structurize 3.9.1__tar.gz → 3.9.3__tar.gz

{structurize-3.9.1/structurize.egg-info → structurize-3.9.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: structurize
-Version: 3.9.1
+Version: 3.9.3
 Summary: Tools to convert from and to JSON Structure from various other schema languages.
 Author-email: Clemens Vasters <clemensv@microsoft.com>
 Classifier: Programming Language :: Python :: 3

{structurize-3.9.1 → structurize-3.9.3}/avrotize/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '3.9.1'
-__version_tuple__ = version_tuple = (3, 9, 1)
+__version__ = version = '3.9.3'
+__version_tuple__ = version_tuple = (3, 9, 3)
 
-__commit_id__ = commit_id = 'gd2cba4eb3'
+__commit_id__ = commit_id = 'g0848ff37a'

{structurize-3.9.1 → structurize-3.9.3}/avrotize/avrotojstruct.py

@@ -13,6 +13,7 @@ class AvroToJsonStructure:
     def __init__(self, avro_encoding: bool = False) -> None:
         self.known_types: set[str] = set()
         self.reference_stack: set[str] = set()
+        self.definition_paths: dict[str, str] = {}
         self.avro_encoding: bool = avro_encoding
 
     # ------------------------------------------------------------------ TOP-LEVEL
@@ -38,10 +39,29 @@ class AvroToJsonStructure:
                     "definitions": {},
                 }
 
-            # TEMPORARY: process only first element
             first = avro_schema[0]
             if isinstance(first, dict):
-                return self.convert(first, namespace)
+                self.known_types.clear()
+                self.reference_stack.clear()
+                self._prepare_definition_paths(avro_schema, namespace)
+
+                current_namespace, name = self.resolve_full_name(
+                    first.get("name", f"AnonymousType_{uuid.uuid4().hex}"),
+                    first.get("namespace", namespace),
+                )
+                fqn = self.get_fqn(current_namespace, name)
+                doc: Dict[str, Any] = {
+                    "$schema": "https://json-structure.org/meta/core/v0/#",
+                    "$id": f"https://example.com/schemas/{fqn}",
+                    "name": name,
+                    "type": "null",
+                    "definitions": {},
+                }
+                for schema in avro_schema:
+                    if isinstance(schema, dict):
+                        self.register_definition(schema, namespace, doc["definitions"])
+                self._apply_extensions(doc)
+                return doc
 
             # First element non-dict → return stub
             bad_name = f"invalid_list_root_{uuid.uuid4().hex[:8]}"
@@ -56,10 +76,11 @@ class AvroToJsonStructure:
         # Reset caches for each top-level conversion
         self.known_types.clear()
         self.reference_stack.clear()
+        self._prepare_definition_paths(avro_schema, namespace)
 
-        current_namespace = avro_schema.get("namespace", namespace)
-        name = self.clean_name(
-            avro_schema.get("name", f"AnonymousType_{uuid.uuid4().hex}")
+        current_namespace, name = self.resolve_full_name(
+            avro_schema.get("name", f"AnonymousType_{uuid.uuid4().hex}"),
+            avro_schema.get("namespace", namespace),
         )
         fqn = self.get_fqn(current_namespace, name)
 
@@ -73,8 +94,26 @@ class AvroToJsonStructure:
 
         # Build definitions – do NOT skip root
         self.register_definition(avro_schema, current_namespace, doc["definitions"])
+        self._apply_extensions(doc)
         return doc
 
+    def _apply_extensions(self, doc: Dict[str, Any]) -> None:
+        uses = set(doc.get("$uses", []))
+        if self._contains_key(doc, "default"):
+            uses.add("JSONStructureValidation")
+        if self._contains_key(doc, "altnames"):
+            uses.add("JSONStructureAlternateNames")
+        if uses:
+            doc["$schema"] = "https://json-structure.org/meta/extended/v0/#"
+            doc["$uses"] = sorted(uses)
+
+    def _contains_key(self, value: Any, key: str) -> bool:
+        if isinstance(value, dict):
+            return key in value or any(self._contains_key(item, key) for item in value.values())
+        if isinstance(value, list):
+            return any(self._contains_key(item, key) for item in value)
+        return False
+
     # ------------------------------------------------------------------ REGISTRATION
 
     def register_definition(
@@ -88,9 +127,9 @@ class AvroToJsonStructure:
         Ensure `avro_schema` has an entry in `definitions`.
         """
 
