SyntaxGraph bridge for Python type system

Summary:
Bridges the Python SyntaxGraph (the AST view of Thrift schemas) into the runtime TypeSystem, letting schema definitions be consumed through the unified TypeSystem interface. The SchemaRegistry becomes the on-demand, lazily-resolving entry point for turning schema URIs into runtime type nodes.

Key changes:
  - Add a new SyntaxGraph → TypeSystem bridge that lazily materializes memoized, fully-resolved type nodes from AST definitions, with cycle-safe construction (recursive types resolve against stable identities) and atomic, all-or-nothing builds (a failed build leaves nothing partially cached)
  - Make SchemaRegistry itself the unbounded TypeSystem view, resolving types on demand via module discovery rather than enumerating all URIs up front
  - Encode the key schema semantics at the boundary: typedefs are erased to their true type, the IDL any type maps to the ANY primitive, unions force optional fields, and terse-write presence is preserved through the graph so field qualifiers map correctly

Reviewed By: praihan

Differential Revision: D107588115

fbshipit-source-id: 6ce2a7f4e30fbf59aafa95de6411921140f22d49

Author: Heejong Lee

third-party/thrift/src/thrift/lib/python/schema/schema_registry.py

@@ -34,6 +34,8 @@
 import zstandard  # @manual=fbsource//third-party/pypi/zstandard:zstandard
 from apache.thrift.type.schema import thrift_types as _schema_types
 from thrift.lib.python.schema.syntax_graph import Definition, SyntaxGraph
+from thrift.lib.python.schema.type_system import DefinitionNode, TypeSystem
+from thrift.lib.python.schema.type_system_bridge import SyntaxGraphBridge
 from thrift.python.serializer import deserialize, Protocol
 
 
@@ -126,8 +128,13 @@ def _get_definition_key(thrift_type: type[Any]) -> bytes | None:
     return key
 
 
-class SchemaRegistry:
-    """Pure Python registry for looking up SyntaxGraph Definitions by type or URI."""
+class SchemaRegistry(TypeSystem):
+    """Pure Python registry for looking up SyntaxGraph Definitions by type or URI.
+
+    The registry is also the unbounded ``TypeSystem`` view: it bridges SyntaxGraph
+    definitions into runtime ``DefinitionNode``s on demand, memoized.
+    ``get_known_uris()`` returns ``None`` because the registry is
+    lazy/module-discovery based and cannot enumerate all URIs up front."""
 
     _instance: SchemaRegistry | None = None
 
@@ -154,6 +161,7 @@ def __init__(self) -> None:
         self._builder: IncrementalGraphBuilder = IncrementalGraphBuilder()
         self._module_resolver: Callable[[str], types.ModuleType | None] | None = None
         self._uri_module_map: dict[str, str] | None = None
+        self._ts_bridge: SyntaxGraphBridge = SyntaxGraphBridge(self)
         self._omnibus_seeded: bool = False
 
     @property
@@ -241,6 +249,29 @@ def get_definition_by_uri(self, uri: str) -> Definition:
 
         raise KeyError(f"Definition not found for URI {uri!r}")
 
+    # -- TypeSystem interface (bridged from SyntaxGraph) --------------------
+
+    def definition_by_uri(self, uri: str) -> Definition | None:
+        """SyntaxResolver hook: AST Definition for ``uri`` (``None`` if unknown).
+
+        Wraps ``get_definition_by_uri`` -- triggering the same lazy module
+        discovery -- but returns ``None`` instead of raising on a miss."""
+        try:
+            return self.get_definition_by_uri(uri)
+        except KeyError:
+            return None
+
+    def get_user_defined_type(self, uri: str) -> DefinitionNode | None:
+        """Resolve a URI to a TypeSystem ``DefinitionNode`` (bridged from the
+        SyntaxGraph, memoized). ``None`` if unknown or not a user-defined type
+        (typedefs/constants/services/interactions are excluded)."""
+        return self._ts_bridge.get_user_defined_type(uri)
+
+    def get_known_uris(self) -> None:
+        """``None`` -- the registry is lazy/module-discovery based and cannot
+        enumerate all URIs up front."""
+        return None
+
     def _resolve_module_for_uri(self, uri: str) -> types.ModuleType | None:
         """Find the module that defines the given URI."""
         self._build_uri_to_module_index()

third-party/thrift/src/thrift/lib/python/schema/syntax_graph.py

@@ -26,11 +26,9 @@
 from typing import Any, Generic, TypeVar
 
 from apache.thrift.protocol.detail.protocol_detail.thrift_types import Value
-from apache.thrift.syntax_graph.syntax_graph.thrift_types import (  # noqa: F401 -- re-export
-    FieldPresenceQualifier,
-    Primitive,
-)
+from apache.thrift.syntax_graph.syntax_graph.thrift_types import Primitive
 from apache.thrift.type.schema import thrift_types as _schema_types
+from apache.thrift.type.schema.thrift_types import FieldQualifier
 from thrift.python.serializer import deserialize, Protocol
 
