structurize 2.16.2__py3-none-any.whl → 2.16.6__py3-none-any.whl

avrotize/structuretocsharp.py

@@ -1,2005 +1,2005 @@
-# pylint: disable=line-too-long
-
-""" StructureToCSharp class for converting JSON Structure schema to C# classes """
-
-import json
-import os
-import re
-from typing import Any, Dict, List, Tuple, Union, cast, Optional
-import uuid
-
-from avrotize.common import pascal, process_template
-import glob
-
-JsonNode = Dict[str, 'JsonNode'] | List['JsonNode'] | str | None
-
-
-INDENT = '    '
-
-
-class StructureToCSharp:
-    """ Converts JSON Structure schema to C# classes """
-
-    def __init__(self, base_namespace: str = '') -> None:
-        self.base_namespace = base_namespace
-        self.project_name: str = ''  # Optional explicit project name, separate from namespace
-        self.schema_doc: JsonNode = None
-        self.output_dir = os.getcwd()
-        self.pascal_properties = False
-        self.system_text_json_annotation = False
-        self.newtonsoft_json_annotation = False
-        self.system_xml_annotation = False
-        self.generated_types: Dict[str,str] = {}
-        self.generated_structure_types: Dict[str, Dict[str, Union[str, Dict, List]]] = {}
-        self.type_dict: Dict[str, Dict] = {}
-        self.definitions: Dict[str, Any] = {}
-        self.schema_registry: Dict[str, Dict] = {}  # Maps $id URIs to schemas
-        self.offers: Dict[str, Any] = {}  # Maps add-in names to property definitions from $offers
-
-    def get_qualified_name(self, namespace: str, name: str) -> str:
-        """ Concatenates namespace and name with a dot separator """
-        return f"{namespace}.{name}" if namespace != '' else name
-
-    def concat_namespace(self, namespace: str, name: str) -> str:
-        """ Concatenates namespace and name with a dot separator """
-        if namespace and name:
-            return f"{namespace}.{name}"
-        elif namespace:
-            return namespace
-        else:
-            return name
-
-    def map_primitive_to_csharp(self, structure_type: str) -> str:
-        """ Maps JSON Structure primitive types to C# types """
-        mapping = {
-            'null': 'void',  # Placeholder, actual handling for nullable types is in the union logic
-            'boolean': 'bool',
-            'string': 'string',
-            'integer': 'int',  # Generic integer type without format
-            'number': 'double',  # Generic number type without format
-            'int8': 'sbyte',
-            'uint8': 'byte',
-            'int16': 'short',
-            'uint16': 'ushort',
-            'int32': 'int',
-            'uint32': 'uint',
-            'int64': 'long',
-            'uint64': 'ulong',
-            'int128': 'System.Int128',
-            'uint128': 'System.UInt128',
-            'float8': 'float',  # Approximation - C# doesn't have native 8-bit float
-            'float': 'float',
-            'double': 'double',
-            'binary32': 'float',  # IEEE 754 binary32
-            'binary64': 'double',  # IEEE 754 binary64
-            'decimal': 'decimal',
-            'binary': 'byte[]',
-            'date': 'DateOnly',
-            'time': 'TimeOnly',
-            'datetime': 'DateTimeOffset',
-            'timestamp': 'DateTimeOffset',
-            'duration': 'TimeSpan',
-            'uuid': 'Guid',
-            'uri': 'Uri',
-            'jsonpointer': 'string',
-            'any': 'object'
-        }
-        qualified_class_name = 'global::'+self.get_qualified_name(pascal(self.base_namespace), pascal(structure_type))
-        if qualified_class_name in self.generated_structure_types:
-            result = qualified_class_name
-        else:
-            result = mapping.get(structure_type, 'object')
-        return result
-
-    def is_csharp_reserved_word(self, word: str) -> bool:
-        """ Checks if a word is a reserved C# keyword """
-        reserved_words = [
-            'abstract', 'as', 'base', 'bool', 'break', 'byte', 'case', 'catch', 'char', 'checked', 'class', 'const',
-            'continue', 'decimal', 'default', 'delegate', 'do', 'double', 'else', 'enum', 'event', 'explicit', 'extern',
-            'false', 'finally', 'fixed', 'float', 'for', 'foreach', 'goto', 'if', 'implicit', 'in', 'int', 'interface',
-            'internal', 'is', 'lock', 'long', 'namespace', 'new', 'null', 'object', 'operator', 'out', 'override',
-            'params', 'private', 'protected', 'public', 'readonly', 'ref', 'return', 'sbyte', 'sealed', 'short', 'sizeof',
-            'stackalloc', 'static', 'string', 'struct', 'switch', 'this', 'throw', 'true', 'try', 'typeof', 'uint', 'ulong',
-            'unchecked', 'unsafe', 'ushort', 'using', 'virtual', 'void', 'volatile', 'while'
-        ]
-        return word in reserved_words
-
-    def is_csharp_primitive_type(self, csharp_type: str) -> bool:
-        """ Checks if a type is a C# primitive type """
-        if csharp_type.endswith('?'):
-            csharp_type = csharp_type[:-1]
-        return csharp_type in ['void', 'bool', 'sbyte', 'byte', 'short', 'ushort', 'int', 'uint', 'long', 'ulong', 
-                               'float', 'double', 'decimal', 'string', 'DateTime', 'DateTimeOffset', 'DateOnly', 
-                               'TimeOnly', 'TimeSpan', 'Guid', 'byte[]', 'object', 'System.Int128', 'System.UInt128', 'Uri']
-
-    def map_csharp_primitive_to_clr_type(self, cs_type: str) -> str:
-        """ Maps C# primitive types to CLR types"""
-        map = {
-            "int": "Int32",
-            "long": "Int64",
-            "float": "Single",
-            "double": "Double",
-            "decimal": "Decimal",
-            "short": "Int16",
-            "sbyte": "SByte",
-            "byte": "Byte",
-            "ushort": "UInt16",
-            "uint": "UInt32",
-            "ulong": "UInt64",
-            "bool": "Boolean",
-            "string": "String",
-            "Guid": "Guid"
-        }
-        return map.get(cs_type, cs_type)
-
-    def resolve_ref(self, ref: str, context_schema: Optional[Dict] = None) -> Optional[Dict]:
-        """ Resolves a $ref to the actual schema definition """
-        # Check if it's an absolute URI reference (schema with $id)
-        if not ref.startswith('#/'):
-            # Try to resolve from schema registry
-            if ref in self.schema_registry:
-                return self.schema_registry[ref]
-            return None
-        
-        # Handle fragment-only references (internal to document)
-        path = ref[2:].split('/')
-        schema = context_schema if context_schema else self.schema_doc
-        
-        for part in path:
-            if not isinstance(schema, dict) or part not in schema:
-                return None
-            schema = schema[part]
-        
-        return schema
-    
-    def validate_abstract_ref(self, ref_schema: Dict, ref: str, is_extends_context: bool = False) -> None:
-        """
-        Validates that abstract types are only referenced in $extends context.
-        Per JSON Structure Core Spec Section 3.10.1, abstract types cannot be 
-        directly instantiated and should only be referenced via $extends.
-        
-        Args:
-            ref_schema: The resolved schema being referenced
-            ref: The $ref string for error reporting
-            is_extends_context: True if this reference is in a $extends context
-        """
-        import sys
-        is_abstract = ref_schema.get('abstract', False)
-        if is_abstract and not is_extends_context:
-            print(f"WARNING: Abstract type referenced outside $extends context: {ref}", file=sys.stderr)
-            print(f"  Abstract types cannot be directly instantiated. Use $extends to inherit from them.", file=sys.stderr)
-    
-    def register_schema_ids(self, schema: Dict, base_uri: str = '') -> None:
-        """ Recursively registers schemas with $id keywords """
-        if not isinstance(schema, dict):
-            return
-        
-        # Register this schema if it has an $id
-        if '$id' in schema:
-            schema_id = schema['$id']
-            # Handle relative URIs
-            if base_uri and not schema_id.startswith(('http://', 'https://', 'urn:')):
-                from urllib.parse import urljoin
-                schema_id = urljoin(base_uri, schema_id)
-            self.schema_registry[schema_id] = schema
-            base_uri = schema_id  # Update base URI for nested schemas
-        
-        # Recursively process definitions
-        if 'definitions' in schema:
-            for def_name, def_schema in schema['definitions'].items():
-                if isinstance(def_schema, dict):
-                    self.register_schema_ids(def_schema, base_uri)
-        
-        # Recursively process properties
-        if 'properties' in schema:
-            for prop_name, prop_schema in schema['properties'].items():
-                if isinstance(prop_schema, dict):
-                    self.register_schema_ids(prop_schema, base_uri)
-        
-        # Recursively process items, values, etc.
-        for key in ['items', 'values', 'additionalProperties']:
-            if key in schema and isinstance(schema[key], dict):
-                self.register_schema_ids(schema[key], base_uri)
-
-    def convert_structure_type_to_csharp(self, class_name: str, field_name: str, structure_type: JsonNode, parent_namespace: str) -> str:
-        """ Converts JSON Structure type to C# type """
-        if isinstance(structure_type, str):
-            return self.map_primitive_to_csharp(structure_type)
-        elif isinstance(structure_type, list):
-            # Handle type unions
-            non_null_types = [t for t in structure_type if t != 'null']
-            if len(non_null_types) == 1:
-                # Nullable type
-                return f"{self.convert_structure_type_to_csharp(class_name, field_name, non_null_types[0], parent_namespace)}?"
-            else:
-                return self.generate_embedded_union(class_name, field_name, non_null_types, parent_namespace, write_file=True)
-        elif isinstance(structure_type, dict):
-            # Handle $ref
-            if '$ref' in structure_type:
-                ref_schema = self.resolve_ref(structure_type['$ref'], self.schema_doc)
-                if ref_schema:
-                    # Validate abstract type usage (Section 3.10.1)
-                    # Abstract types should only be referenced via $extends
-                    self.validate_abstract_ref(ref_schema, structure_type['$ref'], is_extends_context=False)
-                    
-                    # Extract type name from the ref
-                    ref_path = structure_type['$ref'].split('/')
-                    type_name = ref_path[-1]
-                    ref_namespace = '.'.join(ref_path[2:-1]) if len(ref_path) > 3 else parent_namespace
-                    return self.generate_class_or_choice(ref_schema, ref_namespace, write_file=True, explicit_name=type_name)
-                return 'object'
-            
-            # Handle enum keyword - must be checked before 'type'
-            if 'enum' in structure_type:
-                return self.generate_enum(structure_type, field_name, parent_namespace, write_file=True)
-            
-            # Handle type keyword
-            if 'type' not in structure_type:
-                return 'object'
-            
-            struct_type = structure_type['type']
-            
-            # Handle complex types
-            if struct_type == 'object':
-                return self.generate_class(structure_type, parent_namespace, write_file=True)
-            elif struct_type == 'array':
-                items_type = self.convert_structure_type_to_csharp(class_name, field_name+'List', structure_type.get('items', {'type': 'any'}), parent_namespace)
-                return f"List<{items_type}>"
-            elif struct_type == 'set':
-                items_type = self.convert_structure_type_to_csharp(class_name, field_name+'Set', structure_type.get('items', {'type': 'any'}), parent_namespace)
-                return f"HashSet<{items_type}>"
-            elif struct_type == 'map':
-                values_type = self.convert_structure_type_to_csharp(class_name, field_name+'Map', structure_type.get('values', {'type': 'any'}), parent_namespace)
-                return f"Dictionary<string, {values_type}>"
-            elif struct_type == 'choice':
-                return self.generate_choice(structure_type, parent_namespace, write_file=True)
-            elif struct_type == 'tuple':
-                return self.generate_tuple(structure_type, parent_namespace, write_file=True)
-            else:
-                return self.convert_structure_type_to_csharp(class_name, field_name, struct_type, parent_namespace)
-        return 'object'
-
-    def generate_class_or_choice(self, structure_schema: Dict, parent_namespace: str, write_file: bool = True, explicit_name: str = '') -> str:
-        """ Generates a Class or Choice """
-        struct_type = structure_schema.get('type', 'object')
-        if struct_type == 'object':
-            return self.generate_class(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
-        elif struct_type == 'choice':
-            return self.generate_choice(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
-        elif struct_type == 'tuple':
-            return self.generate_tuple(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
-        elif struct_type in ('map', 'array', 'set'):
-            # Root-level container types: generate wrapper class with implicit conversions
-            return self.generate_container_wrapper(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
-        return 'object'
-
-    def generate_class(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
-        """ Generates a Class from JSON Structure object type """
-        class_definition = ''
-        
-        # Get name and namespace
-        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedClass'))
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        xml_namespace = structure_schema.get('xmlns', None)
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-
-        # Check if this is an abstract type (Section 3.10.1)
-        is_abstract = structure_schema.get('abstract', False)
-        
-        # Generate class documentation
-        doc = structure_schema.get('description', structure_schema.get('doc', class_name))
-        class_definition += f"/// <summary>\n/// {doc}\n/// </summary>\n"
-        
-        if is_abstract:
-            class_definition += f"/// <remarks>\n/// This is an abstract type and cannot be instantiated directly.\n/// </remarks>\n"
-
-        # Add Obsolete attribute if deprecated
-        if structure_schema.get('deprecated', False):
-            deprecated_msg = structure_schema.get('description', f'{class_name} is deprecated')
-            class_definition += f"[System.Obsolete(\"{deprecated_msg}\")]\n"
-
-        # Add XML serialization attribute for the class if enabled
-        if self.system_xml_annotation:
-            if xml_namespace:
-                class_definition += f"[XmlRoot(\"{class_name}\", Namespace=\"{xml_namespace}\")]\n"
-            else:
-                class_definition += f"[XmlRoot(\"{class_name}\")]\n"
-
-        # Generate properties
-        properties = structure_schema.get('properties', {})
-        required_props = structure_schema.get('required', [])
-        
-        # Handle alternative required sets
-        is_alternative_required = isinstance(required_props, list) and len(required_props) > 0 and isinstance(required_props[0], list)
-        
-        # Check additionalProperties setting (Section 3.7.8)
-        additional_props = structure_schema.get('additionalProperties', True if is_abstract else None)
-        
-        fields_str = []
-        for prop_name, prop_schema in properties.items():
-            field_def = self.generate_property(prop_name, prop_schema, class_name, schema_namespace, required_props)
-            fields_str.append(field_def)
-        
-        # Add dictionary for additional properties if needed
-        if additional_props is not False and additional_props is not None:
-            if isinstance(additional_props, dict):
-                # additionalProperties is a schema
-                value_type = self.convert_structure_type_to_csharp(class_name, 'additionalValue', additional_props, schema_namespace)
-                fields_str.append(f"{INDENT}/// <summary>\n{INDENT}/// Additional properties not defined in schema\n{INDENT}/// </summary>\n")
-                fields_str.append(f"{INDENT}public Dictionary<string, {value_type}>? AdditionalProperties {{ get; set; }}\n")
-            elif additional_props is True:
-                # Allow any additional properties
-                fields_str.append(f"{INDENT}/// <summary>\n{INDENT}/// Additional properties not defined in schema\n{INDENT}/// </summary>\n")
-                fields_str.append(f"{INDENT}public Dictionary<string, object>? AdditionalProperties {{ get; set; }}\n")
-        
-        class_body = "\n".join(fields_str)
-        
-        # Generate class declaration
-        abstract_modifier = "abstract " if is_abstract else ""
-        sealed_modifier = "sealed " if additional_props is False and not is_abstract else ""
-        
-        class_definition += f"public {abstract_modifier}{sealed_modifier}partial class {class_name}\n{{\n{class_body}"
-        
-        # Add default constructor (not for abstract classes with no concrete constructors)
-        if not is_abstract or properties:
-            class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Default constructor\n{INDENT}/// </summary>\n"
-            constructor_modifier = "protected" if is_abstract else "public"
-            class_definition += f"{INDENT}{constructor_modifier} {class_name}()\n{INDENT}{{\n{INDENT}}}"
-
-        # Add helper methods from template if any annotations are enabled
-        if self.system_text_json_annotation or self.newtonsoft_json_annotation or self.system_xml_annotation:
-            class_definition += process_template(
-                "structuretocsharp/dataclass_core.jinja",
-                class_name=class_name,
-                system_text_json_annotation=self.system_text_json_annotation,
-                newtonsoft_json_annotation=self.newtonsoft_json_annotation,
-                system_xml_annotation=self.system_xml_annotation
-            )
-
-        # Generate Equals and GetHashCode
-        class_definition += self.generate_equals_and_gethashcode(structure_schema, class_name, schema_namespace)
-
-        class_definition += "\n"+"}"
-
-        if write_file:
-            self.write_to_file(namespace, class_name, class_definition)
-
-        self.generated_types[ref] = "class"
-        self.generated_structure_types[ref] = structure_schema
-        return ref
-
-    def generate_property(self, prop_name: str, prop_schema: Dict, class_name: str, parent_namespace: str, required_props: List) -> str:
-        """ Generates a property for a class """
-        property_definition = ''
-        
-        # Resolve property name
-        field_name = prop_name
-        if self.is_csharp_reserved_word(field_name):
-            field_name = f"@{field_name}"
-        if self.pascal_properties:
-            field_name_cs = pascal(field_name)
-        else:
-            field_name_cs = field_name
-        if field_name_cs == class_name:
-            field_name_cs += "_"
-        
-        # Check if this is a const field
-        if 'const' in prop_schema:
-            const_value = prop_schema['const']
-            prop_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
-            
-            # Remove nullable marker for const
-            if prop_type.endswith('?'):
-                prop_type = prop_type[:-1]
-            
-            # Generate documentation
-            doc = prop_schema.get('description', prop_schema.get('doc', field_name_cs))
-            property_definition += f"{INDENT}/// <summary>\n{INDENT}/// {doc}\n{INDENT}/// </summary>\n"
-            
-            # Add JSON property name annotation
-            if self.system_text_json_annotation and field_name != field_name_cs:
-                property_definition += f'{INDENT}[System.Text.Json.Serialization.JsonPropertyName("{prop_name}")]\n'
-            if self.newtonsoft_json_annotation and field_name != field_name_cs:
-                property_definition += f'{INDENT}[Newtonsoft.Json.JsonProperty("{prop_name}")]\n'
-            
-            # Add XML element annotation if enabled
-            if self.system_xml_annotation:
-                property_definition += f'{INDENT}[System.Xml.Serialization.XmlElement("{prop_name}")]\n'
-            
-            # Generate const field
-            const_val = self.format_default_value(const_value, prop_type)
-            property_definition += f"{INDENT}public const {prop_type} {field_name_cs} = {const_val};\n"
-            
-            return property_definition
-        
-        # Determine if required
-        is_required = prop_name in required_props if not isinstance(required_props, list) or len(required_props) == 0 or not isinstance(required_props[0], list) else any(prop_name in req_set for req_set in required_props)
-        
-        # Get property type
-        prop_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
-        
-        # Add nullable marker if not required and not already nullable
-        if not is_required and not prop_type.endswith('?') and not prop_type.startswith('List<') and not prop_type.startswith('HashSet<') and not prop_type.startswith('Dictionary<'):
-            prop_type += '?'