-        current_namespace = avro_schema.get("namespace", namespace)
-        name = self.clean_name(
-            avro_schema.get("name", f"AnonymousType_{uuid.uuid4().hex}")
+        current_namespace, name = self.resolve_full_name(
+            avro_schema.get("name", f"AnonymousType_{uuid.uuid4().hex}"),
+            avro_schema.get("namespace", namespace),
         )
         fqn = self.get_fqn(current_namespace, name)
 
@@ -113,8 +152,10 @@ class AvroToJsonStructure:
 
         avro_type = avro_schema.get("type")
         # Use the schema's own namespace if provided, otherwise fall back to the passed 'namespace'
-        current_schema_namespace = avro_schema.get("namespace", namespace)
-        name = self.clean_name(avro_schema.get("name", f"AnonymousType_{uuid.uuid4().hex}"))
+        current_schema_namespace, name = self.resolve_full_name(
+            avro_schema.get("name", f"AnonymousType_{uuid.uuid4().hex}"),
+            avro_schema.get("namespace", namespace),
+        )
         fqn = self.get_fqn(current_schema_namespace, name)
 
         if fqn in self.reference_stack:
@@ -202,6 +243,10 @@ class AvroToJsonStructure:
         # If a definition was constructed, add it to the definitions map with proper nesting.
         if type_definition_content is not None:
             parts = fqn.split('/')
+            definition_name = parts[-1]
+            if definition_name != name:
+                type_definition_content["name"] = definition_name
+                type_definition_content.setdefault("altnames", {})["avro"] = name
             current_level_dict = definitions
             for i, part_name in enumerate(parts):
                 if i == len(parts) - 1: # Last part is the type name itself
@@ -232,7 +277,10 @@ class AvroToJsonStructure:
                 return {"type": self.get_primitive_types()[avro_type_schema]}
             # Named type reference
             if "." in avro_type_schema:
-                ref_fqn = avro_type_schema.replace(".", "/")
+                ref_namespace, ref_name = self.resolve_full_name(
+                    avro_type_schema, context_namespace
+                )
+                ref_fqn = self.get_fqn(ref_namespace, ref_name)
             else:
                 ref_fqn = self.get_fqn(context_namespace, self.clean_name(avro_type_schema))
             # JSON Structure Core requires a type reference to be the value of the
@@ -268,8 +316,10 @@ class AvroToJsonStructure:
             if category in ("record", "enum", "fixed"):
                 # Ensure definition exists then reference it
                 self.register_definition(avro_type_schema, inline_ns, definitions)
-                ref_name = self.clean_name(avro_type_schema["name"])
-                ref_fqn = self.get_fqn(inline_ns, ref_name)
+                ref_namespace, ref_name = self.resolve_full_name(
+                    avro_type_schema["name"], inline_ns
+                )
+                ref_fqn = self.get_fqn(ref_namespace, ref_name)
                 # Wrap the reference under ``type`` (see note above) so the emitted
                 # property/items/values/choice node is a valid JSON Structure schema.
                 return {"type": {"$ref": f"#/definitions/{ref_fqn}"}}
@@ -350,11 +400,69 @@ class AvroToJsonStructure:
     def clean_name(self, name: str) -> str:
         return name.replace(".", "_")
 
