSON(c)
+			}
+			node["WithColumns"] = cols
+		}
 		if len(r.Columns) > 0 {
 			cols := make([]jsonNode, len(r.Columns))
 			for i, c := range r.Columns {
@@ -17301,11 +17308,17 @@ func openRowsetColumnDefinitionToJSON(col *ast.OpenRowsetColumnDefinition) jsonN
 	node := jsonNode{
 		"$type": "OpenRowsetColumnDefinition",
 	}
+	if col.ColumnOrdinal != nil {
+		node["ColumnOrdinal"] = scalarExpressionToJSON(col.ColumnOrdinal)
+	}
 	if col.ColumnIdentifier != nil {
 		node["ColumnIdentifier"] = identifierToJSON(col.ColumnIdentifier)
 	}
 	if col.DataType != nil {
 		node["DataType"] = dataTypeReferenceToJSON(col.DataType)
 	}
+	if col.Collation != nil {
+		node["Collation"] = identifierToJSON(col.Collation)
+	}
 	return node
 }

parser/parse_dml.go

@@ -687,6 +687,60 @@ func (p *Parser) parseBulkOpenRowset() (*ast.BulkOpenRowset, error) {
 	}
 	p.nextToken()
 
+	// Parse optional WITH (column_definitions) - for schema specification
+	if p.curTok.Type == TokenWith {
+		p.nextToken() // consume WITH
+		if p.curTok.Type == TokenLParen {
+			p.nextToken() // consume (
+			for p.curTok.Type != TokenRParen && p.curTok.Type != TokenEOF {
+				colDef := &ast.OpenRowsetColumnDefinition{}
+				colDef.ColumnIdentifier = p.parseIdentifier()
+
+				// Parse data type
+				dataType, err := p.parseDataTypeReference()
+				if err != nil {
+					return nil, err
+				}
+				colDef.DataType = dataType
+
+				// Parse optional COLLATE
+				if strings.ToUpper(p.curTok.Literal) == "COLLATE" {
+					p.nextToken() // consume COLLATE
+					colDef.Collation = p.parseIdentifier()
+				}
+
+				// Parse optional column ordinal (integer) or JSON path (string)
+				if p.curTok.Type == TokenNumber {
+					colDef.ColumnOrdinal = &ast.IntegerLiteral{
+						LiteralType: "Integer",
+						Value:       p.curTok.Literal,
+					}
+					p.nextToken()
+				} else if p.curTok.Type == TokenString {
+					// JSON path specification like '$.stateName' or 'strict $.population'
+					colDef.ColumnOrdinal = &ast.StringLiteral{
+						LiteralType:   "String",
+						IsNational:    false,
+						IsLargeObject: false,
+						Value:         strings.Trim(p.curTok.Literal, "'"),
+					}
+					p.nextToken()
+				}
+
+				result.WithColumns = append(result.WithColumns, colDef)
+
+				if p.curTok.Type == TokenComma {
+					p.nextToken()
+				} else {
+					break
+				}
+			}
+			if p.curTok.Type == TokenRParen {
+				p.nextToken() // consume )
+			}
+		}
+	}
+
 	// Parse optional alias
 	if p.curTok.Type == TokenAs {
 		p.nextToken()