-        
-        # Generate documentation
-        doc = prop_schema.get('description', prop_schema.get('doc', field_name_cs))
-        property_definition += f"{INDENT}/// <summary>\n{INDENT}/// {doc}\n{INDENT}/// </summary>\n"
-        
-        # Add JSON property name annotation
-        if self.system_text_json_annotation and field_name != field_name_cs:
-            property_definition += f'{INDENT}[System.Text.Json.Serialization.JsonPropertyName("{prop_name}")]\n'
-        if self.newtonsoft_json_annotation and field_name != field_name_cs:
-            property_definition += f'{INDENT}[Newtonsoft.Json.JsonProperty("{prop_name}")]\n'
-        
-        # Add XML element annotation if enabled
-        if self.system_xml_annotation:
-            property_definition += f'{INDENT}[System.Xml.Serialization.XmlElement("{prop_name}")]\n'
-        
-        # Add validation attributes based on schema constraints
-        # StringLength attribute for maxLength
-        if 'maxLength' in prop_schema:
-            max_length = prop_schema['maxLength']
-            if 'minLength' in prop_schema:
-                min_length = prop_schema['minLength']
-                property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.StringLength({max_length}, MinimumLength = {min_length})]\n'
-            else:
-                property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.StringLength({max_length})]\n'
-        elif 'minLength' in prop_schema:
-            # MinLength only (no max)
-            min_length = prop_schema['minLength']
-            property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.MinLength({min_length})]\n'
-        
-        # RegularExpression attribute for pattern
-        if 'pattern' in prop_schema:
-            pattern = prop_schema['pattern'].replace('\\', '\\\\').replace('"', '\\"')
-            property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.RegularExpression(@"{pattern}")]\n'
-        
-        # Range attribute for minimum/maximum on numeric types
-        if 'minimum' in prop_schema or 'maximum' in prop_schema:
-            min_val = prop_schema.get('minimum', prop_schema.get('exclusiveMinimum', 'double.MinValue'))
-            max_val = prop_schema.get('maximum', prop_schema.get('exclusiveMaximum', 'double.MaxValue'))
-            
-            # Convert to appropriate format
-            if isinstance(min_val, (int, float)):
-                min_str = str(min_val)
-            else:
-                min_str = str(min_val)
-            
-            if isinstance(max_val, (int, float)):
-                max_str = str(max_val)
-            else:
-                max_str = str(max_val)
-            
-            property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.Range({min_str}, {max_str})]\n'
-        
-        # Add Obsolete attribute if deprecated
-        if prop_schema.get('deprecated', False):
-            deprecated_msg = prop_schema.get('description', f'{prop_name} is deprecated')
-            property_definition += f'{INDENT}[System.Obsolete("{deprecated_msg}")]\n'
-        
-        # Generate property with required modifier if needed
-        required_modifier = "required " if is_required and not prop_type.endswith('?') else ""
-        
-        # Handle readOnly and writeOnly
-        is_read_only = prop_schema.get('readOnly', False)
-        is_write_only = prop_schema.get('writeOnly', False)
-        
-        if is_read_only:
-            # readOnly: private or init-only setter
-            property_definition += f"{INDENT}public {required_modifier}{prop_type} {field_name_cs} {{ get; init; }}"
-        elif is_write_only:
-            # writeOnly: private getter
-            property_definition += f"{INDENT}public {required_modifier}{prop_type} {field_name_cs} {{ private get; set; }}"
-        else:
-            # Normal property
-            property_definition += f"{INDENT}public {required_modifier}{prop_type} {field_name_cs} {{ get; set; }}"
-        
-        # Add default value if present
-        if 'default' in prop_schema:
-            default_val = self.format_default_value(prop_schema['default'], prop_type)
-            property_definition += f" = {default_val};\n"
-        else:
-            property_definition += "\n"
-        
-        return property_definition
-
-    def format_default_value(self, value: Any, csharp_type: str) -> str:
-        """ Formats a default value for C# """
-        if value is None:
-            return "null"
-        elif isinstance(value, bool):
-            return "true" if value else "false"
-        elif isinstance(value, str):
-            return f'"{value}"'
-        elif isinstance(value, (int, float)):
-            return str(value)
-        elif isinstance(value, list):
-            return f"new {csharp_type}()"
-        elif isinstance(value, dict):
-            return f"new {csharp_type}()"
-        return f"default({csharp_type})"
-
-    def generate_enum(self, structure_schema: Dict, field_name: str, parent_namespace: str, write_file: bool) -> str:
-        """ Generates a C# enum from JSON Structure enum keyword """
-        enum_values = structure_schema.get('enum', [])
-        if not enum_values:
-            return 'object'
-        
-        # Determine enum name from field name
-        enum_name = pascal(field_name) + 'Enum' if field_name else 'UnnamedEnum'
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, enum_name)
-        if ref in self.generated_types:
-            return ref
-        
-        # Determine underlying type
-        base_type = structure_schema.get('type', 'string')
-        
-        # For string enums, we don't specify an underlying type
-        # For numeric enums, we map the type
-        numeric_types = ['int8', 'uint8', 'int16', 'uint16', 'int32', 'uint32', 'int64', 'uint64']
-        is_numeric = base_type in numeric_types
-        
-        enum_definition = ''
-        doc = structure_schema.get('description', structure_schema.get('doc', enum_name))
-        enum_definition += f"/// <summary>\n/// {doc}\n/// </summary>\n"
-        
-        # Add Obsolete attribute if deprecated
-        if structure_schema.get('deprecated', False):
-            deprecated_msg = structure_schema.get('description', f'{enum_name} is deprecated')
-            enum_definition += f"[System.Obsolete(\"{deprecated_msg}\")]\n"
-        
-        if is_numeric:
-            cs_base_type = self.map_primitive_to_csharp(base_type)
-            enum_definition += f"public enum {enum_name} : {cs_base_type}\n{{\n"
-        else:
-            # String enum - for System.Text.Json, use JsonConverter with JsonStringEnumConverter
-            if self.system_text_json_annotation:
-                enum_definition += f"[System.Text.Json.Serialization.JsonConverter(typeof(System.Text.Json.Serialization.JsonStringEnumConverter))]\n"
-            enum_definition += f"public enum {enum_name}\n{{\n"
-        
-        # Generate enum members
-        for i, value in enumerate(enum_values):
-            if is_numeric:
-                # Numeric enum - use the value directly
-                member_name = f"Value{value}"  # Prefix with "Value" since enum members can't start with numbers
-                enum_definition += f"{INDENT}{member_name} = {value}"
-            else:
-                # String enum - create member from the string
-                member_name = pascal(str(value).replace('-', '_').replace(' ', '_'))
-                enum_definition += f"{INDENT}{member_name}"
-            
-            if i < len(enum_values) - 1:
-                enum_definition += ",\n"
-            else:
-                enum_definition += "\n"
-        
-        enum_definition += "}"
-        
-        if write_file:
-            self.write_to_file(namespace, enum_name, enum_definition)
-        
-        self.generated_types[ref] = "enum"
-        self.generated_structure_types[ref] = structure_schema
-        return ref
-
-    def generate_choice(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
-        """ Generates a discriminated union (choice) type """
-        # Choice types in JSON Structure can be:
-        # 1. Tagged unions - single property with the choice type as key
-        # 2. Inline unions - with $extends and selector
-        
-        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedChoice'))
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-        
-        choices = structure_schema.get('choices', {})
-        selector = structure_schema.get('selector')
-        extends = structure_schema.get('$extends')
-        
-        if extends and selector:
-            # Inline union - generate as inheritance hierarchy
-            return self.generate_inline_union(structure_schema, parent_namespace, write_file, explicit_name)
-        else:
-            # Tagged union - generate as a union class similar to Avro
-            return self.generate_tagged_union(structure_schema, parent_namespace, write_file, explicit_name)
-
-    def generate_tagged_union(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
-        """ Generates a tagged union type """
-        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedChoice'))
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-        
-        choices = structure_schema.get('choices', {})
-        choice_types = []
-        
-        for choice_name, choice_schema in choices.items():
-            choice_type = self.convert_structure_type_to_csharp(class_name, choice_name, choice_schema, schema_namespace)
-            choice_types.append((choice_name, choice_type))
-        
-        # Generate the union class similar to Avro unions
-        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name)}\n/// </summary>\n"
-        class_definition += f"public partial class {class_name}\n{{\n"
-        
-        # Generate properties for each choice
-        for choice_name, choice_type in choice_types:
-            prop_name = pascal(choice_name)
-            class_definition += f"{INDENT}/// <summary>\n{INDENT}/// Gets or sets the {prop_name} value\n{INDENT}/// </summary>\n"
-            class_definition += f"{INDENT}public {choice_type}? {prop_name} {{ get; set; }} = null;\n"
-        
-        # Add constructor
-        class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Default constructor\n{INDENT}/// </summary>\n"
-        class_definition += f"{INDENT}public {class_name}()\n{INDENT}{{\n{INDENT}}}\n"
-        
-        # Add constructors for each choice
-        for choice_name, choice_type in choice_types:
-            prop_name = pascal(choice_name)
-            class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Constructor for {prop_name} values\n{INDENT}/// </summary>\n"
-            class_definition += f"{INDENT}public {class_name}({choice_type} {prop_name.lower()})\n{INDENT}{{\n"
-            class_definition += f"{INDENT*2}this.{prop_name} = {prop_name.lower()};\n{INDENT}}}\n"
-        
-        # Generate Equals and GetHashCode for choice types
-        class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Determines whether the specified object is equal to the current object.\n{INDENT}/// </summary>\n"
-        class_definition += f"{INDENT}public override bool Equals(object? obj)\n{INDENT}{{\n"
-        class_definition += f"{INDENT*2}if (obj is not {class_name} other) return false;\n"
-        
-        # Compare each choice property
-        equality_checks = []
-        for choice_name, choice_type in choice_types:
-            prop_name = pascal(choice_name)
-            equality_checks.append(f"Equals(this.{prop_name}, other.{prop_name})")
-        
-        if len(equality_checks) == 1:
-            class_definition += f"{INDENT*2}return {equality_checks[0]};\n"
-        else:
-            class_definition += f"{INDENT*2}return " + f"\n{INDENT*3}&& ".join(equality_checks) + ";\n"
-        
-        class_definition += f"{INDENT}}}\n\n"
-        
-        # Generate GetHashCode
-        class_definition += f"{INDENT}/// <summary>\n{INDENT}/// Serves as the default hash function.\n{INDENT}/// </summary>\n"
-        class_definition += f"{INDENT}public override int GetHashCode()\n{INDENT}{{\n"
-        
-        if len(choice_types) <= 8:
-            hash_fields = [f"this.{pascal(choice_name)}" for choice_name, _ in choice_types]
-            class_definition += f"{INDENT*2}return HashCode.Combine({', '.join(hash_fields)});\n"
-        else:
-            class_definition += f"{INDENT*2}var hash = new HashCode();\n"
-            for choice_name, _ in choice_types:
-                prop_name = pascal(choice_name)
-                class_definition += f"{INDENT*2}hash.Add(this.{prop_name});\n"
-            class_definition += f"{INDENT*2}return hash.ToHashCode();\n"
-        
-        class_definition += f"{INDENT}}}\n"
-        
-        class_definition += "}"
-        
-        if write_file:
-            self.write_to_file(namespace, class_name, class_definition)
-        
-        self.generated_types[ref] = "choice"
-        self.generated_structure_types[ref] = structure_schema
-        return ref
-
-    def generate_inline_union(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
-        """ Generates an inline union type with inheritance """
-        # For inline unions, we generate an abstract base class and derived classes
-        # The selector property indicates which derived class is being used
-        
-        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedChoice'))
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-        
-        # Get base class from $extends
-        extends_ref = structure_schema.get('$extends', '')
-        if extends_ref and isinstance(extends_ref, str):
-            base_schema = self.resolve_ref(extends_ref, self.schema_doc)
-            if not base_schema:
-                # Try resolving relative to the structure_schema itself
-                base_schema = self.resolve_ref(extends_ref, structure_schema)
-            
-            # Validate abstract type usage - $extends is ALLOWED to reference abstract types
-            if base_schema:
-                self.validate_abstract_ref(base_schema, extends_ref, is_extends_context=True)
-        else:
-            base_schema = None
-        
-        if not base_schema:
-            # Fallback to tagged union if no base
-            return self.generate_tagged_union(structure_schema, parent_namespace, write_file, explicit_name)
-        
-        # First, ensure base class is generated (if it's abstract, it won't be referenced directly)
-        base_schema_copy = base_schema.copy()
-        if 'name' not in base_schema_copy:
-            # Extract name from $extends ref
-            base_name = extends_ref.split('/')[-1]
-            base_schema_copy['name'] = base_name
-        base_class_ref = self.generate_class(base_schema_copy, schema_namespace, write_file)
-        base_class_name = base_class_ref.split('::')[-1].split('.')[-1]
-        
-        choices = structure_schema.get('choices', {})
-        selector = structure_schema.get('selector', 'type')
-        
-        # Generate abstract base class with selector property
-        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name + ' (inline union base)')}\n/// </summary>\n"
-        
-        if self.system_text_json_annotation:
-            class_definition += f'[System.Text.Json.Serialization.JsonPolymorphic(TypeDiscriminatorPropertyName = "{selector}")]\n'
-            for choice_name in choices.keys():
-                derived_class_name = pascal(choice_name)
-                class_definition += f'[System.Text.Json.Serialization.JsonDerivedType(typeof({derived_class_name}), "{choice_name}")]\n'
-        
-        class_definition += f"public abstract partial class {class_name}"
-        
-        # Inherit from base class if it exists
-        if base_class_name and base_class_name != class_name:
-            class_definition += f" : {base_class_name}"
-        
-        class_definition += "\n{\n"
-        
-        # Add selector property (not required since derived classes set it in constructor)
-        class_definition += f"{INDENT}/// <summary>\n{INDENT}/// Type discriminator\n{INDENT}/// </summary>\n"
-        if self.system_text_json_annotation:
-            class_definition += f'{INDENT}[System.Text.Json.Serialization.JsonPropertyName("{selector}")]\n'
-        
-        # Check if selector is already in base properties
-        base_has_selector = selector in base_schema.get('properties', {})
-        if base_has_selector:
-            class_definition += f"{INDENT}public new string {pascal(selector)} {{ get; set; }} = \"\";\n"
-        else:
-            class_definition += f"{INDENT}public string {pascal(selector)} {{ get; set; }} = \"\";\n"
-        
-        class_definition += "}"
-        
-        if write_file:
-            self.write_to_file(namespace, class_name, class_definition)
-        
-        # Generate derived classes for each choice with property merging
-        for choice_name, choice_schema_ref in choices.items():
-            # Resolve the choice schema
-            if isinstance(choice_schema_ref, dict) and '$ref' in choice_schema_ref:
-                choice_schema = self.resolve_ref(choice_schema_ref['$ref'], self.schema_doc)
-                if not choice_schema:
-                    # Try resolving relative to the structure_schema itself
-                    choice_schema = self.resolve_ref(choice_schema_ref['$ref'], structure_schema)
-            else:
-                choice_schema = choice_schema_ref
-            
-            if not choice_schema or not isinstance(choice_schema, dict):
-                continue
-            
-            # Mark this choice as generated to prevent duplicate generation in process_definitions
-            derived_class_name = pascal(choice_name)
-            derived_namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-            derived_ref = 'global::'+self.get_qualified_name(derived_namespace, derived_class_name)
-            
-            # Only generate if not already generated
-            if derived_ref in self.generated_types:
-                continue
-            
-            # Merge properties from base schema into choice schema
-            merged_schema = self.merge_inherited_properties(choice_schema, base_schema, class_name)
-            merged_schema['name'] = choice_name
-            merged_schema['namespace'] = schema_namespace
-            
-            # Mark that this extends the union base
-            merged_schema['$extends_inline_union'] = class_name
-            
-            # Generate the derived class
-            self.generate_derived_class(merged_schema, class_name, choice_name, selector, schema_namespace, write_file)
-        
-        self.generated_types[ref] = "choice"
-        self.generated_structure_types[ref] = structure_schema
-        return ref
-    
-    def merge_inherited_properties(self, derived_schema: Dict, base_schema: Dict, union_class_name: str) -> Dict:
-        """ Merges properties from base schema into derived schema """
-        merged = derived_schema.copy()
-        
-        # Get properties from both schemas
-        base_props = base_schema.get('properties', {})
-        derived_props = merged.get('properties', {})
-        
-        # Track which properties come from base (for filtering during generation)
-        base_property_names = list(base_props.keys())
-        
-        # Merge properties (derived overrides base)
-        merged_props = {}
-        merged_props.update(base_props)
-        merged_props.update(derived_props)
-        merged['properties'] = merged_props
-        
-        # Store base property names so we can skip them during code generation
-        merged['$base_properties'] = base_property_names
-        
-        # Merge required fields
-        base_required = base_schema.get('required', [])
-        derived_required = merged.get('required', [])
-        if isinstance(base_required, list) and isinstance(derived_required, list):
-            # Combine and deduplicate
-            merged['required'] = list(set(base_required + derived_required))
-        
-        return merged
-    
-    def generate_derived_class(self, schema: Dict, base_class_name: str, choice_name: str, selector: str, parent_namespace: str, write_file: bool) -> str:
-        """ Generates a derived class for inline union """
-        class_name = pascal(choice_name)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema.get('namespace', parent_namespace)))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-        
-        # Generate class with inheritance
-        doc = schema.get('description', f'{class_name} - {choice_name} variant')
-        class_definition = f"/// <summary>\n/// {doc}\n/// </summary>\n"
-        
-        class_definition += f"public partial class {class_name} : {base_class_name}\n{{\n"
-        
-        # Generate properties (only the derived-specific ones, base properties are inherited)
-        properties = schema.get('properties', {})
-        required_props = schema.get('required', [])
-        
-        # Get base class properties to filter them out
-        # We need to find the base schema to know what properties to exclude
-        # For now, we'll generate all properties from merged schema
-        # NOTE: BaseAddress properties come through the merged schema but shouldn't be redeclared
-        #       We need to identify which properties are from the BASE vs which are NEW
-        
-        # Get the original (non-merged) choice schema to determine NEW properties
-        # The merged schema has all properties; we want only the ones NOT in base
-        # Since we don't have access to the original choice schema here, we'll use a heuristic:
-        # Properties marked with a special flag during merging
-        
-        # Alternative approach: Only generate properties that are NOT in the InlineChoice base
-        # But InlineChoice only has the selector, not the BaseAddress properties
-        # So we need to look further up the chain
-        
-        # SIMPLEST SOLUTION: Filter out properties that come from the extended base
-        # We can detect this by checking if the property exists in the base_schema context
-        # But we don't have base_schema in this method
-        
-        # FOR NOW: Generate all properties but mark inherited ones with 'new'
-        # Actually, C# doesn't allow 'new required' - that's the error
-        # So we MUST skip inherited properties entirely
-        
-        # The merged schema has ALL properties. We need to skip base properties.
-        # The base properties are those NOT in the original choice schema
-        # We need to pass the original choice schema properties to know what to generate
-        
-        # CORRECT FIX: Only generate properties from the ORIGINAL choice schema, not merged
-        # But wait - we're passing merged_schema which has all properties
-        # We need to differentiate
-        
-        # Let's add a marker during merge to track which properties are from base
-        # Or better: pass BOTH original and merged schemas
-        
-        # QUICK FIX: Check if selector property to skip, and skip properties that were
-        # in the base by checking schema metadata
-        
-        # Since schema has '$extends_inline_union', we can use that
-        # But we need the ORIGINAL choice properties, not merged
-        
-        # The issue is we're generating from merged_schema which has ALL properties
-        # We need to know which properties are NEW (from choice) vs inherited (from base)
-        
-        # SOLUTION: Don't generate inherited properties - but how to identify them?