-    def get_fqn(self, namespace: str | None, name: str) -> str:
+    def resolve_full_name(self, name: str, namespace: str | None) -> tuple[str | None, str]:
+        if "." in name:
+            namespace, name = name.rsplit(".", 1)
+        return namespace, self.clean_name(name)
+
+    def _prepare_definition_paths(
+        self, avro_schema: Any, namespace: str | None = None
+    ) -> None:
+        full_names: set[str] = set()
+
+        def collect(node: Any, current_namespace: str | None) -> None:
+            if isinstance(node, list):
+                for item in node:
+                    collect(item, current_namespace)
+                return
+            if not isinstance(node, dict):
+                return
+
+            category = node.get("type")
+            inline_namespace = node.get("namespace", current_namespace)
+            if category in ("record", "enum", "fixed") and node.get("name"):
+                type_namespace, type_name = self.resolve_full_name(
+                    node["name"], inline_namespace
+                )
+                raw_fqn = self.get_raw_fqn(type_namespace, type_name)
+                full_names.add(raw_fqn)
+                if category == "record":
+                    for field in node.get("fields", []):
+                        collect(field.get("type"), type_namespace)
+                return
+
+            if isinstance(category, (dict, list)):
+                collect(category, inline_namespace)
+            elif category == "array":
+                collect(node.get("items"), inline_namespace)
+            elif category == "map":
+                collect(node.get("values"), inline_namespace)
+
+        collect(avro_schema, namespace)
+        reserved_names = {"type", "definitions"}
+        self.definition_paths.clear()
+        for full_name in full_names:
+            parts = full_name.split("/")
+            encoded_parts = []
+            for index, part in enumerate(parts):
+                prefix = "/".join(parts[: index + 1])
+                if index < len(parts) - 1 and prefix in full_names:
+                    part += "_"
+                elif index == len(parts) - 1 and part in reserved_names:
+                    part += "_"
+                encoded_parts.append(part)
+            self.definition_paths[full_name] = "/".join(encoded_parts)
+
+    @staticmethod
+    def get_raw_fqn(namespace: str | None, name: str) -> str:
         if namespace:
             return f"{namespace.replace('.', '/')}/{name}"
         return name
 
+    def get_fqn(self, namespace: str | None, name: str) -> str:
+        raw_fqn = self.get_raw_fqn(namespace, name)
+        return self.definition_paths.get(raw_fqn, raw_fqn)
+
     @staticmethod
     def get_primitive_types() -> Dict[str, str]:
         return {