 
@@ -440,7 +438,7 @@ class FieldNode:
         "_id",
         "_name",
         "_type",
-        "_presence",
+        "_qualifier",
         "_doc_block",
         "_annotations",
         "_default_value",
@@ -449,7 +447,7 @@ class FieldNode:
     _id: int
     _name: str
     _type: TypeRef
-    _presence: FieldPresenceQualifier
+    _qualifier: FieldQualifier
     _doc_block: str | None
     _annotations: list[Annotation]
     _default_value: Value | None
@@ -461,15 +459,15 @@ def __init__(
         id: int,
         name: str,
         type: TypeRef,
-        presence: FieldPresenceQualifier,
+        qualifier: FieldQualifier,
         doc_block: str | None,
         annotations: list[Annotation],
         default_value: Value | None = None,
     ) -> None:
         self._id = id
         self._name = name
         self._type = type
-        self._presence = presence
+        self._qualifier = qualifier
         self._doc_block = doc_block
         self._annotations = annotations
         self._default_value = default_value
@@ -488,8 +486,8 @@ def type(self) -> TypeRef:
         return self._type
 
     @property
-    def presence(self) -> FieldPresenceQualifier:
-        return self._presence
+    def qualifier(self) -> FieldQualifier:
+        return self._qualifier
 
     @property
     def doc_block(self) -> str | None:

third-party/thrift/src/thrift/lib/python/schema/syntax_graph_builder.py

@@ -24,10 +24,7 @@
 import functools
 from typing import Any
 
-from apache.thrift.syntax_graph.syntax_graph.thrift_types import (
-    FieldPresenceQualifier,
-    Primitive,
-)
+from apache.thrift.syntax_graph.syntax_graph.thrift_types import Primitive
 from apache.thrift.type.schema import thrift_types as _schema_types
 from apache.thrift.type.standard import thrift_types as _standard_types
 from apache.thrift.type.type_rep import thrift_types as _type_rep_types
@@ -348,13 +345,6 @@ def _type_of_or_none(
 
     # -- Field conversion --------------------------------------------------
 
-    def _presence_of(
-        self, qualifier: _schema_types.FieldQualifier
-    ) -> FieldPresenceQualifier:
-        if qualifier == _schema_types.FieldQualifier.Optional:
-            return FieldPresenceQualifier.OPTIONAL
-        return FieldPresenceQualifier.UNQUALIFIED
-
     def _doc_block_of(self, attrs: _schema_types.DefinitionAttrs) -> str | None:
         if attrs.docs and attrs.docs.contents:
             return attrs.docs.contents
@@ -373,7 +363,7 @@ def _create_field(self, field: _schema_types.Field) -> FieldNode:
             id=field.id,
             name=field.attrs.name,
             type=self._type_of(field.type),
-            presence=self._presence_of(field.qualifier),
+            qualifier=field.qualifier,
             doc_block=self._doc_block_of(field.attrs),
             annotations=self._create_annotations(field.attrs),
             default_value=self._resolve_value(field.customDefault),

third-party/thrift/src/thrift/lib/python/schema/tests/test_syntax_graph.py

@@ -26,7 +26,7 @@
     ExceptionNode,
     ExceptionTypeRef,
     FieldNode,
-    FieldPresenceQualifier,
+    FieldQualifier,
     FunctionNode,
     InteractionNode,
     ListTypeRef,
@@ -104,11 +104,11 @@ def test_struct_fields(self) -> None:
         self.assertEqual(fields[0].name, "field1")
         self.assertIsInstance(fields[0].type, PrimitiveTypeRef)
         self.assertEqual(fields[0].type.primitive, Primitive.I32)
-        self.assertEqual(fields[0].presence, FieldPresenceQualifier.UNQUALIFIED)
+        self.assertEqual(fields[0].qualifier, FieldQualifier.Default)
 
         self.assertEqual(fields[1].id, 2)
         self.assertEqual(fields[1].name, "field2")
-        self.assertEqual(fields[1].presence, FieldPresenceQualifier.OPTIONAL)
+        self.assertEqual(fields[1].qualifier, FieldQualifier.Optional)
 
     def test_enum_basic(self) -> None:
         prog = self.graph.get_program_by_name("syntax_graph")