-        # We could store in merged_schema a list of base property names
-        
-        # Let me fix this by adding a key to mark base properties
-        base_properties = schema.get('$base_properties', [])
-        
-        for prop_name, prop_schema in properties.items():
-            # Skip selector - it's defined in base as required
-            if prop_name == selector:
-                continue
-            # Skip properties inherited from base schema
-            if prop_name in base_properties:
-                continue
-            field_def = self.generate_property(prop_name, prop_schema, class_name, parent_namespace, required_props)
-            class_definition += field_def
-        
-        # Add constructor that sets the discriminator
-        class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Constructor that sets the discriminator value\n{INDENT}/// </summary>\n"
-        class_definition += f"{INDENT}public {class_name}()\n{INDENT}{{\n"
-        class_definition += f"{INDENT*2}this.{pascal(selector)} = \"{choice_name}\";\n"
-        class_definition += f"{INDENT}}}\n"
-        
-        # Generate Equals and GetHashCode
-        class_definition += self.generate_equals_and_gethashcode(schema, class_name, parent_namespace)
-        
-        class_definition += "}"
-        
-        if write_file:
-            self.write_to_file(namespace, class_name, class_definition)
-        
-        self.generated_types[ref] = "class"
-        self.generated_structure_types[ref] = schema
-        return ref
-
-    def generate_tuple(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
-        """ Generates a tuple type - Per JSON Structure spec, tuples serialize as JSON arrays, not objects """
-        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedTuple'))
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-        
-        properties = structure_schema.get('properties', {})
-        tuple_order = structure_schema.get('tuple', [])
-        
-        # Build list of tuple element types and names in correct order
-        tuple_elements = []
-        for prop_name in tuple_order:
-            if prop_name in properties:
-                prop_schema = properties[prop_name]
-                prop_type = self.convert_structure_type_to_csharp(class_name, prop_name, prop_schema, schema_namespace)
-                field_name = pascal(prop_name) if self.pascal_properties else prop_name
-                tuple_elements.append((prop_type, field_name))
-        
-        # Generate as a C# record struct with positional parameters
-        # Per JSON Structure spec: tuples serialize as JSON arrays like ["Alice", 42]
-        tuple_signature = ', '.join([f"{elem_type} {elem_name}" for elem_type, elem_name in tuple_elements])
-        
-        # Create the tuple record struct
-        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name)}\n/// </summary>\n"
-        class_definition += f"/// <remarks>\n/// JSON Structure tuple type - serializes as JSON array: [{', '.join(['...' for _ in tuple_elements])}]\n/// </remarks>\n"
-        
-        # Add JsonConverter attribute if System.Text.Json annotations are enabled
-        if self.system_text_json_annotation:
-            class_definition += f"[System.Text.Json.Serialization.JsonConverter(typeof(TupleJsonConverter<{class_name}>))]\n"
-        
-        class_definition += f"public record struct {class_name}({tuple_signature});\n"
-        
-        if write_file:
-            self.write_to_file(namespace, class_name, class_definition)
-        
-        self.generated_types[ref] = "tuple"
-        self.generated_structure_types[ref] = structure_schema
-        return ref
-
-    def generate_container_wrapper(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
-        """ Generates a wrapper class for root-level container types (map, array, set) """
-        struct_type = structure_schema.get('type', 'map')
-        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', f'Root{struct_type.capitalize()}'))
-        schema_namespace = structure_schema.get('namespace', parent_namespace)
-        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
-        
-        ref = 'global::'+self.get_qualified_name(namespace, class_name)
-        if ref in self.generated_types:
-            return ref
-        
-        # Determine the underlying collection type
-        value_type = "string"  # Default
-        item_type = "string"   # Default
-        underlying_type = "object"
-        
-        if struct_type == 'map':
-            values_schema = structure_schema.get('values', {'type': 'string'})
-            value_type = self.convert_structure_type_to_csharp(class_name, 'value', values_schema, schema_namespace)
-            underlying_type = f"Dictionary<string, {value_type}>"
-        elif struct_type == 'array':
-            items_schema = structure_schema.get('items', {'type': 'string'})
-            item_type = self.convert_structure_type_to_csharp(class_name, 'item', items_schema, schema_namespace)
-            underlying_type = f"List<{item_type}>"
-        elif struct_type == 'set':
-            items_schema = structure_schema.get('items', {'type': 'string'})
-            item_type = self.convert_structure_type_to_csharp(class_name, 'item', items_schema, schema_namespace)
-            underlying_type = f"HashSet<{item_type}>"
-        
-        # Generate wrapper class with implicit conversions
-        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name)}\n/// </summary>\n"
-        class_definition += f"/// <remarks>\n/// Wrapper for root-level {struct_type} type\n/// </remarks>\n"
-        class_definition += f"public class {class_name}\n{{\n"
-        class_definition += f"{INDENT}private {underlying_type} _value = new();\n\n"
-        
-        # Add indexer or collection access
-        if struct_type == 'map':
-            class_definition += f"{INDENT}public {value_type} this[string key]\n"
-            class_definition += f"{INDENT}{{\n"
-            class_definition += f"{INDENT*2}get => _value[key];\n"
-            class_definition += f"{INDENT*2}set => _value[key] = value;\n"
-            class_definition += f"{INDENT}}}\n\n"
-        elif struct_type in ('array', 'set'):
-            class_definition += f"{INDENT}public {item_type} this[int index]\n"
-            class_definition += f"{INDENT}{{\n"
-            if struct_type == 'array':
-                class_definition += f"{INDENT*2}get => _value[index];\n"
-                class_definition += f"{INDENT*2}set => _value[index] = value;\n"
-            else:  # set
-                class_definition += f"{INDENT*2}get => _value.ElementAt(index);\n"
-                class_definition += f"{INDENT*2}set => throw new NotSupportedException(\"Cannot set items by index in a HashSet\");\n"
-            class_definition += f"{INDENT}}}\n\n"
-        
-        # Add Count property
-        class_definition += f"{INDENT}public int Count => _value.Count;\n\n"
-        
-        # Add Add method for collections
-        if struct_type == 'map':
-            class_definition += f"{INDENT}public void Add(string key, {value_type} value) => _value.Add(key, value);\n\n"
-        elif struct_type in ('array', 'set'):
-            class_definition += f"{INDENT}public void Add({item_type} item) => _value.Add(item);\n\n"
-        
-        # Override Equals and GetHashCode for proper value equality
-        class_definition += f"{INDENT}public override bool Equals(object? obj)\n"
-        class_definition += f"{INDENT}{{\n"
-        class_definition += f"{INDENT*2}if (obj is not {class_name} other) return false;\n"
-        if struct_type == 'map':
-            class_definition += f"{INDENT*2}if (_value.Count != other._value.Count) return false;\n"
-            class_definition += f"{INDENT*2}foreach (var kvp in _value)\n"
-            class_definition += f"{INDENT*2}{{\n"
-            class_definition += f"{INDENT*3}if (!other._value.TryGetValue(kvp.Key, out var otherValue) || !Equals(kvp.Value, otherValue))\n"
-            class_definition += f"{INDENT*4}return false;\n"
-            class_definition += f"{INDENT*2}}}\n"
-            class_definition += f"{INDENT*2}return true;\n"
-        elif struct_type == 'array':
-            class_definition += f"{INDENT*2}return _value.SequenceEqual(other._value);\n"
-        else:  # set
-            class_definition += f"{INDENT*2}return _value.SetEquals(other._value);\n"
-        class_definition += f"{INDENT}}}\n\n"
-        
-        class_definition += f"{INDENT}public override int GetHashCode()\n"
-        class_definition += f"{INDENT}{{\n"
-        class_definition += f"{INDENT*2}var hash = new HashCode();\n"
-        if struct_type == 'map':
-            class_definition += f"{INDENT*2}foreach (var kvp in _value)\n"
-            class_definition += f"{INDENT*2}{{\n"
-            class_definition += f"{INDENT*3}hash.Add(kvp.Key);\n"
-            class_definition += f"{INDENT*3}hash.Add(kvp.Value);\n"
-            class_definition += f"{INDENT*2}}}\n"
-        else:  # array or set
-            class_definition += f"{INDENT*2}foreach (var item in _value)\n"
-            class_definition += f"{INDENT*2}{{\n"
-            class_definition += f"{INDENT*3}hash.Add(item);\n"
-            class_definition += f"{INDENT*2}}}\n"
-        class_definition += f"{INDENT*2}return hash.ToHashCode();\n"
-        class_definition += f"{INDENT}}}\n\n"
-        
-        # Implicit conversion to underlying type
-        class_definition += f"{INDENT}public static implicit operator {underlying_type}({class_name} wrapper) => wrapper._value;\n\n"
-        
-        # Implicit conversion from underlying type
-        class_definition += f"{INDENT}public static implicit operator {class_name}({underlying_type} value) => new() {{ _value = value }};\n"
-        
-        class_definition += "}\n"
-        
-        if write_file:
-            self.write_to_file(namespace, class_name, class_definition)
-        
-        self.generated_types[ref] = "class"
-        self.generated_structure_types[ref] = structure_schema
-        return ref
-
-
-    def generate_embedded_union(self, class_name: str, field_name: str, structure_types: List, parent_namespace: str, write_file: bool) -> str:
-        """ Generates an embedded Union Class """
-        # Similar to Avro's union handling, but for JSON Structure types
-        union_class_name = pascal(field_name)+'Union'
-        ref = class_name+'.'+union_class_name
-        
-        # For simplicity, generate as object type
-        # A complete implementation would generate a proper union class
-        return 'object'
-
-    def generate_equals_and_gethashcode(self, structure_schema: Dict, class_name: str, parent_namespace: str) -> str:
-        """ Generates Equals and GetHashCode methods for value equality """
-        code = "\n"
-        properties = structure_schema.get('properties', {})
-        
-        # Filter out const properties since they're static and same for all instances
-        non_const_properties = {k: v for k, v in properties.items() if 'const' not in v}
-        
-        if not non_const_properties:
-            # Empty class or only const fields - simple implementation
-            code += f"{INDENT}/// <summary>\n{INDENT}/// Determines whether the specified object is equal to the current object.\n{INDENT}/// </summary>\n"
-            code += f"{INDENT}public override bool Equals(object? obj)\n{INDENT}{{\n"
-            code += f"{INDENT*2}return obj is {class_name};\n"
-            code += f"{INDENT}}}\n\n"
-            code += f"{INDENT}/// <summary>\n{INDENT}/// Serves as the default hash function.\n{INDENT}/// </summary>\n"
-            code += f"{INDENT}public override int GetHashCode()\n{INDENT}{{\n"
-            code += f"{INDENT*2}return 0;\n"
-            code += f"{INDENT}}}\n"
-            return code
-        
-        # Generate Equals method
-        code += f"{INDENT}/// <summary>\n{INDENT}/// Determines whether the specified object is equal to the current object.\n{INDENT}/// </summary>\n"
-        code += f"{INDENT}public override bool Equals(object? obj)\n{INDENT}{{\n"
-        code += f"{INDENT*2}if (obj is not {class_name} other) return false;\n"
-        
-        # Build equality comparisons for each non-const property
-        equality_checks = []
-        for prop_name, prop_schema in non_const_properties.items():
-            field_name = prop_name
-            if self.is_csharp_reserved_word(field_name):
-                field_name = f"@{field_name}"
-            if self.pascal_properties:
-                field_name = pascal(field_name)
-            if field_name == class_name:
-                field_name += "_"
-            
-            field_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
-            
-            # Handle different types of comparisons
-            if field_type == 'byte[]' or field_type == 'byte[]?':
-                # Byte arrays need special handling
-                equality_checks.append(f"System.Linq.Enumerable.SequenceEqual(this.{field_name} ?? Array.Empty<byte>(), other.{field_name} ?? Array.Empty<byte>())")
-            elif field_type.startswith('Dictionary<'):
-                # Dictionaries need special comparison - compare keys and values
-                if field_type.endswith('?'):
-                    dict_compare = f"((this.{field_name} == null && other.{field_name} == null) || (this.{field_name} != null && other.{field_name} != null && this.{field_name}.Count == other.{field_name}.Count && this.{field_name}.All(kvp => other.{field_name}.TryGetValue(kvp.Key, out var val) && Equals(kvp.Value, val))))"
-                    equality_checks.append(dict_compare)
-                else:
-                    dict_compare = f"(this.{field_name}.Count == other.{field_name}.Count && this.{field_name}.All(kvp => other.{field_name}.TryGetValue(kvp.Key, out var val) && Equals(kvp.Value, val)))"
-                    equality_checks.append(dict_compare)
-            elif field_type.startswith('List<') or field_type.startswith('HashSet<'):
-                # Lists and HashSets need sequence comparison
-                if field_type.endswith('?'):
-                    equality_checks.append(f"((this.{field_name} == null && other.{field_name} == null) || (this.{field_name} != null && other.{field_name} != null && this.{field_name}.SequenceEqual(other.{field_name})))")
-                else:
-                    equality_checks.append(f"this.{field_name}.SequenceEqual(other.{field_name})")
-            else:
-                # Use Equals for reference types, == for value types
-                if field_type.endswith('?') or not self.is_csharp_primitive_type(field_type):
-                    equality_checks.append(f"Equals(this.{field_name}, other.{field_name})")
-                else:
-                    equality_checks.append(f"this.{field_name} == other.{field_name}")
-        
-        # Join all checks with &&
-        if len(equality_checks) == 1:
-            code += f"{INDENT*2}return {equality_checks[0]};\n"
-        else:
-            code += f"{INDENT*2}return " + f"\n{INDENT*3}&& ".join(equality_checks) + ";\n"
-        
-        code += f"{INDENT}}}\n\n"
-        
-        # Generate GetHashCode method
-        code += f"{INDENT}/// <summary>\n{INDENT}/// Serves as the default hash function.\n{INDENT}/// </summary>\n"
-        code += f"{INDENT}public override int GetHashCode()\n{INDENT}{{\n"
-        
-        # Collect field names for HashCode.Combine (skip const fields)
-        hash_fields = []
-        for prop_name, prop_schema in non_const_properties.items():
-            field_name = prop_name
-            if self.is_csharp_reserved_word(field_name):
-                field_name = f"@{field_name}"
-            if self.pascal_properties:
-                field_name = pascal(field_name)
-            if field_name == class_name:
-                field_name += "_"
-            
-            field_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
-            
-            # Handle special types that need custom hash code computation
-            if field_type == 'byte[]' or field_type == 'byte[]?':
-                hash_fields.append(f"({field_name} != null ? System.Convert.ToBase64String({field_name}).GetHashCode() : 0)")
-            elif field_type.startswith('List<') or field_type.startswith('HashSet<') or field_type.startswith('Dictionary<'):
-                # For collections, compute hash from elements
-                if field_type.endswith('?'):
-                    hash_fields.append(f"({field_name} != null ? {field_name}.Aggregate(0, (acc, item) => HashCode.Combine(acc, item)) : 0)")
-                else:
-                    hash_fields.append(f"{field_name}.Aggregate(0, (acc, item) => HashCode.Combine(acc, item))")
-            else:
-                hash_fields.append(field_name)
-        
-        # HashCode.Combine supports up to 8 parameters
-        if len(hash_fields) <= 8:
-            code += f"{INDENT*2}return HashCode.Combine({', '.join(hash_fields)});\n"
-        else:
-            # For more than 8 fields, use HashCode.Add
-            code += f"{INDENT*2}var hash = new HashCode();\n"
-            for field in hash_fields:
-                code += f"{INDENT*2}hash.Add({field});\n"
-            code += f"{INDENT*2}return hash.ToHashCode();\n"
-        
-        code += f"{INDENT}}}\n"
-        
-        return code
-
-    def write_to_file(self, namespace: str, name: str, definition: str) -> None:
-        """ Writes the class or enum to a file """
-        directory_path = os.path.join(
-            self.output_dir, os.path.join('src', namespace.replace('.', os.sep)))
-        if not os.path.exists(directory_path):
-            os.makedirs(directory_path, exist_ok=True)
-        file_path = os.path.join(directory_path, f"{name}.cs")
-
-        with open(file_path, 'w', encoding='utf-8') as file:
-            # Common using statements (add more as needed)
-            file_content = "using System;\nusing System.Collections.Generic;\n"
-            file_content += "using System.Linq;\n"
-            if self.system_text_json_annotation:
-                file_content += "using System.Text.Json;\n"
-                file_content += "using System.Text.Json.Serialization;\n"
-            if self.newtonsoft_json_annotation:
-                file_content += "using Newtonsoft.Json;\n"
-            if self.system_xml_annotation:  # Add XML serialization using directive
-                file_content += "using System.Xml.Serialization;\n"
-
-            if namespace:
-                # Namespace declaration with correct indentation for the definition
-                file_content += f"\nnamespace {namespace}\n{{\n"
-                indented_definition = '\n'.join(
-                    [f"{INDENT}{line}" for line in definition.split('\n')])
-                file_content += f"{indented_definition}\n}}"
-            else:
-                file_content += definition
-            file.write(file_content)
-
-    def convert(self, structure_schema_path: str, output_dir: str) -> None:
-        """ Converts a JSON Structure schema file to C# classes """
-        self.output_dir = output_dir
-        
-        with open(structure_schema_path, 'r', encoding='utf-8') as file:
-            schema = json.load(file)
-        
-        self.convert_schema(schema, output_dir)
-
-    def convert_schema(self, schema: JsonNode, output_dir: str) -> None:
-        """ Converts a JSON Structure schema to C# classes """
-        if not isinstance(schema, list):
-            schema = [schema]
-
-        # Determine project name: use explicit project_name if set, otherwise derive from base_namespace
-        if self.project_name and self.project_name.strip():
-            # Use explicitly set project name
-            project_name = self.project_name
-        else:
-            # Fall back to using base_namespace as project name
-            project_name = self.base_namespace
-            if not project_name or project_name.strip() == '':
-                # Derive from output directory name as fallback
-                project_name = os.path.basename(os.path.abspath(output_dir))
-                if not project_name or project_name.strip() == '':
-                    project_name = 'Generated'
-                # Clean up the project name
-                project_name = project_name.replace('-', '_').replace(' ', '_')
-                # Update base_namespace to match (only if it was empty)
-                self.base_namespace = project_name
-                import warnings
-                warnings.warn(f"No namespace provided, using '{project_name}' derived from output directory", UserWarning)
-        
-        self.schema_doc = schema
-        if not os.path.exists(output_dir):
-            os.makedirs(output_dir, exist_ok=True)
-        
-        # Create solution file if it doesn't exist
-        if not glob.glob(os.path.join(output_dir, "src", "*.sln")):
-            sln_file = os.path.join(output_dir, f"{project_name}.sln")
-            if not os.path.exists(sln_file):
-                if not os.path.exists(os.path.dirname(sln_file)) and os.path.dirname(sln_file):
-                    os.makedirs(os.path.dirname(sln_file))
-                with open(sln_file, 'w', encoding='utf-8') as file:
-                    file.write(process_template(
-                        "structuretocsharp/project.sln.jinja", 
-                        project_name=project_name, 
-                        uuid=lambda:str(uuid.uuid4()),
-                        system_xml_annotation=self.system_xml_annotation,
-                        system_text_json_annotation=self.system_text_json_annotation,
-                        newtonsoft_json_annotation=self.newtonsoft_json_annotation))
-        
-        # Create main project file if it doesn't exist
-        if not glob.glob(os.path.join(output_dir, "src", "*.csproj")):
-            csproj_file = os.path.join(output_dir, "src", f"{pascal(project_name)}.csproj")
-            if not os.path.exists(csproj_file):
-                if not os.path.exists(os.path.dirname(csproj_file)):
-                    os.makedirs(os.path.dirname(csproj_file))
-                with open(csproj_file, 'w', encoding='utf-8') as file:
-                    file.write(process_template(
-                        "structuretocsharp/project.csproj.jinja",
-                        project_name=project_name, 
-                        system_xml_annotation=self.system_xml_annotation,
-                        system_text_json_annotation=self.system_text_json_annotation,
-                        newtonsoft_json_annotation=self.newtonsoft_json_annotation))
-        
-        # Create test project file if it doesn't exist
-        if not glob.glob(os.path.join(output_dir, "test", "*.csproj")):
-            csproj_test_file = os.path.join(output_dir, "test", f"{pascal(project_name)}.Test.csproj")
-            if not os.path.exists(csproj_test_file):
-                if not os.path.exists(os.path.dirname(csproj_test_file)):
-                    os.makedirs(os.path.dirname(csproj_test_file))
-                with open(csproj_test_file, 'w', encoding='utf-8') as file:
-                    file.write(process_template(
-                        "structuretocsharp/testproject.csproj.jinja", 
-                        project_name=project_name,
-                        system_xml_annotation=self.system_xml_annotation,
-                        system_text_json_annotation=self.system_text_json_annotation,
-                        newtonsoft_json_annotation=self.newtonsoft_json_annotation))
-
-        self.output_dir = output_dir
-        
-        # Register all schemas with $id keywords for cross-references
-        for structure_schema in (s for s in schema if isinstance(s, dict)):
-            self.register_schema_ids(structure_schema)
-        
-        # Process each schema
-        for structure_schema in (s for s in schema if isinstance(s, dict)):
-            # Store definitions for later use
-            if 'definitions' in structure_schema:
-                self.definitions = structure_schema['definitions']
-            
-            # Store $offers for add-in system
-            if '$offers' in structure_schema:
-                self.offers = structure_schema['$offers']
-            
-            # Process root type FIRST so inline unions can generate derived classes
-            if 'type' in structure_schema:
-                self.generate_class_or_choice(structure_schema, '', write_file=True)
-            elif '$root' in structure_schema:
-                root_ref = structure_schema['$root']
-                root_schema = self.resolve_ref(root_ref, structure_schema)
-                if root_schema:
-                    ref_path = root_ref.split('/')
-                    type_name = ref_path[-1]
-                    ref_namespace = '.'.join(ref_path[2:-1]) if len(ref_path) > 3 else ''
-                    self.generate_class_or_choice(root_schema, ref_namespace, write_file=True, explicit_name=type_name)
-            
-            # Now process remaining definitions that weren't generated as part of inline unions
-            if 'definitions' in structure_schema:
-                self.process_definitions(self.definitions, '')
-            
-            # Generate add-in interfaces and extensible wrapper classes
-            if self.offers:
-                self.generate_addins(structure_schema)
-        
-        # Generate tuple converter utility class if needed (after all types processed)
-        if self.system_text_json_annotation:
-            self.generate_tuple_converter(output_dir)
-        
-        # Generate tests
-        self.generate_tests(output_dir)
-        
-        # Generate instance serializer program
-        self.generate_instance_serializer(output_dir)
-
-    def generate_addins(self, structure_schema: Dict) -> None:
-        """
-        Generates add-in interfaces and view classes for types that have $offers.
-        
-        For each add-in in $offers, creates:
-        1. An interface I{AddinName} with the add-in properties
-        2. An internal view class that wraps the Extensions dictionary
-        3. Implicit operators on the base class that convert to the interface
-        """
-        if not self.offers or not isinstance(self.offers, dict):
-            return
-        
-        root_type_name = structure_schema.get('name', 'Document')
-        namespace_pascal = pascal(self.base_namespace)
-        
-        # Generate interface and view class for each add-in
-        view_classes = []
-        for addin_name, addin_def in self.offers.items():
-            self.generate_addin_interface(addin_name, addin_def, namespace_pascal)
-            view_class_name = self.generate_addin_view_class(addin_name, addin_def, namespace_pascal)
-            view_classes.append((addin_name, view_class_name))
-        
-        # Add Extensions dictionary and implicit operators to the base class
-        if 'type' in structure_schema and structure_schema['type'] == 'object':
-            self.add_extensions_to_base_class(root_type_name, view_classes, namespace_pascal)
-
-    def generate_addin_interface(self, addin_name: str, addin_def: Any, namespace: str) -> None:
-        """
-        Generates an interface for an add-in from $offers.
-        
-        Args:
-            addin_name: Name of the add-in (e.g., "AuditInfo")
-            addin_def: Definition of the add-in (either inline properties or a $ref)
-            namespace: Target namespace for the interface
-        """
-        interface_name = f"I{pascal(addin_name)}"
-        
-        # Resolve the add-in definition if it's a reference
-        if isinstance(addin_def, str):
-            # It's a JSON pointer reference
-            addin_def = self.resolve_ref(addin_def, self.schema_doc)
-        elif isinstance(addin_def, dict) and '$ref' in addin_def:
-            addin_def = self.resolve_ref(addin_def['$ref'], self.schema_doc)
-        
-        if not addin_def or not isinstance(addin_def, dict):
-            return
-        
-        properties = addin_def.get('properties', {})
-        if not properties:
-            return
-        
-        # Generate interface definition
-        interface_code = f"{INDENT}/// <summary>\n"
-        interface_code += f"{INDENT}/// Add-in interface: {addin_name}\n"
-        if 'description' in addin_def:
-            interface_code += f"{INDENT}/// {addin_def['description']}\n"
-        interface_code += f"{INDENT}/// </summary>\n"
-        interface_code += f"{INDENT}public interface {interface_name}\n"
-        interface_code += f"{INDENT}{{\n"
-        
-        # Generate properties
-        for prop_name, prop_schema in properties.items():
-            if not isinstance(prop_schema, dict):
-                continue
-                
-            csharp_prop_name = pascal(prop_name) if self.pascal_properties else prop_name
-            csharp_type = self.convert_structure_type_to_csharp(interface_name, prop_name, prop_schema, namespace)
-            
-            # Add XML doc comment
-            if 'description' in prop_schema:
-                interface_code += f"{INDENT}{INDENT}/// <summary>\n"
-                interface_code += f"{INDENT}{INDENT}/// {prop_schema['description']}\n"
-                interface_code += f"{INDENT}{INDENT}/// </summary>\n"
-            
-            # Interface properties are always nullable for add-ins (both value types and reference types)
-            if not csharp_type.endswith('?'):
-                csharp_type += '?'
-            
-            interface_code += f"{INDENT}{INDENT}{csharp_type} {csharp_prop_name} {{ get; set; }}\n"
-        
-        interface_code += f"{INDENT}}}\n"
-        
-        # Write interface to file
-        self.write_to_file(namespace, interface_name, interface_code)
-        
-        # Track as generated
-        qualified_name = 'global::' + self.get_qualified_name(namespace, interface_name)
-        self.generated_types[qualified_name] = "interface"
-
-    def generate_extensible_class(self, base_type_name: str, addin_names: List[str], namespace: str) -> None:
-        """
-        DEPRECATED: Replaced by generate_addin_view_class and add_extensions_to_base_class.
-        This method is kept for backward compatibility but does nothing.
-        """
-        pass
-
-    def generate_addin_view_class(self, addin_name: str, addin_def: Any, namespace: str) -> str:
-        """
-        Generates an internal view class that wraps the Extensions dictionary.