structurize-3.9.3/avrotize/avrotojtd.py

@@ -0,0 +1,404 @@
+"""Avrotize Schema to JSON Type Definition (RFC 8927) converter."""
+
+from __future__ import annotations
+
+import json
+from typing import Any
+
+from avrotize.common import altname
+
+AvroSchema = dict[str, Any] | list[Any] | str
+JtdSchema = dict[str, Any]
+
+
+class AvroToJtdConverter:
+    """Convert Avrotize/Avro schemas to JSON Type Definition schemas."""
+
+    PRIMITIVE_TO_JTD: dict[str, str] = {
+        "boolean": "boolean",
+        "int": "int32",
+        "long": "uint32",
+        "float": "float32",
+        "double": "float64",
+        "string": "string",
+    }
+
+    def __init__(self, record_type: str | None = None) -> None:
+        """Initialize the converter."""
+        self.record_type = record_type
+        self.named_types: dict[str, dict[str, Any]] = {}
+        self.simple_names: dict[str, str] = {}
+        self.inherited_namespaces: dict[int, str] = {}
+        self.inline_named_schema_ids: set[int] = set()
+        self.definitions: dict[str, JtdSchema] = {}
+        self.converting: set[str] = set()
+        self.root_key: str | None = None
+
+    def convert(self, avro_schema: AvroSchema) -> JtdSchema:
+        """Convert an Avro schema object to a JTD schema dictionary."""
+        self._register_named_types(avro_schema)
+        root = self._register_top_level(avro_schema)
+        if self.record_type:
+            root = self._find_named_type(self.record_type)
+        if isinstance(root, dict) and root.get("type") in {"record", "enum"}:
+            self.root_key = self._type_key(root)
+        root_namespace = self._schema_namespace(root) if isinstance(root, dict) else None
+        root_jtd = self._convert_type(root, root_namespace)
+        if isinstance(root, dict) and root.get("type") in {"record", "enum"} and root.get("jtdRoot"):
+            root_name = self._type_key(root)
+            self.definitions[root_name] = root_jtd
+            return {"definitions": self.definitions, "ref": root_name}
+        if self.definitions:
+            if isinstance(root, dict) and root.get("type") in {"record", "enum"}:
+                root_name = self._type_key(root)
+                self.definitions[root_name] = root_jtd
+                return {"definitions": self.definitions, "ref": root_name}
+            root_jtd = dict(root_jtd)
+            root_jtd["definitions"] = self.definitions
+        return root_jtd
+
+    def _register_top_level(self, avro_schema: AvroSchema) -> AvroSchema:
+        if isinstance(avro_schema, list):
+            if self._is_discriminator_union(avro_schema):
+                for item in avro_schema:
+                    self._register_named(item)
+                return avro_schema
+            discriminator_branches = [
+                item for item in avro_schema
+                if isinstance(item, dict) and item.get("type") == "record" and item.get("jtdDiscriminator")
+            ]
+            if self._is_discriminator_union(discriminator_branches):
+                for item in avro_schema:
+                    if isinstance(item, dict) and item.get("type") in {"record", "enum"}:
+                        self._register_named(item)
+                helper_dependencies = self._named_dependencies_of(discriminator_branches)
+                return [
+                    item for item in avro_schema
+                    if not (
+                        isinstance(item, dict)
+                        and item.get("type") in {"record", "enum"}
+                        and not item.get("jtdDiscriminator")
+                        and self._type_key(item) in helper_dependencies
+                    )
+                ]
+            named_items = [
+                item for item in avro_schema
+                if isinstance(item, dict) and item.get("type") in {"record", "enum"}
+            ]
+            root_members = [item for item in avro_schema if item not in named_items]
+            has_explicit_root = any(item.get("jtdRoot") for item in named_items) or any(
+                isinstance(item, dict) and item.get("jtdRoot") for item in root_members
+            )
+            if named_items and root_members and has_explicit_root:
+                for item in named_items:
+                    self._register_named(item)
+                if len(root_members) == 1:
+                    return root_members[0]
+                return root_members
+            if not all(isinstance(item, dict) and item.get("type") in {"record", "enum"} for item in avro_schema):
+                return avro_schema
+            for item in avro_schema:
+                self._register_named(item)
+            root = next((item for item in avro_schema if item.get("jtdRoot")), avro_schema[-1] if avro_schema else None)
+            if root is None:
+                raise ValueError("Avro schema list is empty")
+            return root
+        if isinstance(avro_schema, dict) and avro_schema.get("type") in {"record", "enum"}:
+            self._register_named(avro_schema)
+        return avro_schema
+
+    def _register_named_types(
+        self,
+        avro_type: AvroSchema,
+        seen: set[int] | None = None,
+        inherited_namespace: str | None = None,
+        inline: bool = False,
+    ) -> None:
+        if seen is None:
+            seen = set()
+        if isinstance(avro_type, list):
+            for member in avro_type:
+                self._register_named_types(member, seen, inherited_namespace, inline)
+            return
+        if not isinstance(avro_type, dict) or id(avro_type) in seen:
+            return
+        seen.add(id(avro_type))
+
+        schema_type = avro_type.get("type")
+        if schema_type in {"record", "enum"} and avro_type.get("name"):
+            namespace = avro_type.get("namespace", inherited_namespace)
+            if namespace:
+                self.inherited_namespaces[id(avro_type)] = str(namespace)
+            if inline:
+                self.inline_named_schema_ids.add(id(avro_type))
+            self._register_named(avro_type)
+            inherited_namespace = str(namespace) if namespace else None
+        if schema_type == "record":
+            for field in avro_type.get("fields", []):
+                self._register_named_types(field.get("type", "string"), seen, inherited_namespace, True)
+        elif schema_type == "array":
+            self._register_named_types(avro_type.get("items", "string"), seen, inherited_namespace, True)
+        elif schema_type == "map":
+            self._register_named_types(avro_type.get("values", "string"), seen, inherited_namespace, True)
+        elif isinstance(schema_type, (dict, list)):
+            self._register_named_types(schema_type, seen, inherited_namespace, True)
+
+    def _register_named(self, schema: dict[str, Any]) -> None:
+        key = self._type_key(schema)
+        self.named_types[key] = schema
+        self.simple_names[schema["name"]] = key
+
+    def _named_dependencies_of(self, roots: list[Any]) -> set[str]:
+        root_keys = {
+            self._type_key(root)
+            for root in roots
+            if isinstance(root, dict) and root.get("type") in {"record", "enum"}
+        }
+        dependencies: set[str] = set()
+        pending = list(roots)
+        while pending:
+            for reference in self._named_references(pending.pop()):
+                if reference in root_keys or reference in dependencies:
+                    continue
+                dependencies.add(reference)
+                pending.append(self.named_types[reference])
+        return dependencies
+
+    def _named_references(self, avro_type: AvroSchema, namespace: str | None = None) -> set[str]:
+        if isinstance(avro_type, str):
+            key = self._resolve_named_key(avro_type, namespace)
+            return {key} if key else set()
+        if isinstance(avro_type, list):
+            references: set[str] = set()
+            for member in avro_type:
+                references.update(self._named_references(member, namespace))
+            return references
+        if not isinstance(avro_type, dict):
+            return set()
+
+        schema_type = avro_type.get("type")
+        if schema_type == "record":
+            record_namespace = self._schema_namespace(avro_type, namespace)
+            references: set[str] = set()
+            for field in avro_type.get("fields", []):
+                references.update(self._named_references(field.get("type", "string"), record_namespace))
+            return references
+        if schema_type == "array":
+            return self._named_references(avro_type.get("items", "string"), namespace)
+        if schema_type == "map":
+            return self._named_references(avro_type.get("values", "string"), namespace)
+        if isinstance(schema_type, (dict, list, str)):
+            return self._named_references(schema_type, namespace)
+        return set()
+
+    def _find_named_type(self, record_type: str) -> dict[str, Any]:
+        key = self._resolve_named_key(record_type)
+        if key and key in self.named_types:
+            return self.named_types[key]
+        raise ValueError(f"Avro record type '{record_type}' not found")
+
+    def _type_key(self, schema: dict[str, Any]) -> str:
+        namespace = self._schema_namespace(schema)
+        return f"{namespace}.{schema['name']}" if namespace else schema["name"]
+
+    def _schema_namespace(self, schema: dict[str, Any], fallback: str | None = None) -> str | None:
+        if "namespace" in schema:
+            namespace = schema.get("namespace")
+        else:
+            namespace = self.inherited_namespaces.get(id(schema), fallback)