-        
-        Example output:
-        internal sealed class AuditInfoView : IAuditInfo
-        {
-            private readonly Dictionary<string, object?> _extensions;
-            
-            public AuditInfoView(Dictionary<string, object?> extensions)
-            {
-                _extensions = extensions;
-            }
-            
-            public string? CreatedBy
-            {
-                get => _extensions.TryGetValue("createdBy", out var val) ? val as string : null;
-                set { if (value != null) _extensions["createdBy"] = value; else _extensions.Remove("createdBy"); }
-            }
-        }
-        
-        Args:
-            addin_name: Name of the add-in (e.g., "AuditInfo")
-            addin_def: Definition of the add-in
-            namespace: Target namespace
-            
-        Returns:
-            The name of the generated view class
-        """
-        view_class_name = f"{pascal(addin_name)}View"
-        interface_name = f"I{pascal(addin_name)}"
-        
-        # Resolve the add-in definition if it's a reference
-        if isinstance(addin_def, str):
-            addin_def = self.resolve_ref(addin_def, self.schema_doc)
-        elif isinstance(addin_def, dict) and '$ref' in addin_def:
-            addin_def = self.resolve_ref(addin_def['$ref'], self.schema_doc)
-        
-        if not addin_def or not isinstance(addin_def, dict):
-            return view_class_name
-        
-        properties = addin_def.get('properties', {})
-        if not properties:
-            return view_class_name
-        
-        # Generate class definition
-        class_code = f"{INDENT}/// <summary>\n"
-        class_code += f"{INDENT}/// View class wrapping Extensions dictionary for {addin_name} add-in\n"
-        if 'description' in addin_def:
-            class_code += f"{INDENT}/// {addin_def['description']}\n"
-        class_code += f"{INDENT}/// </summary>\n"
-        class_code += f"{INDENT}public sealed class {view_class_name} : {interface_name}\n"
-        class_code += f"{INDENT}{{\n"
-        
-        # Add private field
-        class_code += f"{INDENT}{INDENT}private readonly Dictionary<string, object?> _extensions;\n\n"
-        
-        # Add constructor
-        class_code += f"{INDENT}{INDENT}public {view_class_name}(Dictionary<string, object?> extensions)\n"
-        class_code += f"{INDENT}{INDENT}{{\n"
-        class_code += f"{INDENT}{INDENT}{INDENT}_extensions = extensions;\n"
-        class_code += f"{INDENT}{INDENT}}}\n\n"
-        
-        # Generate properties
-        for prop_name, prop_schema in properties.items():
-            if not isinstance(prop_schema, dict):
-                continue
-                
-            csharp_prop_name = pascal(prop_name) if self.pascal_properties else prop_name
-            csharp_type = self.convert_structure_type_to_csharp(view_class_name, prop_name, prop_schema, namespace)
-            
-            # Remove nullable marker for determining base type
-            base_csharp_type = csharp_type.rstrip('?')
-            is_nullable = csharp_type.endswith('?')
-            
-            # Ensure nullable for add-ins
-            if not is_nullable:
-                csharp_type += '?'
-            
-            # Add XML doc comment
-            if 'description' in prop_schema:
-                class_code += f"{INDENT}{INDENT}/// <summary>\n"
-                class_code += f"{INDENT}{INDENT}/// {prop_schema['description']}\n"
-                class_code += f"{INDENT}{INDENT}/// </summary>\n"
-            
-            # Generate getter that reads from dictionary
-            class_code += f"{INDENT}{INDENT}public {csharp_type} {csharp_prop_name}\n"
-            class_code += f"{INDENT}{INDENT}{{\n"
-            
-            # Getter - use TryGetValue with type-specific conversion
-            class_code += f'{INDENT}{INDENT}{INDENT}get => _extensions.TryGetValue("{prop_name}", out var val) && val != null ? '
-            
-            # Add appropriate conversion based on type
-            if base_csharp_type in ['string', 'bool', 'int', 'long', 'float', 'double', 'decimal']:
-                if base_csharp_type == 'string':
-                    class_code += 'val as string : null;\n'
-                elif base_csharp_type == 'bool':
-                    class_code += 'Convert.ToBoolean(val) : null;\n'
-                elif base_csharp_type in ['int', 'long', 'float', 'double', 'decimal']:
-                    class_code += f'Convert.To{base_csharp_type.capitalize()}(val) : null;\n'
-                else:
-                    class_code += 'val : null;\n'
-            else:
-                # For complex types, try direct cast
-                class_code += f'({base_csharp_type})val : null;\n'
-            
-            # Setter - write to dictionary or remove if null
-            class_code += f'{INDENT}{INDENT}{INDENT}set {{ if (value != null) _extensions["{prop_name}"] = value; else _extensions.Remove("{prop_name}"); }}\n'
-            
-            class_code += f"{INDENT}{INDENT}}}\n\n"
-        
-        class_code += f"{INDENT}}}\n"
-        
-        # Write class to file
-        self.write_to_file(namespace, view_class_name, class_code)
-        
-        # Track as generated (internal, not exported)
-        qualified_name = 'global::' + self.get_qualified_name(namespace, view_class_name)
-        self.generated_types[qualified_name] = "view_class"
-        
-        return view_class_name
-
-    def add_extensions_to_base_class(self, base_type_name: str, view_classes: List[tuple], namespace: str) -> None:
-        """
-        Adds Extensions dictionary property and implicit operators to the base class.
-        
-        Appends to the existing base class file:
-        - Extensions property (Dictionary<string, object?>)
-        - Implicit operators for each add-in interface
-        
-        Args:
-            base_type_name: Name of the base type
-            view_classes: List of (addin_name, view_class_name) tuples
-            namespace: Target namespace
-        """
-        base_class_name = pascal(base_type_name)
-        
-        # Generate the partial class extension code
-        extension_code = f"{INDENT}/// <summary>\n"
-        extension_code += f"{INDENT}/// Partial class extension for {base_class_name} with add-in support\n"
-        extension_code += f"{INDENT}/// </summary>\n"
-        extension_code += f"{INDENT}public partial class {base_class_name}\n"
-        extension_code += f"{INDENT}{{\n"
-        
-        # Add Extensions property
-        extension_code += f"{INDENT}{INDENT}/// <summary>\n"
-        extension_code += f"{INDENT}{INDENT}/// Extension properties storage for add-ins.\n"
-        extension_code += f"{INDENT}{INDENT}/// Unknown JSON properties are automatically captured here during deserialization.\n"
-        extension_code += f"{INDENT}{INDENT}/// </summary>\n"
-        
-        if self.system_text_json_annotation:
-            extension_code += f'{INDENT}{INDENT}[System.Text.Json.Serialization.JsonExtensionData]\n'
-        if self.newtonsoft_json_annotation:
-            extension_code += f'{INDENT}{INDENT}[Newtonsoft.Json.JsonExtensionData]\n'
-        
-        extension_code += f"{INDENT}{INDENT}public Dictionary<string, object?> Extensions {{ get; set; }} = new();\n\n"
-        
-        # Add implicit operators for each add-in
-        for addin_name, view_class_name in view_classes:
-            interface_name = f"I{pascal(addin_name)}"
-            
-            extension_code += f"{INDENT}{INDENT}/// <summary>\n"
-            extension_code += f"{INDENT}{INDENT}/// Implicit conversion to {interface_name} view\n"
-            extension_code += f"{INDENT}{INDENT}/// </summary>\n"
-            extension_code += f"{INDENT}{INDENT}public static implicit operator {view_class_name}({base_class_name} obj)\n"
-            extension_code += f"{INDENT}{INDENT}{INDENT}=> new {view_class_name}(obj.Extensions);\n\n"
-        
-        extension_code += f"{INDENT}}}\n"
-        
-        # Write to a separate file (e.g., ProductExtensions.cs)
-        extension_file_name = f"{base_class_name}Extensions"
-        self.write_to_file(namespace, extension_file_name, extension_code)
-
-    def process_definitions(self, definitions: Dict, namespace_path: str) -> None:
-        """ Processes the definitions section recursively """
-        for name, definition in definitions.items():
-            if isinstance(definition, dict):
-                if 'type' in definition:
-                    # This is a type definition
-                    current_namespace = self.concat_namespace(namespace_path, '')
-                    # Check if this type was already generated (e.g., as part of inline union)
-                    check_namespace = pascal(self.concat_namespace(self.base_namespace, current_namespace))
-                    check_name = pascal(name)
-                    check_ref = 'global::'+self.get_qualified_name(check_namespace, check_name)
-                    if check_ref not in self.generated_types:
-                        self.generate_class_or_choice(definition, current_namespace, write_file=True, explicit_name=name)
-                else:
-                    # This is a namespace
-                    new_namespace = self.concat_namespace(namespace_path, name)
-                    self.process_definitions(definition, new_namespace)
-
-    def generate_tests(self, output_dir: str) -> None:
-        """ Generates unit tests for all the generated C# classes and enums """
-        test_directory_path = os.path.join(output_dir, "test")
-        if not os.path.exists(test_directory_path):
-            os.makedirs(test_directory_path, exist_ok=True)
-
-        for class_name, type_kind in self.generated_types.items():
-            # Skip test generation for:
-            # 1. View classes (internal wrappers for Extensions dictionary)
-            # 2. Extension partial classes (add implicit operators to base classes)
-            base_name = class_name.split('.')[-1]
-            
-            # Skip view classes (e.g., AuditInfoView)
-            if type_kind == "view_class" or base_name.endswith('View'):
-                continue
-            
-            # Skip extension partial classes (e.g., ProductExtensions)
-            if base_name.endswith('Extensions'):
-                continue
-            
-            if type_kind in ["class", "enum"]:
-                self.generate_test_class(class_name, type_kind, test_directory_path)
-
-    def generate_tuple_converter(self, output_dir: str) -> None:
-        """ Generates the TupleJsonConverter utility class for JSON array serialization """
-        # Check if any tuples were generated
-        has_tuples = any(type_kind == "tuple" for type_kind in self.generated_types.values())
-        if not has_tuples:
-            return  # No tuples, no need for converter
-
-        # Convert base namespace to PascalCase for consistency with other generated classes
-        namespace_pascal = pascal(self.base_namespace)
-        
-        # Generate the converter class
-        converter_definition = process_template(
-            "structuretocsharp/tuple_converter.cs.jinja",
-            namespace=namespace_pascal
-        )
-
-        # Write to the same directory structure as other classes (using PascalCase path)
-        directory_path = os.path.join(
-            output_dir, os.path.join('src', namespace_pascal.replace('.', os.sep)))
-        if not os.path.exists(directory_path):
-            os.makedirs(directory_path, exist_ok=True)
-        converter_file_path = os.path.join(directory_path, "TupleJsonConverter.cs")
-        
-        # Add using statements
-        file_content = "using System;\n"
-        file_content += "using System.Linq;\n"
-        file_content += "using System.Reflection;\n"
-        file_content += "using System.Text.Json;\n"
-        file_content += "using System.Text.Json.Serialization;\n\n"
-        file_content += converter_definition
-        
-        with open(converter_file_path, 'w', encoding='utf-8') as converter_file:
-            converter_file.write(file_content)
-
-    def generate_instance_serializer(self, output_dir: str) -> None:
-        """ Generates InstanceSerializer.cs that creates instances and serializes them to JSON """
-        test_directory_path = os.path.join(output_dir, "test")
-        if not os.path.exists(test_directory_path):
-            os.makedirs(test_directory_path, exist_ok=True)
-
-        # Collect all classes (not enums, tuples, or other types) that have test classes
-        # Skip abstract classes since they cannot be instantiated
-        # Skip view classes and extension partial classes
-        classes = []
-        for class_name, type_kind in self.generated_types.items():
-            if type_kind == "class":
-                base_name = class_name.split('.')[-1]
-                
-                # Skip view classes (internal wrappers for Extensions dictionary)
-                if base_name.endswith('View'):
-                    continue
-                
-                # Skip extension partial classes
-                if base_name.endswith('Extensions'):
-                    continue
-                
-                # Skip abstract classes
-                structure_schema = cast(Dict[str, JsonNode], self.generated_structure_types.get(class_name, {}))
-                if structure_schema.get('abstract', False):
-                    continue
-                    
-                if class_name.startswith("global::"):
-                    class_name = class_name[8:]
-                test_class_name = f"{class_name.split('.')[-1]}Tests"
-                class_base_name = class_name.split('.')[-1]
-                
-                # Get proper namespace from class_name
-                if '.' in class_name:
-                    namespace = ".".join(class_name.split('.')[:-1])
-                else:
-                    namespace = self.base_namespace if self.base_namespace else ''
-                
-                # Build fully qualified test name
-                full_qualified_test_name = f"{namespace}.{test_class_name}" if namespace else test_class_name
-                
-                classes.append({
-                    'class_name': class_base_name,
-                    'test_class_name': test_class_name,
-                    'full_name': class_name,
-                    'full_qualified_test_name': full_qualified_test_name
-                })
-
-        if not classes:
-            return  # No classes to serialize
-
-        program_definition = process_template(
-            "structuretocsharp/program.cs.jinja",
-            classes=classes
-        )
-
-        program_file_path = os.path.join(test_directory_path, "InstanceSerializer.cs")
-        with open(program_file_path, 'w', encoding='utf-8') as program_file:
-            program_file.write(program_definition)
-
-    def generate_test_class(self, class_name: str, type_kind: str, test_directory_path: str) -> None:
-        """ Generates a unit test class for a given C# class or enum """
-        structure_schema: Dict[str, JsonNode] = cast(Dict[str, JsonNode], self.generated_structure_types.get(class_name, {}))
-        if class_name.startswith("global::"):
-            class_name = class_name[8:]
-        test_class_name = f"{class_name.split('.')[-1]}Tests"
-        namespace = ".".join(class_name.split('.')[:-1])
-        class_base_name = class_name.split('.')[-1]
-
-        # Skip test generation for abstract classes (cannot be instantiated)
-        if type_kind == "class" and structure_schema.get('abstract', False):
-            return
-
-        if type_kind == "class":
-            fields = self.get_class_test_fields(structure_schema, class_base_name)
-            test_class_definition = process_template(
-                "structuretocsharp/class_test.cs.jinja",
-                namespace=namespace,
-                test_class_name=test_class_name,
-                class_base_name=class_base_name,
-                fields=fields,
-                system_xml_annotation=self.system_xml_annotation,
-                system_text_json_annotation=self.system_text_json_annotation,
-                newtonsoft_json_annotation=self.newtonsoft_json_annotation
-            )
-        elif type_kind == "enum":
-            # For enums, extract symbols from the enum schema
-            enum_values = structure_schema.get('enum', [])
-            symbols = []
-            if enum_values:
-                for value in enum_values:
-                    if isinstance(value, str):
-                        # Convert to PascalCase enum member name
-                        symbol_name = ''.join(word.capitalize() for word in re.split(r'[_\-\s]+', value))
-                        symbols.append(symbol_name)
-                    else:
-                        # For numeric enums, use Value1, Value2, etc.
-                        symbols.append(f"Value{value}")
-            
-            test_class_definition = process_template(
-                "structuretocsharp/enum_test.cs.jinja",
-                namespace=namespace,
-                test_class_name=test_class_name,
-                enum_base_name=class_base_name,
-                symbols=symbols,
-                system_xml_annotation=self.system_xml_annotation,
-                system_text_json_annotation=self.system_text_json_annotation,
-                newtonsoft_json_annotation=self.newtonsoft_json_annotation
-            )
-        else:
-            return
-
-        test_file_path = os.path.join(test_directory_path, f"{test_class_name}.cs")
-        with open(test_file_path, 'w', encoding='utf-8') as test_file:
-            test_file.write(test_class_definition)
-
-    def get_class_test_fields(self, structure_schema: Dict[str, JsonNode], class_name: str) -> List[Any]:
-        """ Retrieves fields for a given class name """
-
-        class Field:
-            def __init__(self, fn: str, ft: str, tv: Any, ct: bool, pm: bool):
-                self.field_name = fn
-                self.field_type = ft
-                self.test_value = tv
-                self.is_const = ct
-                self.is_primitive = pm
-
-        fields: List[Field] = []
-        if structure_schema and 'properties' in structure_schema:
-            for prop_name, prop_schema in cast(Dict[str, Dict], structure_schema['properties']).items():
-                field_name = prop_name
-                if self.pascal_properties:
-                    field_name = pascal(field_name)
-                if field_name == class_name:
-                    field_name += "_"
-                if self.is_csharp_reserved_word(field_name):
-                    field_name = f"@{field_name}"
-                
-                field_type = self.convert_structure_type_to_csharp(
-                    class_name, field_name, prop_schema, str(structure_schema.get('namespace', '')))
-                is_class = field_type in self.generated_types and self.generated_types[field_type] == "class"
-                
-                # Check if this is a const field
-                is_const = 'const' in prop_schema
-                test_value = self.get_test_value(field_type) if not is_const else self.format_default_value(prop_schema['const'], field_type)
-                
-                f = Field(field_name, field_type, test_value, is_const, not is_class)
-                fields.append(f)
-        return cast(List[Any], fields)
-
-    def get_test_value(self, csharp_type: str) -> str:
-        """Returns a default test value based on the C# type"""
-        # For nullable object types, return typed null to avoid var issues
-        if csharp_type == "object?" or csharp_type == "object":
-            return "null"  # Use null for object types (typically unions) to avoid reference inequality
-        
-        test_values = {
-            'string': '"test_string"',
-            'bool': 'true',
-            'sbyte': '(sbyte)42',
-            'byte': '(byte)42',
-            'short': '(short)42',
-            'ushort': '(ushort)42',
-            'int': '42',
-            'uint': '42U',
-            'long': '42L',
-            'ulong': '42UL',
-            'System.Int128': 'new System.Int128(0, 42)',
-            'System.UInt128': 'new System.UInt128(0, 42)',
-            'float': '3.14f',
-            'double': '3.14',
-            'decimal': '3.14m',
-            'byte[]': 'new byte[] { 0x01, 0x02, 0x03 }',
-            'DateOnly': 'new DateOnly(2024, 1, 1)',
-            'TimeOnly': 'new TimeOnly(12, 0, 0)',
-            'DateTimeOffset': 'new DateTimeOffset(2024, 1, 1, 12, 0, 0, TimeSpan.Zero)',
-            'TimeSpan': 'TimeSpan.FromHours(1)',
-            'Guid': 'new Guid("12345678-1234-1234-1234-123456789012")',
-            'Uri': 'new Uri("https://example.com")',
-            'null': 'null'
-        }
-        if csharp_type.endswith('?'):
-            csharp_type = csharp_type[:-1]
-        
-        # Normalize to use qualified reference (strip global:: prefix if present, then add it)
-        base_type = csharp_type.replace('global::', '')
-        qualified_ref = f'global::{base_type}'
-        
-        # Check if this is a tuple type (generated_types tracks what we've created)
-        if qualified_ref in self.generated_types and self.generated_types[qualified_ref] == "tuple":
-            # For tuple types, we need to construct with test values based on the schema
-            schema = self.generated_structure_types.get(qualified_ref)
-            if schema:
-                tuple_order = schema.get('tuple', [])
-                properties = schema.get('properties', {})
-                test_params = []
-                for prop_name in tuple_order:
-                    if prop_name in properties:
-                        prop_schema = properties[prop_name]
-                        prop_type = self.convert_structure_type_to_csharp(base_type, prop_name, prop_schema, str(schema.get('namespace', '')))
-                        test_params.append(self.get_test_value(prop_type))
-                if test_params:
-                    return f'new {base_type}({", ".join(test_params)})'
-        
-        # Check if this is a choice type (discriminated union)
-        if qualified_ref in self.generated_types and self.generated_types[qualified_ref] == "choice":
-            # For choice types, create an instance with the first choice property set
-            schema = self.generated_structure_types.get(qualified_ref)
-            if schema:
-                choices = schema.get('choices', {})
-                if choices:
-                    # Get the first choice property
-                    first_choice_name, first_choice_schema = next(iter(choices.items()))
-                    choice_type = self.convert_structure_type_to_csharp(base_type, first_choice_name, first_choice_schema, str(schema.get('namespace', '')))
-                    choice_test_value = self.get_test_value(choice_type)
-                    # Use the constructor that takes the first choice
-                    return f'new {base_type}({choice_test_value})'
-        
-        return test_values.get(base_type, test_values.get(csharp_type, f'new {csharp_type}()'))
-
-
-
-
-def convert_structure_to_csharp(
-    structure_schema_path: str, 
-    cs_file_path: str, 
-    base_namespace: str = '',
-    project_name: str = '',
-    pascal_properties: bool = False, 
-    system_text_json_annotation: bool = False, 
-    newtonsoft_json_annotation: bool = False, 
-    system_xml_annotation: bool = False
-):
-    """Converts JSON Structure schema to C# classes
-
-    Args:
-        structure_schema_path (str): JSON Structure input schema path
-        cs_file_path (str): Output C# file path
-        base_namespace (str, optional): Base namespace. Defaults to ''.
-        project_name (str, optional): Explicit project name for .csproj files (separate from namespace). Defaults to ''.
-        pascal_properties (bool, optional): Pascal case properties. Defaults to False.
-        system_text_json_annotation (bool, optional): Use System.Text.Json annotations. Defaults to False.
-        newtonsoft_json_annotation (bool, optional): Use Newtonsoft.Json annotations. Defaults to False.
-        system_xml_annotation (bool, optional): Use System.Xml.Serialization annotations. Defaults to False.
-    """
-
-    if not base_namespace:
-        base_namespace = os.path.splitext(os.path.basename(cs_file_path))[0].replace('-', '_')
-    
-    structtocs = StructureToCSharp(base_namespace)
-    structtocs.project_name = project_name
-    structtocs.pascal_properties = pascal_properties
-    structtocs.system_text_json_annotation = system_text_json_annotation
-    structtocs.newtonsoft_json_annotation = newtonsoft_json_annotation
-    structtocs.system_xml_annotation = system_xml_annotation
-    structtocs.convert(structure_schema_path, cs_file_path)
-
-
-def convert_structure_schema_to_csharp(
-    structure_schema: JsonNode, 
-    output_dir: str, 
-    base_namespace: str = '',
-    project_name: str = '',
-    pascal_properties: bool = False, 
-    system_text_json_annotation: bool = False, 
-    newtonsoft_json_annotation: bool = False, 
-    system_xml_annotation: bool = False
-):
-    """Converts JSON Structure schema to C# classes
-
-    Args:
-        structure_schema (JsonNode): JSON Structure schema to convert
-        output_dir (str): Output directory
-        base_namespace (str, optional): Base namespace for the generated classes. Defaults to ''.
-        project_name (str, optional): Explicit project name for .csproj files (separate from namespace). Defaults to ''.
-        pascal_properties (bool, optional): Pascal case properties. Defaults to False.
-        system_text_json_annotation (bool, optional): Use System.Text.Json annotations. Defaults to False.
-        newtonsoft_json_annotation (bool, optional): Use Newtonsoft.Json annotations. Defaults to False.
-        system_xml_annotation (bool, optional): Use System.Xml.Serialization annotations. Defaults to False.