+        return str(namespace) if namespace else None
+
+    def _resolve_named_key(self, reference: str, namespace: str | None = None) -> str | None:
+        if reference in self.named_types:
+            return reference
+        if namespace and "." not in reference:
+            relative_key = f"{namespace}.{reference}"
+            if relative_key in self.named_types:
+                return relative_key
+        return self.simple_names.get(reference)
+
+    def _convert_type(self, avro_type: AvroSchema, namespace: str | None = None) -> JtdSchema:
+        nullable, non_null = self._strip_nullable(avro_type)
+        if self._is_discriminator_union(non_null):
+            jtd = self._convert_discriminator(non_null, namespace)  # type: ignore[arg-type]
+        elif isinstance(non_null, dict) and non_null.get("type") == "record" and non_null.get("jtdDiscriminator"):
+            jtd = self._convert_discriminator([non_null], namespace)
+        else:
+            jtd = self._convert_non_nullable(non_null, namespace)
+        if nullable:
+            jtd = dict(jtd)
+            jtd["nullable"] = True
+        return jtd
+
+    def _convert_non_nullable(self, avro_type: AvroSchema, namespace: str | None = None) -> JtdSchema:
+        if isinstance(avro_type, str):
+            if avro_type == "null":
+                return {"nullable": True}
+            if avro_type in self.PRIMITIVE_TO_JTD:
+                return {"type": self.PRIMITIVE_TO_JTD[avro_type]}
+            ref_key = self._resolve_named_key(avro_type, namespace)
+            if ref_key:
+                self._ensure_definition(ref_key)
+                return {"ref": ref_key}
+            return {"type": "string"}
+        if isinstance(avro_type, list):
+            return self._convert_union(avro_type, namespace)
+        if not isinstance(avro_type, dict):
+            return {"type": "string"}
+
+        if avro_type.get("jtdWrappedDefinition"):
+            fields = avro_type.get("fields", [])
+            if len(fields) == 1:
+                return self._convert_type(fields[0].get("type", "string"), namespace)
+        if avro_type.get("jtdType"):
+            return {"type": avro_type["jtdType"]}
+        if avro_type.get("logicalType") == "timestamp-millis" and avro_type.get("type") == "long":
+            return {"type": "timestamp"}
+
+        schema_type = avro_type.get("type")
+        if isinstance(schema_type, dict) or isinstance(schema_type, list):
+            return self._convert_type(schema_type, namespace)
+        if schema_type in self.PRIMITIVE_TO_JTD:
+            return {"type": self.PRIMITIVE_TO_JTD[schema_type]}
+        if schema_type == "record":
+            key = self._type_key(avro_type)
+            if id(avro_type) in self.inline_named_schema_ids and key != self.root_key:
+                self._ensure_definition(key)
+                return {"ref": key}
+            return self._convert_record(avro_type, namespace)
+        if schema_type == "enum":
+            key = self._type_key(avro_type)
+            if id(avro_type) in self.inline_named_schema_ids and key != self.root_key:
+                self._ensure_definition(key)
+                return {"ref": key}
+            return self._convert_enum(avro_type)
+        if schema_type == "array":
+            return {"elements": self._convert_type(avro_type.get("items", "string"), namespace)}
+        if schema_type == "map":
+            return {"values": self._convert_type(avro_type.get("values", "string"), namespace)}
+        if isinstance(schema_type, str):
+            ref_key = self._resolve_named_key(schema_type, namespace)
+            if ref_key:
+                self._ensure_definition(ref_key)
+                return {"ref": ref_key}
+        return {"type": "string"}
+
+    def _convert_record(self, schema: dict[str, Any], parent_namespace: str | None = None) -> JtdSchema:
+        key = self._type_key(schema)
+        if key in self.converting:
+            return {"ref": key}
+        namespace = self._schema_namespace(schema, parent_namespace)
+        self.converting.add(key)
+        properties: dict[str, JtdSchema] = {}
+        optional_properties: dict[str, JtdSchema] = {}
+        for field in schema.get("fields", []):
+            if field.get("name") == schema.get("jtdDiscriminator"):
+                continue
+            property_name = altname(field, "jtd")
+            field_nullable, field_type = self._strip_nullable(field.get("type", "string"))
+            converted = self._convert_type(field_type, namespace)
+            if field_nullable and field.get("default") is None:
+                optional_properties[property_name] = converted
+            else:
+                if field_nullable:
+                    converted = dict(converted)
+                    converted["nullable"] = True