-    """
-    structtocs = StructureToCSharp(base_namespace)
-    structtocs.project_name = project_name
-    structtocs.pascal_properties = pascal_properties
-    structtocs.system_text_json_annotation = system_text_json_annotation
-    structtocs.newtonsoft_json_annotation = newtonsoft_json_annotation
-    structtocs.system_xml_annotation = system_xml_annotation
-    structtocs.convert_schema(structure_schema, output_dir)
+# pylint: disable=line-too-long
+
+""" StructureToCSharp class for converting JSON Structure schema to C# classes """
+
+import json
+import os
+import re
+from typing import Any, Dict, List, Tuple, Union, cast, Optional
+import uuid
+
+from avrotize.common import pascal, process_template
+import glob
+
+JsonNode = Dict[str, 'JsonNode'] | List['JsonNode'] | str | None
+
+
+INDENT = '    '
+
+
+class StructureToCSharp:
+    """ Converts JSON Structure schema to C# classes """
+
+    def __init__(self, base_namespace: str = '') -> None:
+        self.base_namespace = base_namespace
+        self.project_name: str = ''  # Optional explicit project name, separate from namespace
+        self.schema_doc: JsonNode = None
+        self.output_dir = os.getcwd()
+        self.pascal_properties = False
+        self.system_text_json_annotation = False
+        self.newtonsoft_json_annotation = False
+        self.system_xml_annotation = False
+        self.generated_types: Dict[str,str] = {}
+        self.generated_structure_types: Dict[str, Dict[str, Union[str, Dict, List]]] = {}
+        self.type_dict: Dict[str, Dict] = {}
+        self.definitions: Dict[str, Any] = {}
+        self.schema_registry: Dict[str, Dict] = {}  # Maps $id URIs to schemas
+        self.offers: Dict[str, Any] = {}  # Maps add-in names to property definitions from $offers
+
+    def get_qualified_name(self, namespace: str, name: str) -> str:
+        """ Concatenates namespace and name with a dot separator """
+        return f"{namespace}.{name}" if namespace != '' else name
+
+    def concat_namespace(self, namespace: str, name: str) -> str:
+        """ Concatenates namespace and name with a dot separator """
+        if namespace and name:
+            return f"{namespace}.{name}"
+        elif namespace:
+            return namespace
+        else:
+            return name
+
+    def map_primitive_to_csharp(self, structure_type: str) -> str:
+        """ Maps JSON Structure primitive types to C# types """
+        mapping = {
+            'null': 'void',  # Placeholder, actual handling for nullable types is in the union logic
+            'boolean': 'bool',
+            'string': 'string',
+            'integer': 'int',  # Generic integer type without format
+            'number': 'double',  # Generic number type without format
+            'int8': 'sbyte',
+            'uint8': 'byte',
+            'int16': 'short',
+            'uint16': 'ushort',
+            'int32': 'int',
+            'uint32': 'uint',
+            'int64': 'long',
+            'uint64': 'ulong',
+            'int128': 'System.Int128',
+            'uint128': 'System.UInt128',
+            'float8': 'float',  # Approximation - C# doesn't have native 8-bit float
+            'float': 'float',
+            'double': 'double',
+            'binary32': 'float',  # IEEE 754 binary32
+            'binary64': 'double',  # IEEE 754 binary64
+            'decimal': 'decimal',
+            'binary': 'byte[]',
+            'date': 'DateOnly',
+            'time': 'TimeOnly',
+            'datetime': 'DateTimeOffset',
+            'timestamp': 'DateTimeOffset',
+            'duration': 'TimeSpan',
+            'uuid': 'Guid',
+            'uri': 'Uri',
+            'jsonpointer': 'string',
+            'any': 'object'
+        }
+        qualified_class_name = 'global::'+self.get_qualified_name(pascal(self.base_namespace), pascal(structure_type))
+        if qualified_class_name in self.generated_structure_types:
+            result = qualified_class_name
+        else:
+            result = mapping.get(structure_type, 'object')
+        return result
+
+    def is_csharp_reserved_word(self, word: str) -> bool:
+        """ Checks if a word is a reserved C# keyword """
+        reserved_words = [
+            'abstract', 'as', 'base', 'bool', 'break', 'byte', 'case', 'catch', 'char', 'checked', 'class', 'const',
+            'continue', 'decimal', 'default', 'delegate', 'do', 'double', 'else', 'enum', 'event', 'explicit', 'extern',
+            'false', 'finally', 'fixed', 'float', 'for', 'foreach', 'goto', 'if', 'implicit', 'in', 'int', 'interface',
+            'internal', 'is', 'lock', 'long', 'namespace', 'new', 'null', 'object', 'operator', 'out', 'override',
+            'params', 'private', 'protected', 'public', 'readonly', 'ref', 'return', 'sbyte', 'sealed', 'short', 'sizeof',
+            'stackalloc', 'static', 'string', 'struct', 'switch', 'this', 'throw', 'true', 'try', 'typeof', 'uint', 'ulong',
+            'unchecked', 'unsafe', 'ushort', 'using', 'virtual', 'void', 'volatile', 'while'
+        ]
+        return word in reserved_words
+
+    def is_csharp_primitive_type(self, csharp_type: str) -> bool:
+        """ Checks if a type is a C# primitive type """
+        if csharp_type.endswith('?'):
+            csharp_type = csharp_type[:-1]
+        return csharp_type in ['void', 'bool', 'sbyte', 'byte', 'short', 'ushort', 'int', 'uint', 'long', 'ulong', 
+                               'float', 'double', 'decimal', 'string', 'DateTime', 'DateTimeOffset', 'DateOnly', 
+                               'TimeOnly', 'TimeSpan', 'Guid', 'byte[]', 'object', 'System.Int128', 'System.UInt128', 'Uri']
+
+    def map_csharp_primitive_to_clr_type(self, cs_type: str) -> str:
+        """ Maps C# primitive types to CLR types"""
+        map = {
+            "int": "Int32",
+            "long": "Int64",
+            "float": "Single",
+            "double": "Double",
+            "decimal": "Decimal",
+            "short": "Int16",
+            "sbyte": "SByte",
+            "byte": "Byte",
+            "ushort": "UInt16",
+            "uint": "UInt32",
+            "ulong": "UInt64",
+            "bool": "Boolean",
+            "string": "String",
+            "Guid": "Guid"
+        }
+        return map.get(cs_type, cs_type)
+
+    def resolve_ref(self, ref: str, context_schema: Optional[Dict] = None) -> Optional[Dict]:
+        """ Resolves a $ref to the actual schema definition """
+        # Check if it's an absolute URI reference (schema with $id)
+        if not ref.startswith('#/'):
+            # Try to resolve from schema registry
+            if ref in self.schema_registry:
+                return self.schema_registry[ref]
+            return None
+        
+        # Handle fragment-only references (internal to document)
+        path = ref[2:].split('/')
+        schema = context_schema if context_schema else self.schema_doc
+        
+        for part in path:
+            if not isinstance(schema, dict) or part not in schema:
+                return None
+            schema = schema[part]
+        
+        return schema
+    
+    def validate_abstract_ref(self, ref_schema: Dict, ref: str, is_extends_context: bool = False) -> None:
+        """
+        Validates that abstract types are only referenced in $extends context.
+        Per JSON Structure Core Spec Section 3.10.1, abstract types cannot be 
+        directly instantiated and should only be referenced via $extends.
+        
+        Args:
+            ref_schema: The resolved schema being referenced
+            ref: The $ref string for error reporting
+            is_extends_context: True if this reference is in a $extends context
+        """
+        import sys
+        is_abstract = ref_schema.get('abstract', False)
+        if is_abstract and not is_extends_context:
+            print(f"WARNING: Abstract type referenced outside $extends context: {ref}", file=sys.stderr)
+            print(f"  Abstract types cannot be directly instantiated. Use $extends to inherit from them.", file=sys.stderr)
+    
+    def register_schema_ids(self, schema: Dict, base_uri: str = '') -> None:
+        """ Recursively registers schemas with $id keywords """
+        if not isinstance(schema, dict):
+            return
+        
+        # Register this schema if it has an $id
+        if '$id' in schema:
+            schema_id = schema['$id']
+            # Handle relative URIs
+            if base_uri and not schema_id.startswith(('http://', 'https://', 'urn:')):
+                from urllib.parse import urljoin
+                schema_id = urljoin(base_uri, schema_id)
+            self.schema_registry[schema_id] = schema
+            base_uri = schema_id  # Update base URI for nested schemas
+        
+        # Recursively process definitions
+        if 'definitions' in schema:
+            for def_name, def_schema in schema['definitions'].items():
+                if isinstance(def_schema, dict):
+                    self.register_schema_ids(def_schema, base_uri)
+        
+        # Recursively process properties
+        if 'properties' in schema:
+            for prop_name, prop_schema in schema['properties'].items():
+                if isinstance(prop_schema, dict):
+                    self.register_schema_ids(prop_schema, base_uri)
+        
+        # Recursively process items, values, etc.
+        for key in ['items', 'values', 'additionalProperties']:
+            if key in schema and isinstance(schema[key], dict):
+                self.register_schema_ids(schema[key], base_uri)
+
+    def convert_structure_type_to_csharp(self, class_name: str, field_name: str, structure_type: JsonNode, parent_namespace: str) -> str:
+        """ Converts JSON Structure type to C# type """
+        if isinstance(structure_type, str):
+            return self.map_primitive_to_csharp(structure_type)
+        elif isinstance(structure_type, list):
+            # Handle type unions
+            non_null_types = [t for t in structure_type if t != 'null']
+            if len(non_null_types) == 1:
+                # Nullable type
+                return f"{self.convert_structure_type_to_csharp(class_name, field_name, non_null_types[0], parent_namespace)}?"
+            else:
+                return self.generate_embedded_union(class_name, field_name, non_null_types, parent_namespace, write_file=True)
+        elif isinstance(structure_type, dict):
+            # Handle $ref
+            if '$ref' in structure_type:
+                ref_schema = self.resolve_ref(structure_type['$ref'], self.schema_doc)
+                if ref_schema:
+                    # Validate abstract type usage (Section 3.10.1)
+                    # Abstract types should only be referenced via $extends
+                    self.validate_abstract_ref(ref_schema, structure_type['$ref'], is_extends_context=False)
+                    
+                    # Extract type name from the ref
+                    ref_path = structure_type['$ref'].split('/')
+                    type_name = ref_path[-1]
+                    ref_namespace = '.'.join(ref_path[2:-1]) if len(ref_path) > 3 else parent_namespace
+                    return self.generate_class_or_choice(ref_schema, ref_namespace, write_file=True, explicit_name=type_name)
+                return 'object'
+            
+            # Handle enum keyword - must be checked before 'type'
+            if 'enum' in structure_type:
+                return self.generate_enum(structure_type, field_name, parent_namespace, write_file=True)
+            
+            # Handle type keyword
+            if 'type' not in structure_type:
+                return 'object'
+            
+            struct_type = structure_type['type']
+            
+            # Handle complex types
+            if struct_type == 'object':
+                return self.generate_class(structure_type, parent_namespace, write_file=True)
+            elif struct_type == 'array':
+                items_type = self.convert_structure_type_to_csharp(class_name, field_name+'List', structure_type.get('items', {'type': 'any'}), parent_namespace)
+                return f"List<{items_type}>"
+            elif struct_type == 'set':
+                items_type = self.convert_structure_type_to_csharp(class_name, field_name+'Set', structure_type.get('items', {'type': 'any'}), parent_namespace)
+                return f"HashSet<{items_type}>"
+            elif struct_type == 'map':
+                values_type = self.convert_structure_type_to_csharp(class_name, field_name+'Map', structure_type.get('values', {'type': 'any'}), parent_namespace)
+                return f"Dictionary<string, {values_type}>"
+            elif struct_type == 'choice':
+                return self.generate_choice(structure_type, parent_namespace, write_file=True)
+            elif struct_type == 'tuple':
+                return self.generate_tuple(structure_type, parent_namespace, write_file=True)
+            else:
+                return self.convert_structure_type_to_csharp(class_name, field_name, struct_type, parent_namespace)
+        return 'object'
+
+    def generate_class_or_choice(self, structure_schema: Dict, parent_namespace: str, write_file: bool = True, explicit_name: str = '') -> str:
+        """ Generates a Class or Choice """
+        struct_type = structure_schema.get('type', 'object')
+        if struct_type == 'object':
+            return self.generate_class(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
+        elif struct_type == 'choice':
+            return self.generate_choice(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
+        elif struct_type == 'tuple':
+            return self.generate_tuple(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
+        elif struct_type in ('map', 'array', 'set'):
+            # Root-level container types: generate wrapper class with implicit conversions
+            return self.generate_container_wrapper(structure_schema, parent_namespace, write_file, explicit_name=explicit_name)
+        return 'object'
+
+    def generate_class(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
+        """ Generates a Class from JSON Structure object type """
+        class_definition = ''
+        
+        # Get name and namespace
+        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedClass'))
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        xml_namespace = structure_schema.get('xmlns', None)
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+
+        # Check if this is an abstract type (Section 3.10.1)
+        is_abstract = structure_schema.get('abstract', False)
+        
+        # Generate class documentation
+        doc = structure_schema.get('description', structure_schema.get('doc', class_name))
+        class_definition += f"/// <summary>\n/// {doc}\n/// </summary>\n"
+        
+        if is_abstract:
+            class_definition += f"/// <remarks>\n/// This is an abstract type and cannot be instantiated directly.\n/// </remarks>\n"
+
+        # Add Obsolete attribute if deprecated
+        if structure_schema.get('deprecated', False):
+            deprecated_msg = structure_schema.get('description', f'{class_name} is deprecated')
+            class_definition += f"[System.Obsolete(\"{deprecated_msg}\")]\n"
+
+        # Add XML serialization attribute for the class if enabled
+        if self.system_xml_annotation:
+            if xml_namespace:
+                class_definition += f"[XmlRoot(\"{class_name}\", Namespace=\"{xml_namespace}\")]\n"
+            else:
+                class_definition += f"[XmlRoot(\"{class_name}\")]\n"
+
+        # Generate properties
+        properties = structure_schema.get('properties', {})
+        required_props = structure_schema.get('required', [])
+        
+        # Handle alternative required sets
+        is_alternative_required = isinstance(required_props, list) and len(required_props) > 0 and isinstance(required_props[0], list)
+        
+        # Check additionalProperties setting (Section 3.7.8)
+        additional_props = structure_schema.get('additionalProperties', True if is_abstract else None)
+        
+        fields_str = []
+        for prop_name, prop_schema in properties.items():
+            field_def = self.generate_property(prop_name, prop_schema, class_name, schema_namespace, required_props)
+            fields_str.append(field_def)
+        
+        # Add dictionary for additional properties if needed
+        if additional_props is not False and additional_props is not None:
+            if isinstance(additional_props, dict):
+                # additionalProperties is a schema
+                value_type = self.convert_structure_type_to_csharp(class_name, 'additionalValue', additional_props, schema_namespace)
+                fields_str.append(f"{INDENT}/// <summary>\n{INDENT}/// Additional properties not defined in schema\n{INDENT}/// </summary>\n")
+                fields_str.append(f"{INDENT}public Dictionary<string, {value_type}>? AdditionalProperties {{ get; set; }}\n")
+            elif additional_props is True:
+                # Allow any additional properties
+                fields_str.append(f"{INDENT}/// <summary>\n{INDENT}/// Additional properties not defined in schema\n{INDENT}/// </summary>\n")
+                fields_str.append(f"{INDENT}public Dictionary<string, object>? AdditionalProperties {{ get; set; }}\n")
+        
+        class_body = "\n".join(fields_str)
+        
+        # Generate class declaration
+        abstract_modifier = "abstract " if is_abstract else ""
+        sealed_modifier = "sealed " if additional_props is False and not is_abstract else ""
+        
+        class_definition += f"public {abstract_modifier}{sealed_modifier}partial class {class_name}\n{{\n{class_body}"
+        
+        # Add default constructor (not for abstract classes with no concrete constructors)
+        if not is_abstract or properties:
+            class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Default constructor\n{INDENT}/// </summary>\n"
+            constructor_modifier = "protected" if is_abstract else "public"
+            class_definition += f"{INDENT}{constructor_modifier} {class_name}()\n{INDENT}{{\n{INDENT}}}"
+
+        # Add helper methods from template if any annotations are enabled
+        if self.system_text_json_annotation or self.newtonsoft_json_annotation or self.system_xml_annotation:
+            class_definition += process_template(
+                "structuretocsharp/dataclass_core.jinja",
+                class_name=class_name,
+                system_text_json_annotation=self.system_text_json_annotation,
+                newtonsoft_json_annotation=self.newtonsoft_json_annotation,
+                system_xml_annotation=self.system_xml_annotation
+            )
+
+        # Generate Equals and GetHashCode
+        class_definition += self.generate_equals_and_gethashcode(structure_schema, class_name, schema_namespace)
+
+        class_definition += "\n"+"}"
+
+        if write_file:
+            self.write_to_file(namespace, class_name, class_definition)
+
+        self.generated_types[ref] = "class"
+        self.generated_structure_types[ref] = structure_schema
+        return ref
+
+    def generate_property(self, prop_name: str, prop_schema: Dict, class_name: str, parent_namespace: str, required_props: List) -> str:
+        """ Generates a property for a class """
+        property_definition = ''
+        
+        # Resolve property name
+        field_name = prop_name
+        if self.is_csharp_reserved_word(field_name):
+            field_name = f"@{field_name}"
+        if self.pascal_properties:
+            field_name_cs = pascal(field_name)
+        else:
+            field_name_cs = field_name
+        if field_name_cs == class_name:
+            field_name_cs += "_"
+        
+        # Check if this is a const field
+        if 'const' in prop_schema:
+            const_value = prop_schema['const']
+            prop_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
+            
+            # Remove nullable marker for const
+            if prop_type.endswith('?'):
+                prop_type = prop_type[:-1]
+            
+            # Generate documentation
+            doc = prop_schema.get('description', prop_schema.get('doc', field_name_cs))
+            property_definition += f"{INDENT}/// <summary>\n{INDENT}/// {doc}\n{INDENT}/// </summary>\n"
+            
+            # Add JSON property name annotation
+            if self.system_text_json_annotation and field_name != field_name_cs:
+                property_definition += f'{INDENT}[System.Text.Json.Serialization.JsonPropertyName("{prop_name}")]\n'
+            if self.newtonsoft_json_annotation and field_name != field_name_cs:
+                property_definition += f'{INDENT}[Newtonsoft.Json.JsonProperty("{prop_name}")]\n'
+            
+            # Add XML element annotation if enabled
+            if self.system_xml_annotation:
+                property_definition += f'{INDENT}[System.Xml.Serialization.XmlElement("{prop_name}")]\n'
+            
+            # Generate const field
+            const_val = self.format_default_value(const_value, prop_type)
+            property_definition += f"{INDENT}public const {prop_type} {field_name_cs} = {const_val};\n"
+            
+            return property_definition
+        
+        # Determine if required
+        is_required = prop_name in required_props if not isinstance(required_props, list) or len(required_props) == 0 or not isinstance(required_props[0], list) else any(prop_name in req_set for req_set in required_props)
+        
+        # Get property type
+        prop_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
+        
+        # Add nullable marker if not required and not already nullable
+        if not is_required and not prop_type.endswith('?') and not prop_type.startswith('List<') and not prop_type.startswith('HashSet<') and not prop_type.startswith('Dictionary<'):
+            prop_type += '?'
+        
+        # Generate documentation
+        doc = prop_schema.get('description', prop_schema.get('doc', field_name_cs))
+        property_definition += f"{INDENT}/// <summary>\n{INDENT}/// {doc}\n{INDENT}/// </summary>\n"
+        
+        # Add JSON property name annotation
+        if self.system_text_json_annotation and field_name != field_name_cs:
+            property_definition += f'{INDENT}[System.Text.Json.Serialization.JsonPropertyName("{prop_name}")]\n'
+        if self.newtonsoft_json_annotation and field_name != field_name_cs:
+            property_definition += f'{INDENT}[Newtonsoft.Json.JsonProperty("{prop_name}")]\n'
+        
+        # Add XML element annotation if enabled
+        if self.system_xml_annotation:
+            property_definition += f'{INDENT}[System.Xml.Serialization.XmlElement("{prop_name}")]\n'
+        
+        # Add validation attributes based on schema constraints
+        # StringLength attribute for maxLength
+        if 'maxLength' in prop_schema:
+            max_length = prop_schema['maxLength']
+            if 'minLength' in prop_schema:
+                min_length = prop_schema['minLength']
+                property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.StringLength({max_length}, MinimumLength = {min_length})]\n'
+            else:
+                property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.StringLength({max_length})]\n'
+        elif 'minLength' in prop_schema:
+            # MinLength only (no max)
+            min_length = prop_schema['minLength']
+            property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.MinLength({min_length})]\n'
+        
+        # RegularExpression attribute for pattern
+        if 'pattern' in prop_schema:
+            pattern = prop_schema['pattern'].replace('\\', '\\\\').replace('"', '\\"')
+            property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.RegularExpression(@"{pattern}")]\n'
+        
+        # Range attribute for minimum/maximum on numeric types
+        if 'minimum' in prop_schema or 'maximum' in prop_schema:
+            min_val = prop_schema.get('minimum', prop_schema.get('exclusiveMinimum', 'double.MinValue'))
+            max_val = prop_schema.get('maximum', prop_schema.get('exclusiveMaximum', 'double.MaxValue'))
+            
+            # Convert to appropriate format
+            if isinstance(min_val, (int, float)):
+                min_str = str(min_val)
+            else:
+                min_str = str(min_val)
+            
+            if isinstance(max_val, (int, float)):
+                max_str = str(max_val)
+            else:
+                max_str = str(max_val)
+            
+            property_definition += f'{INDENT}[System.ComponentModel.DataAnnotations.Range({min_str}, {max_str})]\n'
+        
+        # Add Obsolete attribute if deprecated
+        if prop_schema.get('deprecated', False):
+            deprecated_msg = prop_schema.get('description', f'{prop_name} is deprecated')
+            property_definition += f'{INDENT}[System.Obsolete("{deprecated_msg}")]\n'
+        
+        # Generate property with required modifier if needed
+        required_modifier = "required " if is_required and not prop_type.endswith('?') else ""
+        
+        # Handle readOnly and writeOnly
+        is_read_only = prop_schema.get('readOnly', False)
+        is_write_only = prop_schema.get('writeOnly', False)
+        
+        if is_read_only:
+            # readOnly: private or init-only setter
+            property_definition += f"{INDENT}public {required_modifier}{prop_type} {field_name_cs} {{ get; init; }}"
+        elif is_write_only:
+            # writeOnly: private getter
+            property_definition += f"{INDENT}public {required_modifier}{prop_type} {field_name_cs} {{ private get; set; }}"
+        else:
+            # Normal property
+            property_definition += f"{INDENT}public {required_modifier}{prop_type} {field_name_cs} {{ get; set; }}"
+        
+        # Add default value if present
+        if 'default' in prop_schema:
+            default_val = self.format_default_value(prop_schema['default'], prop_type)
+            property_definition += f" = {default_val};\n"
+        else:
+            property_definition += "\n"
+        
+        return property_definition
+
+    def format_default_value(self, value: Any, csharp_type: str) -> str:
+        """ Formats a default value for C# """
+        if value is None:
+            return "null"
+        elif isinstance(value, bool):
+            return "true" if value else "false"
+        elif isinstance(value, str):
+            return f'"{value}"'
+        elif isinstance(value, (int, float)):
+            return str(value)
+        elif isinstance(value, list):
+            return f"new {csharp_type}()"
+        elif isinstance(value, dict):
+            return f"new {csharp_type}()"
+        return f"default({csharp_type})"
+
+    def generate_enum(self, structure_schema: Dict, field_name: str, parent_namespace: str, write_file: bool) -> str:
+        """ Generates a C# enum from JSON Structure enum keyword """
+        enum_values = structure_schema.get('enum', [])
+        if not enum_values:
+            return 'object'
+        
+        # Determine enum name from field name
+        enum_name = pascal(field_name) + 'Enum' if field_name else 'UnnamedEnum'
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, enum_name)
+        if ref in self.generated_types:
+            return ref
+        
+        # Determine underlying type
+        base_type = structure_schema.get('type', 'string')
+        
+        # For string enums, we don't specify an underlying type
+        # For numeric enums, we map the type
+        numeric_types = ['int8', 'uint8', 'int16', 'uint16', 'int32', 'uint32', 'int64', 'uint64']
+        is_numeric = base_type in numeric_types
+        
+        enum_definition = ''
+        doc = structure_schema.get('description', structure_schema.get('doc', enum_name))
+        enum_definition += f"/// <summary>\n/// {doc}\n/// </summary>\n"
+        
+        # Add Obsolete attribute if deprecated
+        if structure_schema.get('deprecated', False):
+            deprecated_msg = structure_schema.get('description', f'{enum_name} is deprecated')
+            enum_definition += f"[System.Obsolete(\"{deprecated_msg}\")]\n"
+        
+        if is_numeric:
+            cs_base_type = self.map_primitive_to_csharp(base_type)
+            enum_definition += f"public enum {enum_name} : {cs_base_type}\n{{\n"
+        else:
+            # String enum - for System.Text.Json, use JsonConverter with JsonStringEnumConverter
+            if self.system_text_json_annotation:
+                enum_definition += f"[System.Text.Json.Serialization.JsonConverter(typeof(System.Text.Json.Serialization.JsonStringEnumConverter))]\n"
+            enum_definition += f"public enum {enum_name}\n{{\n"
+        
+        # Generate enum members
+        for i, value in enumerate(enum_values):
+            if is_numeric:
+                # Numeric enum - use the value directly
+                member_name = f"Value{value}"  # Prefix with "Value" since enum members can't start with numbers
+                enum_definition += f"{INDENT}{member_name} = {value}"
+            else:
+                # String enum - create member from the string
+                member_name = pascal(str(value).replace('-', '_').replace(' ', '_'))
+                enum_definition += f"{INDENT}{member_name}"
+            
+            if i < len(enum_values) - 1:
+                enum_definition += ",\n"
+            else:
+                enum_definition += "\n"
+        
+        enum_definition += "}"
+        
+        if write_file:
+            self.write_to_file(namespace, enum_name, enum_definition)
+        
+        self.generated_types[ref] = "enum"
+        self.generated_structure_types[ref] = structure_schema
+        return ref
+
+    def generate_choice(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
+        """ Generates a discriminated union (choice) type """
+        # Choice types in JSON Structure can be:
+        # 1. Tagged unions - single property with the choice type as key
+        # 2. Inline unions - with $extends and selector
+        
+        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedChoice'))
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+        
+        choices = structure_schema.get('choices', {})
+        selector = structure_schema.get('selector')
+        extends = structure_schema.get('$extends')
+        
+        if extends and selector:
+            # Inline union - generate as inheritance hierarchy
+            return self.generate_inline_union(structure_schema, parent_namespace, write_file, explicit_name)
+        else:
+            # Tagged union - generate as a union class similar to Avro
+            return self.generate_tagged_union(structure_schema, parent_namespace, write_file, explicit_name)
+
+    def generate_tagged_union(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
+        """ Generates a tagged union type """
+        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedChoice'))
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+        
+        choices = structure_schema.get('choices', {})
+        choice_types = []
+        
+        for choice_name, choice_schema in choices.items():
+            choice_type = self.convert_structure_type_to_csharp(class_name, choice_name, choice_schema, schema_namespace)
+            choice_types.append((choice_name, choice_type))
+        
+        # Generate the union class similar to Avro unions
+        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name)}\n/// </summary>\n"
+        class_definition += f"public partial class {class_name}\n{{\n"
+        
+        # Generate properties for each choice
+        for choice_name, choice_type in choice_types:
+            prop_name = pascal(choice_name)
+            class_definition += f"{INDENT}/// <summary>\n{INDENT}/// Gets or sets the {prop_name} value\n{INDENT}/// </summary>\n"
+            class_definition += f"{INDENT}public {choice_type}? {prop_name} {{ get; set; }} = null;\n"
+        
+        # Add constructor
+        class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Default constructor\n{INDENT}/// </summary>\n"
+        class_definition += f"{INDENT}public {class_name}()\n{INDENT}{{\n{INDENT}}}\n"
+        
+        # Add constructors for each choice
+        for choice_name, choice_type in choice_types:
+            prop_name = pascal(choice_name)
+            class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Constructor for {prop_name} values\n{INDENT}/// </summary>\n"
+            class_definition += f"{INDENT}public {class_name}({choice_type} {prop_name.lower()})\n{INDENT}{{\n"
+            class_definition += f"{INDENT*2}this.{prop_name} = {prop_name.lower()};\n{INDENT}}}\n"
+        
+        # Generate Equals and GetHashCode for choice types
+        class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Determines whether the specified object is equal to the current object.\n{INDENT}/// </summary>\n"
+        class_definition += f"{INDENT}public override bool Equals(object? obj)\n{INDENT}{{\n"
+        class_definition += f"{INDENT*2}if (obj is not {class_name} other) return false;\n"
+        
+        # Compare each choice property
+        equality_checks = []
+        for choice_name, choice_type in choice_types:
+            prop_name = pascal(choice_name)
+            equality_checks.append(f"Equals(this.{prop_name}, other.{prop_name})")
+        
+        if len(equality_checks) == 1:
+            class_definition += f"{INDENT*2}return {equality_checks[0]};\n"
+        else:
+            class_definition += f"{INDENT*2}return " + f"\n{INDENT*3}&& ".join(equality_checks) + ";\n"
+        
+        class_definition += f"{INDENT}}}\n\n"
+        
+        # Generate GetHashCode
+        class_definition += f"{INDENT}/// <summary>\n{INDENT}/// Serves as the default hash function.\n{INDENT}/// </summary>\n"
+        class_definition += f"{INDENT}public override int GetHashCode()\n{INDENT}{{\n"
+        
+        if len(choice_types) <= 8:
+            hash_fields = [f"this.{pascal(choice_name)}" for choice_name, _ in choice_types]
+            class_definition += f"{INDENT*2}return HashCode.Combine({', '.join(hash_fields)});\n"
+        else:
+            class_definition += f"{INDENT*2}var hash = new HashCode();\n"
+            for choice_name, _ in choice_types:
+                prop_name = pascal(choice_name)
+                class_definition += f"{INDENT*2}hash.Add(this.{prop_name});\n"
+            class_definition += f"{INDENT*2}return hash.ToHashCode();\n"
+        
+        class_definition += f"{INDENT}}}\n"
+        
+        class_definition += "}"
+        
+        if write_file:
+            self.write_to_file(namespace, class_name, class_definition)
+        
+        self.generated_types[ref] = "choice"
+        self.generated_structure_types[ref] = structure_schema
+        return ref
+
+    def generate_inline_union(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
+        """ Generates an inline union type with inheritance """
+        # For inline unions, we generate an abstract base class and derived classes
+        # The selector property indicates which derived class is being used
+        
+        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedChoice'))
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+        
+        # Get base class from $extends
+        extends_ref = structure_schema.get('$extends', '')
+        if extends_ref and isinstance(extends_ref, str):
+            base_schema = self.resolve_ref(extends_ref, self.schema_doc)
+            if not base_schema:
+                # Try resolving relative to the structure_schema itself
+                base_schema = self.resolve_ref(extends_ref, structure_schema)
+            
+            # Validate abstract type usage - $extends is ALLOWED to reference abstract types
+            if base_schema:
+                self.validate_abstract_ref(base_schema, extends_ref, is_extends_context=True)
+        else:
+            base_schema = None
+        
+        if not base_schema:
+            # Fallback to tagged union if no base
+            return self.generate_tagged_union(structure_schema, parent_namespace, write_file, explicit_name)
+        
+        # First, ensure base class is generated (if it's abstract, it won't be referenced directly)
+        base_schema_copy = base_schema.copy()
+        if 'name' not in base_schema_copy:
+            # Extract name from $extends ref
+            base_name = extends_ref.split('/')[-1]
+            base_schema_copy['name'] = base_name
+        base_class_ref = self.generate_class(base_schema_copy, schema_namespace, write_file)
+        base_class_name = base_class_ref.split('::')[-1].split('.')[-1]
+        
+        choices = structure_schema.get('choices', {})
+        selector = structure_schema.get('selector', 'type')
+        
+        # Generate abstract base class with selector property
+        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name + ' (inline union base)')}\n/// </summary>\n"
+        
+        if self.system_text_json_annotation:
+            class_definition += f'[System.Text.Json.Serialization.JsonPolymorphic(TypeDiscriminatorPropertyName = "{selector}")]\n'
+            for choice_name in choices.keys():
+                derived_class_name = pascal(choice_name)
+                class_definition += f'[System.Text.Json.Serialization.JsonDerivedType(typeof({derived_class_name}), "{choice_name}")]\n'
+        
+        class_definition += f"public abstract partial class {class_name}"
+        
+        # Inherit from base class if it exists
+        if base_class_name and base_class_name != class_name:
+            class_definition += f" : {base_class_name}"
+        
+        class_definition += "\n{\n"
+        
+        # Add selector property (not required since derived classes set it in constructor)
+        class_definition += f"{INDENT}/// <summary>\n{INDENT}/// Type discriminator\n{INDENT}/// </summary>\n"
+        if self.system_text_json_annotation:
+            class_definition += f'{INDENT}[System.Text.Json.Serialization.JsonPropertyName("{selector}")]\n'
+        
+        # Check if selector is already in base properties
+        base_has_selector = selector in base_schema.get('properties', {})
+        if base_has_selector:
+            class_definition += f"{INDENT}public new string {pascal(selector)} {{ get; set; }} = \"\";\n"
+        else:
+            class_definition += f"{INDENT}public string {pascal(selector)} {{ get; set; }} = \"\";\n"
+        
+        class_definition += "}"
+        
+        if write_file:
+            self.write_to_file(namespace, class_name, class_definition)
+        
+        # Generate derived classes for each choice with property merging
+        for choice_name, choice_schema_ref in choices.items():
+            # Resolve the choice schema
+            if isinstance(choice_schema_ref, dict) and '$ref' in choice_schema_ref:
+                choice_schema = self.resolve_ref(choice_schema_ref['$ref'], self.schema_doc)
+                if not choice_schema:
+                    # Try resolving relative to the structure_schema itself
+                    choice_schema = self.resolve_ref(choice_schema_ref['$ref'], structure_schema)
+            else:
+                choice_schema = choice_schema_ref
+            
+            if not choice_schema or not isinstance(choice_schema, dict):
+                continue
+            
+            # Mark this choice as generated to prevent duplicate generation in process_definitions
+            derived_class_name = pascal(choice_name)
+            derived_namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+            derived_ref = 'global::'+self.get_qualified_name(derived_namespace, derived_class_name)
+            
+            # Only generate if not already generated
+            if derived_ref in self.generated_types:
+                continue
+            
+            # Merge properties from base schema into choice schema
+            merged_schema = self.merge_inherited_properties(choice_schema, base_schema, class_name)
+            merged_schema['name'] = choice_name
+            merged_schema['namespace'] = schema_namespace
+            
+            # Mark that this extends the union base
+            merged_schema['$extends_inline_union'] = class_name
+            
+            # Generate the derived class
+            self.generate_derived_class(merged_schema, class_name, choice_name, selector, schema_namespace, write_file)
+        
+        self.generated_types[ref] = "choice"
+        self.generated_structure_types[ref] = structure_schema
+        return ref
+    
+    def merge_inherited_properties(self, derived_schema: Dict, base_schema: Dict, union_class_name: str) -> Dict:
+        """ Merges properties from base schema into derived schema """
+        merged = derived_schema.copy()
+        
+        # Get properties from both schemas
+        base_props = base_schema.get('properties', {})
+        derived_props = merged.get('properties', {})
+        
+        # Track which properties come from base (for filtering during generation)
+        base_property_names = list(base_props.keys())
+        
+        # Merge properties (derived overrides base)
+        merged_props = {}
+        merged_props.update(base_props)
+        merged_props.update(derived_props)
+        merged['properties'] = merged_props
+        
+        # Store base property names so we can skip them during code generation
+        merged['$base_properties'] = base_property_names
+        
+        # Merge required fields
+        base_required = base_schema.get('required', [])
+        derived_required = merged.get('required', [])
+        if isinstance(base_required, list) and isinstance(derived_required, list):
+            # Combine and deduplicate
+            merged['required'] = list(set(base_required + derived_required))
+        
+        return merged
+    
+    def generate_derived_class(self, schema: Dict, base_class_name: str, choice_name: str, selector: str, parent_namespace: str, write_file: bool) -> str:
+        """ Generates a derived class for inline union """
+        class_name = pascal(choice_name)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema.get('namespace', parent_namespace)))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+        
+        # Generate class with inheritance
+        doc = schema.get('description', f'{class_name} - {choice_name} variant')
+        class_definition = f"/// <summary>\n/// {doc}\n/// </summary>\n"
+        
+        class_definition += f"public partial class {class_name} : {base_class_name}\n{{\n"
+        
+        # Generate properties (only the derived-specific ones, base properties are inherited)
+        properties = schema.get('properties', {})
+        required_props = schema.get('required', [])
+        
+        # Get base class properties to filter them out
+        # We need to find the base schema to know what properties to exclude
+        # For now, we'll generate all properties from merged schema
+        # NOTE: BaseAddress properties come through the merged schema but shouldn't be redeclared
+        #       We need to identify which properties are from the BASE vs which are NEW
+        
+        # Get the original (non-merged) choice schema to determine NEW properties
+        # The merged schema has all properties; we want only the ones NOT in base
+        # Since we don't have access to the original choice schema here, we'll use a heuristic:
+        # Properties marked with a special flag during merging
+        
+        # Alternative approach: Only generate properties that are NOT in the InlineChoice base
+        # But InlineChoice only has the selector, not the BaseAddress properties
+        # So we need to look further up the chain
+        
+        # SIMPLEST SOLUTION: Filter out properties that come from the extended base
+        # We can detect this by checking if the property exists in the base_schema context
+        # But we don't have base_schema in this method
+        
+        # FOR NOW: Generate all properties but mark inherited ones with 'new'
+        # Actually, C# doesn't allow 'new required' - that's the error
+        # So we MUST skip inherited properties entirely
+        
+        # The merged schema has ALL properties. We need to skip base properties.
+        # The base properties are those NOT in the original choice schema
+        # We need to pass the original choice schema properties to know what to generate
+        
+        # CORRECT FIX: Only generate properties from the ORIGINAL choice schema, not merged
+        # But wait - we're passing merged_schema which has all properties
+        # We need to differentiate
+        
+        # Let's add a marker during merge to track which properties are from base
+        # Or better: pass BOTH original and merged schemas
+        
+        # QUICK FIX: Check if selector property to skip, and skip properties that were
+        # in the base by checking schema metadata
+        
+        # Since schema has '$extends_inline_union', we can use that
+        # But we need the ORIGINAL choice properties, not merged
+        
+        # The issue is we're generating from merged_schema which has ALL properties
+        # We need to know which properties are NEW (from choice) vs inherited (from base)
+        
+        # SOLUTION: Don't generate inherited properties - but how to identify them?