+                properties[property_name] = converted
+        self.converting.remove(key)
+        result: JtdSchema = {}
+        if properties:
+            result["properties"] = properties
+        if optional_properties:
+            result["optionalProperties"] = optional_properties
+        if "jtdAdditionalProperties" in schema:
+            result["additionalProperties"] = bool(schema["jtdAdditionalProperties"])
+        return result
+
+    @staticmethod
+    def _convert_enum(schema: dict[str, Any]) -> JtdSchema:
+        original_symbols = schema.get("jtdEnumSymbols", {})
+        return {"enum": [original_symbols.get(symbol, symbol) for symbol in schema.get("symbols", [])]}
+
+    def _convert_union(self, union: list[Any], namespace: str | None = None) -> JtdSchema:
+        non_null = [item for item in union if item != "null"]
+        if len(non_null) == 1:
+            return self._convert_type(non_null[0], namespace)
+        if self._is_discriminator_union(non_null):
+            return self._convert_discriminator(non_null, namespace)
+        discriminator_branches = [
+            item for item in non_null
+            if isinstance(item, dict) and item.get("type") == "record" and item.get("jtdDiscriminator")
+        ]
+        if self._is_discriminator_union(discriminator_branches):
+            tagged_ids = {id(branch) for branch in discriminator_branches}
+            converted: list[JtdSchema] = []
+            discriminator_added = False
+            for item in non_null:
+                if id(item) in tagged_ids:
+                    if not discriminator_added:
+                        converted.append(self._convert_discriminator(discriminator_branches, namespace))
+                        discriminator_added = True
+                else:
+                    converted.append(self._convert_type(item, namespace))
+            return {"metadata": {"avrotize-union": converted}}
+        return {"metadata": {"avrotize-union": [self._convert_type(item, namespace) for item in non_null]}}
+
+    def _convert_discriminator(self, union: list[Any], namespace: str | None = None) -> JtdSchema:
+        discriminator = union[0].get("jtdDiscriminator")
+        mapping: dict[str, JtdSchema] = {}
+        for branch in union:
+            tag = branch.get("jtdMappingKey")
+            if tag is None:
+                tag = branch.get("name")
+            mapping[str(tag)] = self._convert_record(branch, namespace)
+        return {"discriminator": discriminator, "mapping": mapping}
+
+    @staticmethod
+    def _strip_nullable(avro_type: AvroSchema) -> tuple[bool, AvroSchema]:
+        if isinstance(avro_type, list) and "null" in avro_type:
+            non_null = [item for item in avro_type if item != "null"]
+            if len(non_null) == 1:
+                return True, non_null[0]
+            return True, non_null
+        return False, avro_type
+
+    @staticmethod
+    def _is_discriminator_union(avro_type: AvroSchema) -> bool:
+        return (
+            isinstance(avro_type, list)
+            and bool(avro_type)
+            and all(isinstance(item, dict) and item.get("type") == "record" and item.get("jtdDiscriminator") for item in avro_type)
+            and len({item.get("jtdDiscriminator") for item in avro_type if isinstance(item, dict)}) == 1
+        )
+
+    def _ensure_definition(self, key: str) -> None:
+        if key in self.definitions:
+            return
+        if key not in self.named_types:
+            return
+        if key in self.converting:
+            self.definitions[key] = {"ref": key}
+            return
+        self.definitions[key] = {}
+        try:
+            schema = self.named_types[key]
+            namespace = self._schema_namespace(schema)
+            if id(schema) in self.inline_named_schema_ids and schema.get("type") == "record":
+                self.definitions[key] = self._convert_record(schema, namespace)
+            elif id(schema) in self.inline_named_schema_ids and schema.get("type") == "enum":
+                self.definitions[key] = self._convert_enum(schema)
+            else:
+                self.definitions[key] = self._convert_non_nullable(schema, namespace)
+        except Exception:
+            self.definitions.pop(key, None)
+            raise
+
+
+def convert_avro_to_jtd(avro_schema_path: str, jtd_file_path: str, record_type: str | None = None) -> None:
+    """Convert an Avrotize Schema file to a JSON Type Definition file."""
+    with open(avro_schema_path, "r", encoding="utf-8") as avro_file:
+        avro_schema = json.load(avro_file)
+    converter = AvroToJtdConverter(record_type=record_type)
+    jtd_schema = converter.convert(avro_schema)
+    with open(jtd_file_path, "w", encoding="utf-8") as jtd_file:
+        json.dump(jtd_schema, jtd_file, indent=2)