+        # We could store in merged_schema a list of base property names
+        
+        # Let me fix this by adding a key to mark base properties
+        base_properties = schema.get('$base_properties', [])
+        
+        for prop_name, prop_schema in properties.items():
+            # Skip selector - it's defined in base as required
+            if prop_name == selector:
+                continue
+            # Skip properties inherited from base schema
+            if prop_name in base_properties:
+                continue
+            field_def = self.generate_property(prop_name, prop_schema, class_name, parent_namespace, required_props)
+            class_definition += field_def
+        
+        # Add constructor that sets the discriminator
+        class_definition += f"\n{INDENT}/// <summary>\n{INDENT}/// Constructor that sets the discriminator value\n{INDENT}/// </summary>\n"
+        class_definition += f"{INDENT}public {class_name}()\n{INDENT}{{\n"
+        class_definition += f"{INDENT*2}this.{pascal(selector)} = \"{choice_name}\";\n"
+        class_definition += f"{INDENT}}}\n"
+        
+        # Generate Equals and GetHashCode
+        class_definition += self.generate_equals_and_gethashcode(schema, class_name, parent_namespace)
+        
+        class_definition += "}"
+        
+        if write_file:
+            self.write_to_file(namespace, class_name, class_definition)
+        
+        self.generated_types[ref] = "class"
+        self.generated_structure_types[ref] = schema
+        return ref
+
+    def generate_tuple(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
+        """ Generates a tuple type - Per JSON Structure spec, tuples serialize as JSON arrays, not objects """
+        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', 'UnnamedTuple'))
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+        
+        properties = structure_schema.get('properties', {})
+        tuple_order = structure_schema.get('tuple', [])
+        
+        # Build list of tuple element types and names in correct order
+        tuple_elements = []
+        for prop_name in tuple_order:
+            if prop_name in properties:
+                prop_schema = properties[prop_name]
+                prop_type = self.convert_structure_type_to_csharp(class_name, prop_name, prop_schema, schema_namespace)
+                field_name = pascal(prop_name) if self.pascal_properties else prop_name
+                tuple_elements.append((prop_type, field_name))
+        
+        # Generate as a C# record struct with positional parameters
+        # Per JSON Structure spec: tuples serialize as JSON arrays like ["Alice", 42]
+        tuple_signature = ', '.join([f"{elem_type} {elem_name}" for elem_type, elem_name in tuple_elements])
+        
+        # Create the tuple record struct
+        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name)}\n/// </summary>\n"
+        class_definition += f"/// <remarks>\n/// JSON Structure tuple type - serializes as JSON array: [{', '.join(['...' for _ in tuple_elements])}]\n/// </remarks>\n"
+        
+        # Add JsonConverter attribute if System.Text.Json annotations are enabled
+        if self.system_text_json_annotation:
+            class_definition += f"[System.Text.Json.Serialization.JsonConverter(typeof(TupleJsonConverter<{class_name}>))]\n"
+        
+        class_definition += f"public record struct {class_name}({tuple_signature});\n"
+        
+        if write_file:
+            self.write_to_file(namespace, class_name, class_definition)
+        
+        self.generated_types[ref] = "tuple"
+        self.generated_structure_types[ref] = structure_schema
+        return ref
+
+    def generate_container_wrapper(self, structure_schema: Dict, parent_namespace: str, write_file: bool, explicit_name: str = '') -> str:
+        """ Generates a wrapper class for root-level container types (map, array, set) """
+        struct_type = structure_schema.get('type', 'map')
+        class_name = pascal(explicit_name if explicit_name else structure_schema.get('name', f'Root{struct_type.capitalize()}'))
+        schema_namespace = structure_schema.get('namespace', parent_namespace)
+        namespace = pascal(self.concat_namespace(self.base_namespace, schema_namespace))
+        
+        ref = 'global::'+self.get_qualified_name(namespace, class_name)
+        if ref in self.generated_types:
+            return ref
+        
+        # Determine the underlying collection type
+        value_type = "string"  # Default
+        item_type = "string"   # Default
+        underlying_type = "object"
+        
+        if struct_type == 'map':
+            values_schema = structure_schema.get('values', {'type': 'string'})
+            value_type = self.convert_structure_type_to_csharp(class_name, 'value', values_schema, schema_namespace)
+            underlying_type = f"Dictionary<string, {value_type}>"
+        elif struct_type == 'array':
+            items_schema = structure_schema.get('items', {'type': 'string'})
+            item_type = self.convert_structure_type_to_csharp(class_name, 'item', items_schema, schema_namespace)
+            underlying_type = f"List<{item_type}>"
+        elif struct_type == 'set':
+            items_schema = structure_schema.get('items', {'type': 'string'})
+            item_type = self.convert_structure_type_to_csharp(class_name, 'item', items_schema, schema_namespace)
+            underlying_type = f"HashSet<{item_type}>"
+        
+        # Generate wrapper class with implicit conversions
+        class_definition = f"/// <summary>\n/// {structure_schema.get('description', class_name)}\n/// </summary>\n"
+        class_definition += f"/// <remarks>\n/// Wrapper for root-level {struct_type} type\n/// </remarks>\n"
+        class_definition += f"public class {class_name}\n{{\n"
+        class_definition += f"{INDENT}private {underlying_type} _value = new();\n\n"
+        
+        # Add indexer or collection access
+        if struct_type == 'map':
+            class_definition += f"{INDENT}public {value_type} this[string key]\n"
+            class_definition += f"{INDENT}{{\n"
+            class_definition += f"{INDENT*2}get => _value[key];\n"
+            class_definition += f"{INDENT*2}set => _value[key] = value;\n"
+            class_definition += f"{INDENT}}}\n\n"
+        elif struct_type in ('array', 'set'):
+            class_definition += f"{INDENT}public {item_type} this[int index]\n"
+            class_definition += f"{INDENT}{{\n"
+            if struct_type == 'array':
+                class_definition += f"{INDENT*2}get => _value[index];\n"
+                class_definition += f"{INDENT*2}set => _value[index] = value;\n"
+            else:  # set
+                class_definition += f"{INDENT*2}get => _value.ElementAt(index);\n"
+                class_definition += f"{INDENT*2}set => throw new NotSupportedException(\"Cannot set items by index in a HashSet\");\n"
+            class_definition += f"{INDENT}}}\n\n"
+        
+        # Add Count property
+        class_definition += f"{INDENT}public int Count => _value.Count;\n\n"
+        
+        # Add Add method for collections
+        if struct_type == 'map':
+            class_definition += f"{INDENT}public void Add(string key, {value_type} value) => _value.Add(key, value);\n\n"
+        elif struct_type in ('array', 'set'):
+            class_definition += f"{INDENT}public void Add({item_type} item) => _value.Add(item);\n\n"
+        
+        # Override Equals and GetHashCode for proper value equality
+        class_definition += f"{INDENT}public override bool Equals(object? obj)\n"
+        class_definition += f"{INDENT}{{\n"
+        class_definition += f"{INDENT*2}if (obj is not {class_name} other) return false;\n"
+        if struct_type == 'map':
+            class_definition += f"{INDENT*2}if (_value.Count != other._value.Count) return false;\n"
+            class_definition += f"{INDENT*2}foreach (var kvp in _value)\n"
+            class_definition += f"{INDENT*2}{{\n"
+            class_definition += f"{INDENT*3}if (!other._value.TryGetValue(kvp.Key, out var otherValue) || !Equals(kvp.Value, otherValue))\n"
+            class_definition += f"{INDENT*4}return false;\n"
+            class_definition += f"{INDENT*2}}}\n"
+            class_definition += f"{INDENT*2}return true;\n"
+        elif struct_type == 'array':
+            class_definition += f"{INDENT*2}return _value.SequenceEqual(other._value);\n"
+        else:  # set
+            class_definition += f"{INDENT*2}return _value.SetEquals(other._value);\n"
+        class_definition += f"{INDENT}}}\n\n"
+        
+        class_definition += f"{INDENT}public override int GetHashCode()\n"
+        class_definition += f"{INDENT}{{\n"
+        class_definition += f"{INDENT*2}var hash = new HashCode();\n"
+        if struct_type == 'map':
+            class_definition += f"{INDENT*2}foreach (var kvp in _value)\n"
+            class_definition += f"{INDENT*2}{{\n"
+            class_definition += f"{INDENT*3}hash.Add(kvp.Key);\n"
+            class_definition += f"{INDENT*3}hash.Add(kvp.Value);\n"
+            class_definition += f"{INDENT*2}}}\n"
+        else:  # array or set
+            class_definition += f"{INDENT*2}foreach (var item in _value)\n"
+            class_definition += f"{INDENT*2}{{\n"
+            class_definition += f"{INDENT*3}hash.Add(item);\n"
+            class_definition += f"{INDENT*2}}}\n"
+        class_definition += f"{INDENT*2}return hash.ToHashCode();\n"
+        class_definition += f"{INDENT}}}\n\n"
+        
+        # Implicit conversion to underlying type
+        class_definition += f"{INDENT}public static implicit operator {underlying_type}({class_name} wrapper) => wrapper._value;\n\n"
+        
+        # Implicit conversion from underlying type
+        class_definition += f"{INDENT}public static implicit operator {class_name}({underlying_type} value) => new() {{ _value = value }};\n"
+        
+        class_definition += "}\n"
+        
+        if write_file:
+            self.write_to_file(namespace, class_name, class_definition)
+        
+        self.generated_types[ref] = "class"
+        self.generated_structure_types[ref] = structure_schema
+        return ref
+
+
+    def generate_embedded_union(self, class_name: str, field_name: str, structure_types: List, parent_namespace: str, write_file: bool) -> str:
+        """ Generates an embedded Union Class """
+        # Similar to Avro's union handling, but for JSON Structure types
+        union_class_name = pascal(field_name)+'Union'
+        ref = class_name+'.'+union_class_name
+        
+        # For simplicity, generate as object type
+        # A complete implementation would generate a proper union class
+        return 'object'
+
+    def generate_equals_and_gethashcode(self, structure_schema: Dict, class_name: str, parent_namespace: str) -> str:
+        """ Generates Equals and GetHashCode methods for value equality """
+        code = "\n"
+        properties = structure_schema.get('properties', {})
+        
+        # Filter out const properties since they're static and same for all instances
+        non_const_properties = {k: v for k, v in properties.items() if 'const' not in v}
+        
+        if not non_const_properties:
+            # Empty class or only const fields - simple implementation
+            code += f"{INDENT}/// <summary>\n{INDENT}/// Determines whether the specified object is equal to the current object.\n{INDENT}/// </summary>\n"
+            code += f"{INDENT}public override bool Equals(object? obj)\n{INDENT}{{\n"
+            code += f"{INDENT*2}return obj is {class_name};\n"
+            code += f"{INDENT}}}\n\n"
+            code += f"{INDENT}/// <summary>\n{INDENT}/// Serves as the default hash function.\n{INDENT}/// </summary>\n"
+            code += f"{INDENT}public override int GetHashCode()\n{INDENT}{{\n"
+            code += f"{INDENT*2}return 0;\n"
+            code += f"{INDENT}}}\n"
+            return code
+        
+        # Generate Equals method
+        code += f"{INDENT}/// <summary>\n{INDENT}/// Determines whether the specified object is equal to the current object.\n{INDENT}/// </summary>\n"
+        code += f"{INDENT}public override bool Equals(object? obj)\n{INDENT}{{\n"
+        code += f"{INDENT*2}if (obj is not {class_name} other) return false;\n"
+        
+        # Build equality comparisons for each non-const property
+        equality_checks = []
+        for prop_name, prop_schema in non_const_properties.items():
+            field_name = prop_name
+            if self.is_csharp_reserved_word(field_name):
+                field_name = f"@{field_name}"
+            if self.pascal_properties:
+                field_name = pascal(field_name)
+            if field_name == class_name:
+                field_name += "_"
+            
+            field_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
+            
+            # Handle different types of comparisons
+            if field_type == 'byte[]' or field_type == 'byte[]?':
+                # Byte arrays need special handling
+                equality_checks.append(f"System.Linq.Enumerable.SequenceEqual(this.{field_name} ?? Array.Empty<byte>(), other.{field_name} ?? Array.Empty<byte>())")
+            elif field_type.startswith('Dictionary<'):
+                # Dictionaries need special comparison - compare keys and values
+                if field_type.endswith('?'):
+                    dict_compare = f"((this.{field_name} == null && other.{field_name} == null) || (this.{field_name} != null && other.{field_name} != null && this.{field_name}.Count == other.{field_name}.Count && this.{field_name}.All(kvp => other.{field_name}.TryGetValue(kvp.Key, out var val) && Equals(kvp.Value, val))))"
+                    equality_checks.append(dict_compare)
+                else:
+                    dict_compare = f"(this.{field_name}.Count == other.{field_name}.Count && this.{field_name}.All(kvp => other.{field_name}.TryGetValue(kvp.Key, out var val) && Equals(kvp.Value, val)))"
+                    equality_checks.append(dict_compare)
+            elif field_type.startswith('List<') or field_type.startswith('HashSet<'):
+                # Lists and HashSets need sequence comparison
+                if field_type.endswith('?'):
+                    equality_checks.append(f"((this.{field_name} == null && other.{field_name} == null) || (this.{field_name} != null && other.{field_name} != null && this.{field_name}.SequenceEqual(other.{field_name})))")
+                else:
+                    equality_checks.append(f"this.{field_name}.SequenceEqual(other.{field_name})")
+            else:
+                # Use Equals for reference types, == for value types
+                if field_type.endswith('?') or not self.is_csharp_primitive_type(field_type):
+                    equality_checks.append(f"Equals(this.{field_name}, other.{field_name})")
+                else:
+                    equality_checks.append(f"this.{field_name} == other.{field_name}")
+        
+        # Join all checks with &&
+        if len(equality_checks) == 1:
+            code += f"{INDENT*2}return {equality_checks[0]};\n"
+        else:
+            code += f"{INDENT*2}return " + f"\n{INDENT*3}&& ".join(equality_checks) + ";\n"
+        
+        code += f"{INDENT}}}\n\n"
+        
+        # Generate GetHashCode method
+        code += f"{INDENT}/// <summary>\n{INDENT}/// Serves as the default hash function.\n{INDENT}/// </summary>\n"
+        code += f"{INDENT}public override int GetHashCode()\n{INDENT}{{\n"
+        
+        # Collect field names for HashCode.Combine (skip const fields)
+        hash_fields = []
+        for prop_name, prop_schema in non_const_properties.items():
+            field_name = prop_name
+            if self.is_csharp_reserved_word(field_name):
+                field_name = f"@{field_name}"
+            if self.pascal_properties:
+                field_name = pascal(field_name)
+            if field_name == class_name:
+                field_name += "_"
+            
+            field_type = self.convert_structure_type_to_csharp(class_name, field_name, prop_schema, parent_namespace)
+            
+            # Handle special types that need custom hash code computation
+            if field_type == 'byte[]' or field_type == 'byte[]?':
+                hash_fields.append(f"({field_name} != null ? System.Convert.ToBase64String({field_name}).GetHashCode() : 0)")
+            elif field_type.startswith('List<') or field_type.startswith('HashSet<') or field_type.startswith('Dictionary<'):
+                # For collections, compute hash from elements
+                if field_type.endswith('?'):
+                    hash_fields.append(f"({field_name} != null ? {field_name}.Aggregate(0, (acc, item) => HashCode.Combine(acc, item)) : 0)")
+                else:
+                    hash_fields.append(f"{field_name}.Aggregate(0, (acc, item) => HashCode.Combine(acc, item))")
+            else:
+                hash_fields.append(field_name)
+        
+        # HashCode.Combine supports up to 8 parameters
+        if len(hash_fields) <= 8:
+            code += f"{INDENT*2}return HashCode.Combine({', '.join(hash_fields)});\n"
+        else:
+            # For more than 8 fields, use HashCode.Add
+            code += f"{INDENT*2}var hash = new HashCode();\n"
+            for field in hash_fields:
+                code += f"{INDENT*2}hash.Add({field});\n"
+            code += f"{INDENT*2}return hash.ToHashCode();\n"
+        
+        code += f"{INDENT}}}\n"
+        
+        return code
+
+    def write_to_file(self, namespace: str, name: str, definition: str) -> None:
+        """ Writes the class or enum to a file """
+        directory_path = os.path.join(
+            self.output_dir, os.path.join('src', namespace.replace('.', os.sep)))
+        if not os.path.exists(directory_path):
+            os.makedirs(directory_path, exist_ok=True)
+        file_path = os.path.join(directory_path, f"{name}.cs")
+
+        with open(file_path, 'w', encoding='utf-8') as file:
+            # Common using statements (add more as needed)
+            file_content = "using System;\nusing System.Collections.Generic;\n"
+            file_content += "using System.Linq;\n"
+            if self.system_text_json_annotation:
+                file_content += "using System.Text.Json;\n"
+                file_content += "using System.Text.Json.Serialization;\n"
+            if self.newtonsoft_json_annotation:
+                file_content += "using Newtonsoft.Json;\n"
+            if self.system_xml_annotation:  # Add XML serialization using directive
+                file_content += "using System.Xml.Serialization;\n"
+
+            if namespace:
+                # Namespace declaration with correct indentation for the definition
+                file_content += f"\nnamespace {namespace}\n{{\n"
+                indented_definition = '\n'.join(
+                    [f"{INDENT}{line}" for line in definition.split('\n')])
+                file_content += f"{indented_definition}\n}}"
+            else:
+                file_content += definition
+            file.write(file_content)
+
+    def convert(self, structure_schema_path: str, output_dir: str) -> None:
+        """ Converts a JSON Structure schema file to C# classes """
+        self.output_dir = output_dir
+        
+        with open(structure_schema_path, 'r', encoding='utf-8') as file:
+            schema = json.load(file)
+        
+        self.convert_schema(schema, output_dir)
+
+    def convert_schema(self, schema: JsonNode, output_dir: str) -> None:
+        """ Converts a JSON Structure schema to C# classes """
+        if not isinstance(schema, list):
+            schema = [schema]
+
+        # Determine project name: use explicit project_name if set, otherwise derive from base_namespace
+        if self.project_name and self.project_name.strip():
+            # Use explicitly set project name
+            project_name = self.project_name
+        else:
+            # Fall back to using base_namespace as project name
+            project_name = self.base_namespace
+            if not project_name or project_name.strip() == '':
+                # Derive from output directory name as fallback
+                project_name = os.path.basename(os.path.abspath(output_dir))
+                if not project_name or project_name.strip() == '':
+                    project_name = 'Generated'
+                # Clean up the project name
+                project_name = project_name.replace('-', '_').replace(' ', '_')
+                # Update base_namespace to match (only if it was empty)
+                self.base_namespace = project_name
+                import warnings
+                warnings.warn(f"No namespace provided, using '{project_name}' derived from output directory", UserWarning)
+        
+        self.schema_doc = schema
+        if not os.path.exists(output_dir):
+            os.makedirs(output_dir, exist_ok=True)
+        
+        # Create solution file if it doesn't exist
+        if not glob.glob(os.path.join(output_dir, "src", "*.sln")):
+            sln_file = os.path.join(output_dir, f"{project_name}.sln")
+            if not os.path.exists(sln_file):
+                if not os.path.exists(os.path.dirname(sln_file)) and os.path.dirname(sln_file):
+                    os.makedirs(os.path.dirname(sln_file))
+                with open(sln_file, 'w', encoding='utf-8') as file:
+                    file.write(process_template(
+                        "structuretocsharp/project.sln.jinja", 
+                        project_name=project_name, 
+                        uuid=lambda:str(uuid.uuid4()),
+                        system_xml_annotation=self.system_xml_annotation,
+                        system_text_json_annotation=self.system_text_json_annotation,
+                        newtonsoft_json_annotation=self.newtonsoft_json_annotation))
+        
+        # Create main project file if it doesn't exist
+        if not glob.glob(os.path.join(output_dir, "src", "*.csproj")):
+            csproj_file = os.path.join(output_dir, "src", f"{pascal(project_name)}.csproj")
+            if not os.path.exists(csproj_file):
+                if not os.path.exists(os.path.dirname(csproj_file)):
+                    os.makedirs(os.path.dirname(csproj_file))
+                with open(csproj_file, 'w', encoding='utf-8') as file:
+                    file.write(process_template(
+                        "structuretocsharp/project.csproj.jinja",
+                        project_name=project_name, 
+                        system_xml_annotation=self.system_xml_annotation,
+                        system_text_json_annotation=self.system_text_json_annotation,
+                        newtonsoft_json_annotation=self.newtonsoft_json_annotation))
+        
+        # Create test project file if it doesn't exist
+        if not glob.glob(os.path.join(output_dir, "test", "*.csproj")):
+            csproj_test_file = os.path.join(output_dir, "test", f"{pascal(project_name)}.Test.csproj")
+            if not os.path.exists(csproj_test_file):
+                if not os.path.exists(os.path.dirname(csproj_test_file)):
+                    os.makedirs(os.path.dirname(csproj_test_file))
+                with open(csproj_test_file, 'w', encoding='utf-8') as file:
+                    file.write(process_template(
+                        "structuretocsharp/testproject.csproj.jinja", 
+                        project_name=project_name,
+                        system_xml_annotation=self.system_xml_annotation,
+                        system_text_json_annotation=self.system_text_json_annotation,
+                        newtonsoft_json_annotation=self.newtonsoft_json_annotation))
+
+        self.output_dir = output_dir
+        
+        # Register all schemas with $id keywords for cross-references
+        for structure_schema in (s for s in schema if isinstance(s, dict)):
+            self.register_schema_ids(structure_schema)
+        
+        # Process each schema
+        for structure_schema in (s for s in schema if isinstance(s, dict)):
+            # Store definitions for later use
+            if 'definitions' in structure_schema:
+                self.definitions = structure_schema['definitions']
+            
+            # Store $offers for add-in system
+            if '$offers' in structure_schema:
+                self.offers = structure_schema['$offers']
+            
+            # Process root type FIRST so inline unions can generate derived classes
+            if 'type' in structure_schema:
+                self.generate_class_or_choice(structure_schema, '', write_file=True)
+            elif '$root' in structure_schema:
+                root_ref = structure_schema['$root']
+                root_schema = self.resolve_ref(root_ref, structure_schema)
+                if root_schema:
+                    ref_path = root_ref.split('/')
+                    type_name = ref_path[-1]
+                    ref_namespace = '.'.join(ref_path[2:-1]) if len(ref_path) > 3 else ''
+                    self.generate_class_or_choice(root_schema, ref_namespace, write_file=True, explicit_name=type_name)
+            
+            # Now process remaining definitions that weren't generated as part of inline unions
+            if 'definitions' in structure_schema:
+                self.process_definitions(self.definitions, '')
+            
+            # Generate add-in interfaces and extensible wrapper classes
+            if self.offers:
+                self.generate_addins(structure_schema)
+        
+        # Generate tuple converter utility class if needed (after all types processed)
+        if self.system_text_json_annotation:
+            self.generate_tuple_converter(output_dir)
+        
+        # Generate tests
+        self.generate_tests(output_dir)
+        
+        # Generate instance serializer program
+        self.generate_instance_serializer(output_dir)
+
+    def generate_addins(self, structure_schema: Dict) -> None:
+        """
+        Generates add-in interfaces and view classes for types that have $offers.
+        
+        For each add-in in $offers, creates:
+        1. An interface I{AddinName} with the add-in properties
+        2. An internal view class that wraps the Extensions dictionary
+        3. Implicit operators on the base class that convert to the interface
+        """
+        if not self.offers or not isinstance(self.offers, dict):
+            return
+        
+        root_type_name = structure_schema.get('name', 'Document')
+        namespace_pascal = pascal(self.base_namespace)
+        
+        # Generate interface and view class for each add-in
+        view_classes = []
+        for addin_name, addin_def in self.offers.items():
+            self.generate_addin_interface(addin_name, addin_def, namespace_pascal)
+            view_class_name = self.generate_addin_view_class(addin_name, addin_def, namespace_pascal)
+            view_classes.append((addin_name, view_class_name))
+        
+        # Add Extensions dictionary and implicit operators to the base class
+        if 'type' in structure_schema and structure_schema['type'] == 'object':
+            self.add_extensions_to_base_class(root_type_name, view_classes, namespace_pascal)
+
+    def generate_addin_interface(self, addin_name: str, addin_def: Any, namespace: str) -> None:
+        """
+        Generates an interface for an add-in from $offers.
+        
+        Args:
+            addin_name: Name of the add-in (e.g., "AuditInfo")
+            addin_def: Definition of the add-in (either inline properties or a $ref)
+            namespace: Target namespace for the interface
+        """
+        interface_name = f"I{pascal(addin_name)}"
+        
+        # Resolve the add-in definition if it's a reference
+        if isinstance(addin_def, str):
+            # It's a JSON pointer reference
+            addin_def = self.resolve_ref(addin_def, self.schema_doc)
+        elif isinstance(addin_def, dict) and '$ref' in addin_def:
+            addin_def = self.resolve_ref(addin_def['$ref'], self.schema_doc)
+        
+        if not addin_def or not isinstance(addin_def, dict):
+            return
+        
+        properties = addin_def.get('properties', {})
+        if not properties:
+            return
+        
+        # Generate interface definition
+        interface_code = f"{INDENT}/// <summary>\n"
+        interface_code += f"{INDENT}/// Add-in interface: {addin_name}\n"
+        if 'description' in addin_def:
+            interface_code += f"{INDENT}/// {addin_def['description']}\n"
+        interface_code += f"{INDENT}/// </summary>\n"
+        interface_code += f"{INDENT}public interface {interface_name}\n"
+        interface_code += f"{INDENT}{{\n"
+        
+        # Generate properties
+        for prop_name, prop_schema in properties.items():
+            if not isinstance(prop_schema, dict):
+                continue
+                
+            csharp_prop_name = pascal(prop_name) if self.pascal_properties else prop_name
+            csharp_type = self.convert_structure_type_to_csharp(interface_name, prop_name, prop_schema, namespace)
+            
+            # Add XML doc comment
+            if 'description' in prop_schema:
+                interface_code += f"{INDENT}{INDENT}/// <summary>\n"
+                interface_code += f"{INDENT}{INDENT}/// {prop_schema['description']}\n"
+                interface_code += f"{INDENT}{INDENT}/// </summary>\n"
+            
+            # Interface properties are always nullable for add-ins (both value types and reference types)
+            if not csharp_type.endswith('?'):
+                csharp_type += '?'
+            
+            interface_code += f"{INDENT}{INDENT}{csharp_type} {csharp_prop_name} {{ get; set; }}\n"
+        
+        interface_code += f"{INDENT}}}\n"
+        
+        # Write interface to file
+        self.write_to_file(namespace, interface_name, interface_code)
+        
+        # Track as generated
+        qualified_name = 'global::' + self.get_qualified_name(namespace, interface_name)
+        self.generated_types[qualified_name] = "interface"
+
+    def generate_extensible_class(self, base_type_name: str, addin_names: List[str], namespace: str) -> None:
+        """
+        DEPRECATED: Replaced by generate_addin_view_class and add_extensions_to_base_class.
+        This method is kept for backward compatibility but does nothing.
+        """
+        pass
+
+    def generate_addin_view_class(self, addin_name: str, addin_def: Any, namespace: str) -> str:
+        """
+        Generates an internal view class that wraps the Extensions dictionary.
+        
+        Example output:
+        internal sealed class AuditInfoView : IAuditInfo
+        {
+            private readonly Dictionary<string, object?> _extensions;
+            
+            public AuditInfoView(Dictionary<string, object?> extensions)
+            {
+                _extensions = extensions;
+            }
+            
+            public string? CreatedBy
+            {
+                get => _extensions.TryGetValue("createdBy", out var val) ? val as string : null;
+                set { if (value != null) _extensions["createdBy"] = value; else _extensions.Remove("createdBy"); }
+            }
+        }
+        
+        Args:
+            addin_name: Name of the add-in (e.g., "AuditInfo")
+            addin_def: Definition of the add-in
+            namespace: Target namespace
+            
+        Returns:
+            The name of the generated view class
+        """
+        view_class_name = f"{pascal(addin_name)}View"
+        interface_name = f"I{pascal(addin_name)}"
+        
+        # Resolve the add-in definition if it's a reference
+        if isinstance(addin_def, str):
+            addin_def = self.resolve_ref(addin_def, self.schema_doc)
+        elif isinstance(addin_def, dict) and '$ref' in addin_def:
+            addin_def = self.resolve_ref(addin_def['$ref'], self.schema_doc)
+        
+        if not addin_def or not isinstance(addin_def, dict):
+            return view_class_name
+        
+        properties = addin_def.get('properties', {})
+        if not properties:
+            return view_class_name
+        
+        # Generate class definition
+        class_code = f"{INDENT}/// <summary>\n"
+        class_code += f"{INDENT}/// View class wrapping Extensions dictionary for {addin_name} add-in\n"
+        if 'description' in addin_def:
+            class_code += f"{INDENT}/// {addin_def['description']}\n"
+        class_code += f"{INDENT}/// </summary>\n"
+        class_code += f"{INDENT}public sealed class {view_class_name} : {interface_name}\n"
+        class_code += f"{INDENT}{{\n"
+        
+        # Add private field
+        class_code += f"{INDENT}{INDENT}private readonly Dictionary<string, object?> _extensions;\n\n"
+        
+        # Add constructor
+        class_code += f"{INDENT}{INDENT}public {view_class_name}(Dictionary<string, object?> extensions)\n"
+        class_code += f"{INDENT}{INDENT}{{\n"
+        class_code += f"{INDENT}{INDENT}{INDENT}_extensions = extensions;\n"
+        class_code += f"{INDENT}{INDENT}}}\n\n"
+        
+        # Generate properties
+        for prop_name, prop_schema in properties.items():
+            if not isinstance(prop_schema, dict):
+                continue
+                
+            csharp_prop_name = pascal(prop_name) if self.pascal_properties else prop_name
+            csharp_type = self.convert_structure_type_to_csharp(view_class_name, prop_name, prop_schema, namespace)
+            
+            # Remove nullable marker for determining base type
+            base_csharp_type = csharp_type.rstrip('?')
+            is_nullable = csharp_type.endswith('?')
+            
+            # Ensure nullable for add-ins
+            if not is_nullable:
+                csharp_type += '?'
+            
+            # Add XML doc comment
+            if 'description' in prop_schema:
+                class_code += f"{INDENT}{INDENT}/// <summary>\n"
+                class_code += f"{INDENT}{INDENT}/// {prop_schema['description']}\n"
+                class_code += f"{INDENT}{INDENT}/// </summary>\n"
+            
+            # Generate getter that reads from dictionary
+            class_code += f"{INDENT}{INDENT}public {csharp_type} {csharp_prop_name}\n"
+            class_code += f"{INDENT}{INDENT}{{\n"
+            
+            # Getter - use TryGetValue with type-specific conversion
+            class_code += f'{INDENT}{INDENT}{INDENT}get => _extensions.TryGetValue("{prop_name}", out var val) && val != null ? '
+            
+            # Add appropriate conversion based on type
+            if base_csharp_type in ['string', 'bool', 'int', 'long', 'float', 'double', 'decimal']:
+                if base_csharp_type == 'string':
+                    class_code += 'val as string : null;\n'
+                elif base_csharp_type == 'bool':
+                    class_code += 'Convert.ToBoolean(val) : null;\n'
+                elif base_csharp_type in ['int', 'long', 'float', 'double', 'decimal']:
+                    class_code += f'Convert.To{base_csharp_type.capitalize()}(val) : null;\n'
+                else:
+                    class_code += 'val : null;\n'
+            else:
+                # For complex types, try direct cast
+                class_code += f'({base_csharp_type})val : null;\n'
+            
+            # Setter - write to dictionary or remove if null
+            class_code += f'{INDENT}{INDENT}{INDENT}set {{ if (value != null) _extensions["{prop_name}"] = value; else _extensions.Remove("{prop_name}"); }}\n'
+            
+            class_code += f"{INDENT}{INDENT}}}\n\n"
+        
+        class_code += f"{INDENT}}}\n"
+        
+        # Write class to file
+        self.write_to_file(namespace, view_class_name, class_code)
+        
+        # Track as generated (internal, not exported)
+        qualified_name = 'global::' + self.get_qualified_name(namespace, view_class_name)
+        self.generated_types[qualified_name] = "view_class"
+        
+        return view_class_name
+
+    def add_extensions_to_base_class(self, base_type_name: str, view_classes: List[tuple], namespace: str) -> None:
+        """
+        Adds Extensions dictionary property and implicit operators to the base class.
+        
+        Appends to the existing base class file:
+        - Extensions property (Dictionary<string, object?>)
+        - Implicit operators for each add-in interface
+        
+        Args:
+            base_type_name: Name of the base type
+            view_classes: List of (addin_name, view_class_name) tuples
+            namespace: Target namespace
+        """
+        base_class_name = pascal(base_type_name)
+        
+        # Generate the partial class extension code
+        extension_code = f"{INDENT}/// <summary>\n"
+        extension_code += f"{INDENT}/// Partial class extension for {base_class_name} with add-in support\n"
+        extension_code += f"{INDENT}/// </summary>\n"
+        extension_code += f"{INDENT}public partial class {base_class_name}\n"
+        extension_code += f"{INDENT}{{\n"
+        
+        # Add Extensions property
+        extension_code += f"{INDENT}{INDENT}/// <summary>\n"
+        extension_code += f"{INDENT}{INDENT}/// Extension properties storage for add-ins.\n"
+        extension_code += f"{INDENT}{INDENT}/// Unknown JSON properties are automatically captured here during deserialization.\n"
+        extension_code += f"{INDENT}{INDENT}/// </summary>\n"
+        
+        if self.system_text_json_annotation:
+            extension_code += f'{INDENT}{INDENT}[System.Text.Json.Serialization.JsonExtensionData]\n'
+        if self.newtonsoft_json_annotation:
+            extension_code += f'{INDENT}{INDENT}[Newtonsoft.Json.JsonExtensionData]\n'
+        
+        extension_code += f"{INDENT}{INDENT}public Dictionary<string, object?> Extensions {{ get; set; }} = new();\n\n"
+        
+        # Add implicit operators for each add-in
+        for addin_name, view_class_name in view_classes:
+            interface_name = f"I{pascal(addin_name)}"
+            
+            extension_code += f"{INDENT}{INDENT}/// <summary>\n"
+            extension_code += f"{INDENT}{INDENT}/// Implicit conversion to {interface_name} view\n"
+            extension_code += f"{INDENT}{INDENT}/// </summary>\n"
+            extension_code += f"{INDENT}{INDENT}public static implicit operator {view_class_name}({base_class_name} obj)\n"
+            extension_code += f"{INDENT}{INDENT}{INDENT}=> new {view_class_name}(obj.Extensions);\n\n"
+        
+        extension_code += f"{INDENT}}}\n"
+        
+        # Write to a separate file (e.g., ProductExtensions.cs)
+        extension_file_name = f"{base_class_name}Extensions"
+        self.write_to_file(namespace, extension_file_name, extension_code)
+
+    def process_definitions(self, definitions: Dict, namespace_path: str) -> None:
+        """ Processes the definitions section recursively """
+        for name, definition in definitions.items():
+            if isinstance(definition, dict):
+                if 'type' in definition:
+                    # This is a type definition
+                    current_namespace = self.concat_namespace(namespace_path, '')
+                    # Check if this type was already generated (e.g., as part of inline union)
+                    check_namespace = pascal(self.concat_namespace(self.base_namespace, current_namespace))
+                    check_name = pascal(name)
+                    check_ref = 'global::'+self.get_qualified_name(check_namespace, check_name)
+                    if check_ref not in self.generated_types:
+                        self.generate_class_or_choice(definition, current_namespace, write_file=True, explicit_name=name)
+                else:
+                    # This is a namespace
+                    new_namespace = self.concat_namespace(namespace_path, name)
+                    self.process_definitions(definition, new_namespace)
+
+    def generate_tests(self, output_dir: str) -> None:
+        """ Generates unit tests for all the generated C# classes and enums """
+        test_directory_path = os.path.join(output_dir, "test")
+        if not os.path.exists(test_directory_path):
+            os.makedirs(test_directory_path, exist_ok=True)
+
+        for class_name, type_kind in self.generated_types.items():
+            # Skip test generation for:
+            # 1. View classes (internal wrappers for Extensions dictionary)
+            # 2. Extension partial classes (add implicit operators to base classes)
+            base_name = class_name.split('.')[-1]
+            
+            # Skip view classes (e.g., AuditInfoView)
+            if type_kind == "view_class" or base_name.endswith('View'):
+                continue
+            
+            # Skip extension partial classes (e.g., ProductExtensions)
+            if base_name.endswith('Extensions'):
+                continue
+            
+            if type_kind in ["class", "enum"]:
+                self.generate_test_class(class_name, type_kind, test_directory_path)
+
+    def generate_tuple_converter(self, output_dir: str) -> None:
+        """ Generates the TupleJsonConverter utility class for JSON array serialization """
+        # Check if any tuples were generated
+        has_tuples = any(type_kind == "tuple" for type_kind in self.generated_types.values())
+        if not has_tuples:
+            return  # No tuples, no need for converter
+
+        # Convert base namespace to PascalCase for consistency with other generated classes
+        namespace_pascal = pascal(self.base_namespace)
+        
+        # Generate the converter class
+        converter_definition = process_template(
+            "structuretocsharp/tuple_converter.cs.jinja",
+            namespace=namespace_pascal
+        )
+
+        # Write to the same directory structure as other classes (using PascalCase path)
+        directory_path = os.path.join(
+            output_dir, os.path.join('src', namespace_pascal.replace('.', os.sep)))
+        if not os.path.exists(directory_path):
+            os.makedirs(directory_path, exist_ok=True)
+        converter_file_path = os.path.join(directory_path, "TupleJsonConverter.cs")
+        
+        # Add using statements
+        file_content = "using System;\n"
+        file_content += "using System.Linq;\n"
+        file_content += "using System.Reflection;\n"
+        file_content += "using System.Text.Json;\n"
+        file_content += "using System.Text.Json.Serialization;\n\n"
+        file_content += converter_definition
+        
+        with open(converter_file_path, 'w', encoding='utf-8') as converter_file:
+            converter_file.write(file_content)
+
+    def generate_instance_serializer(self, output_dir: str) -> None:
+        """ Generates InstanceSerializer.cs that creates instances and serializes them to JSON """
+        test_directory_path = os.path.join(output_dir, "test")
+        if not os.path.exists(test_directory_path):
+            os.makedirs(test_directory_path, exist_ok=True)
+
+        # Collect all classes (not enums, tuples, or other types) that have test classes
+        # Skip abstract classes since they cannot be instantiated
+        # Skip view classes and extension partial classes
+        classes = []
+        for class_name, type_kind in self.generated_types.items():
+            if type_kind == "class":
+                base_name = class_name.split('.')[-1]
+                
+                # Skip view classes (internal wrappers for Extensions dictionary)
+                if base_name.endswith('View'):
+                    continue
+                
+                # Skip extension partial classes
+                if base_name.endswith('Extensions'):
+                    continue
+                
+                # Skip abstract classes
+                structure_schema = cast(Dict[str, JsonNode], self.generated_structure_types.get(class_name, {}))
+                if structure_schema.get('abstract', False):
+                    continue
+                    
+                if class_name.startswith("global::"):
+                    class_name = class_name[8:]
+                test_class_name = f"{class_name.split('.')[-1]}Tests"
+                class_base_name = class_name.split('.')[-1]
+                
+                # Get proper namespace from class_name
+                if '.' in class_name:
+                    namespace = ".".join(class_name.split('.')[:-1])
+                else:
+                    namespace = self.base_namespace if self.base_namespace else ''
+                
+                # Build fully qualified test name
+                full_qualified_test_name = f"{namespace}.{test_class_name}" if namespace else test_class_name
+                
+                classes.append({
+                    'class_name': class_base_name,
+                    'test_class_name': test_class_name,
+                    'full_name': class_name,
+                    'full_qualified_test_name': full_qualified_test_name
+                })
+
+        if not classes:
+            return  # No classes to serialize
+
+        program_definition = process_template(
+            "structuretocsharp/program.cs.jinja",
+            classes=classes
+        )
+
+        program_file_path = os.path.join(test_directory_path, "InstanceSerializer.cs")
+        with open(program_file_path, 'w', encoding='utf-8') as program_file:
+            program_file.write(program_definition)
+
+    def generate_test_class(self, class_name: str, type_kind: str, test_directory_path: str) -> None:
+        """ Generates a unit test class for a given C# class or enum """
+        structure_schema: Dict[str, JsonNode] = cast(Dict[str, JsonNode], self.generated_structure_types.get(class_name, {}))
+        if class_name.startswith("global::"):
+            class_name = class_name[8:]
+        test_class_name = f"{class_name.split('.')[-1]}Tests"
+        namespace = ".".join(class_name.split('.')[:-1])
+        class_base_name = class_name.split('.')[-1]
+
+        # Skip test generation for abstract classes (cannot be instantiated)
+        if type_kind == "class" and structure_schema.get('abstract', False):
+            return
+
+        if type_kind == "class":
+            fields = self.get_class_test_fields(structure_schema, class_base_name)
+            test_class_definition = process_template(
+                "structuretocsharp/class_test.cs.jinja",
+                namespace=namespace,
+                test_class_name=test_class_name,
+                class_base_name=class_base_name,
+                fields=fields,
+                system_xml_annotation=self.system_xml_annotation,
+                system_text_json_annotation=self.system_text_json_annotation,
+                newtonsoft_json_annotation=self.newtonsoft_json_annotation
+            )
+        elif type_kind == "enum":
+            # For enums, extract symbols from the enum schema
+            enum_values = structure_schema.get('enum', [])
+            symbols = []
+            if enum_values:
+                for value in enum_values:
+                    if isinstance(value, str):
+                        # Convert to PascalCase enum member name
+                        symbol_name = ''.join(word.capitalize() for word in re.split(r'[_\-\s]+', value))
+                        symbols.append(symbol_name)
+                    else:
+                        # For numeric enums, use Value1, Value2, etc.
+                        symbols.append(f"Value{value}")
+            
+            test_class_definition = process_template(
+                "structuretocsharp/enum_test.cs.jinja",
+                namespace=namespace,
+                test_class_name=test_class_name,
+                enum_base_name=class_base_name,
+                symbols=symbols,
+                system_xml_annotation=self.system_xml_annotation,
+                system_text_json_annotation=self.system_text_json_annotation,
+                newtonsoft_json_annotation=self.newtonsoft_json_annotation
+            )
+        else:
+            return
+
+        test_file_path = os.path.join(test_directory_path, f"{test_class_name}.cs")
+        with open(test_file_path, 'w', encoding='utf-8') as test_file:
+            test_file.write(test_class_definition)
+
+    def get_class_test_fields(self, structure_schema: Dict[str, JsonNode], class_name: str) -> List[Any]:
+        """ Retrieves fields for a given class name """
+
+        class Field:
+            def __init__(self, fn: str, ft: str, tv: Any, ct: bool, pm: bool):
+                self.field_name = fn
+                self.field_type = ft
+                self.test_value = tv
+                self.is_const = ct
+                self.is_primitive = pm
+
+        fields: List[Field] = []
+        if structure_schema and 'properties' in structure_schema:
+            for prop_name, prop_schema in cast(Dict[str, Dict], structure_schema['properties']).items():
+                field_name = prop_name
+                if self.pascal_properties:
+                    field_name = pascal(field_name)
+                if field_name == class_name:
+                    field_name += "_"
+                if self.is_csharp_reserved_word(field_name):
+                    field_name = f"@{field_name}"
+                
+                field_type = self.convert_structure_type_to_csharp(
+                    class_name, field_name, prop_schema, str(structure_schema.get('namespace', '')))
+                is_class = field_type in self.generated_types and self.generated_types[field_type] == "class"
+                
+                # Check if this is a const field
+                is_const = 'const' in prop_schema
+                test_value = self.get_test_value(field_type) if not is_const else self.format_default_value(prop_schema['const'], field_type)
+                
+                f = Field(field_name, field_type, test_value, is_const, not is_class)
+                fields.append(f)
+        return cast(List[Any], fields)
+
+    def get_test_value(self, csharp_type: str) -> str:
+        """Returns a default test value based on the C# type"""
+        # For nullable object types, return typed null to avoid var issues
+        if csharp_type == "object?" or csharp_type == "object":
+            return "null"  # Use null for object types (typically unions) to avoid reference inequality
+        
+        test_values = {
+            'string': '"test_string"',
+            'bool': 'true',
+            'sbyte': '(sbyte)42',
+            'byte': '(byte)42',
+            'short': '(short)42',
+            'ushort': '(ushort)42',
+            'int': '42',
+            'uint': '42U',
+            'long': '42L',
+            'ulong': '42UL',
+            'System.Int128': 'new System.Int128(0, 42)',
+            'System.UInt128': 'new System.UInt128(0, 42)',
+            'float': '3.14f',
+            'double': '3.14',
+            'decimal': '3.14m',
+            'byte[]': 'new byte[] { 0x01, 0x02, 0x03 }',
+            'DateOnly': 'new DateOnly(2024, 1, 1)',
+            'TimeOnly': 'new TimeOnly(12, 0, 0)',
+            'DateTimeOffset': 'new DateTimeOffset(2024, 1, 1, 12, 0, 0, TimeSpan.Zero)',
+            'TimeSpan': 'TimeSpan.FromHours(1)',
+            'Guid': 'new Guid("12345678-1234-1234-1234-123456789012")',
+            'Uri': 'new Uri("https://example.com")',
+            'null': 'null'
+        }
+        if csharp_type.endswith('?'):
+            csharp_type = csharp_type[:-1]
+        
+        # Normalize to use qualified reference (strip global:: prefix if present, then add it)
+        base_type = csharp_type.replace('global::', '')
+        qualified_ref = f'global::{base_type}'
+        
+        # Check if this is a tuple type (generated_types tracks what we've created)
+        if qualified_ref in self.generated_types and self.generated_types[qualified_ref] == "tuple":
+            # For tuple types, we need to construct with test values based on the schema
+            schema = self.generated_structure_types.get(qualified_ref)
+            if schema:
+                tuple_order = schema.get('tuple', [])
+                properties = schema.get('properties', {})
+                test_params = []
+                for prop_name in tuple_order:
+                    if prop_name in properties:
+                        prop_schema = properties[prop_name]
+                        prop_type = self.convert_structure_type_to_csharp(base_type, prop_name, prop_schema, str(schema.get('namespace', '')))
+                        test_params.append(self.get_test_value(prop_type))
+                if test_params:
+                    return f'new {base_type}({", ".join(test_params)})'
+        
+        # Check if this is a choice type (discriminated union)
+        if qualified_ref in self.generated_types and self.generated_types[qualified_ref] == "choice":
+            # For choice types, create an instance with the first choice property set
+            schema = self.generated_structure_types.get(qualified_ref)
+            if schema:
+                choices = schema.get('choices', {})
+                if choices:
+                    # Get the first choice property
+                    first_choice_name, first_choice_schema = next(iter(choices.items()))
+                    choice_type = self.convert_structure_type_to_csharp(base_type, first_choice_name, first_choice_schema, str(schema.get('namespace', '')))
+                    choice_test_value = self.get_test_value(choice_type)
+                    # Use the constructor that takes the first choice
+                    return f'new {base_type}({choice_test_value})'
+        
+        return test_values.get(base_type, test_values.get(csharp_type, f'new {csharp_type}()'))
+
+
+
+
+def convert_structure_to_csharp(
+    structure_schema_path: str, 
+    cs_file_path: str, 
+    base_namespace: str = '',
+    project_name: str = '',
+    pascal_properties: bool = False, 
+    system_text_json_annotation: bool = False, 
+    newtonsoft_json_annotation: bool = False, 
+    system_xml_annotation: bool = False
+):
+    """Converts JSON Structure schema to C# classes
+
+    Args:
+        structure_schema_path (str): JSON Structure input schema path
+        cs_file_path (str): Output C# file path
+        base_namespace (str, optional): Base namespace. Defaults to ''.
+        project_name (str, optional): Explicit project name for .csproj files (separate from namespace). Defaults to ''.
+        pascal_properties (bool, optional): Pascal case properties. Defaults to False.
+        system_text_json_annotation (bool, optional): Use System.Text.Json annotations. Defaults to False.
+        newtonsoft_json_annotation (bool, optional): Use Newtonsoft.Json annotations. Defaults to False.
+        system_xml_annotation (bool, optional): Use System.Xml.Serialization annotations. Defaults to False.
+    """
+
+    if not base_namespace:
+        base_namespace = os.path.splitext(os.path.basename(cs_file_path))[0].replace('-', '_')
+    
+    structtocs = StructureToCSharp(base_namespace)
+    structtocs.project_name = project_name
+    structtocs.pascal_properties = pascal_properties
+    structtocs.system_text_json_annotation = system_text_json_annotation
+    structtocs.newtonsoft_json_annotation = newtonsoft_json_annotation
+    structtocs.system_xml_annotation = system_xml_annotation
+    structtocs.convert(structure_schema_path, cs_file_path)
+
+
+def convert_structure_schema_to_csharp(
+    structure_schema: JsonNode, 
+    output_dir: str, 
+    base_namespace: str = '',
+    project_name: str = '',
+    pascal_properties: bool = False, 
+    system_text_json_annotation: bool = False, 
+    newtonsoft_json_annotation: bool = False, 
+    system_xml_annotation: bool = False
+):
+    """Converts JSON Structure schema to C# classes
+
+    Args:
+        structure_schema (JsonNode): JSON Structure schema to convert
+        output_dir (str): Output directory
+        base_namespace (str, optional): Base namespace for the generated classes. Defaults to ''.
+        project_name (str, optional): Explicit project name for .csproj files (separate from namespace). Defaults to ''.
+        pascal_properties (bool, optional): Pascal case properties. Defaults to False.
+        system_text_json_annotation (bool, optional): Use System.Text.Json annotations. Defaults to False.
+        newtonsoft_json_annotation (bool, optional): Use Newtonsoft.Json annotations. Defaults to False.
+        system_xml_annotation (bool, optional): Use System.Xml.Serialization annotations. Defaults to False.
+    """
+    structtocs = StructureToCSharp(base_namespace)
+    structtocs.project_name = project_name
+    structtocs.pascal_properties = pascal_properties
+    structtocs.system_text_json_annotation = system_text_json_annotation
+    structtocs.newtonsoft_json_annotation = newtonsoft_json_annotation
+    structtocs.system_xml_annotation = system_xml_annotation
+    structtocs.convert_schema(structure_schema, output_dir)