Performance: remove methodtools/wirerope, use direct lru_cache()

setup.cfg

@@ -40,7 +40,6 @@ install_requires =
     requests
     jinja2
     mappyfile
-    methodtools
     jsonpath-rw
     orjson
     more-ds

src/schematools/_utils.py

@@ -0,0 +1,31 @@
+"""Internal utils, not meant to be used outside schematools."""
+
+import functools
+import weakref
+
+
+def cached_method(*lru_args, **lru_kwargs):
+    """A simple lru-cache per object.
+    This removed the need for methodtools.lru_cache(), which uses wirerope for purity.
+    The usage of wirerope started showing up as 5% of the request time, hence it's significant to remove.
+    """
+
+    def decorator(func):
+        @functools.wraps(func)
+        def initial_wrapped_func(self, *args, **kwargs):
+            # Not storing the wrapped method inside the instance. If we had
+            # a strong reference to self the instance would never die.
+            self_weak = weakref.ref(self)
+
+            @functools.wraps(func)
+            @functools.lru_cache(*lru_args, **lru_kwargs)
+            def cached_method(*args, **kwargs):
+                return func(self_weak(), *args, **kwargs)
+
+            # Assigns to the self reference (preserving the cache), and optimizes the next access.
+            setattr(self, func.__name__, cached_method)
+            return cached_method(*args, **kwargs)
+
+        return initial_wrapped_func
+
+    return decorator

src/schematools/permissions/auth.py

@@ -3,12 +3,12 @@
 The :class:`UserScopes` class handles whether a dataset, table or field can be accessed.
 The other classes in this module ease to retrieval of permission objects.
 """
+
 from __future__ import annotations
 
 from typing import Iterable, Iterator
 
-import methodtools
-
+from schematools._utils import cached_method
 from schematools.types import (
     DatasetFieldSchema,
     DatasetSchema,
@@ -74,15 +74,15 @@ def add_query_params(self, params: list[str]):
         """
         self._query_param_names.extend(params)
 
-    @methodtools.lru_cache()  # type: ignore[misc]
+    @cached_method()  # type: ignore[misc]
     def has_all_scopes(self, needed_scopes: frozenset[str]) -> bool:
         """Check whether the request has all scopes.
 
         This performs an AND check: all scopes should be present.
         """
         return self._scopes.issuperset(needed_scopes)
 
-    @methodtools.lru_cache()  # type: ignore[misc]
+    @cached_method()  # type: ignore[misc]
     def has_any_scope(self, needed_scopes: frozenset[str]) -> bool:
         """Check whether the request grants one of the given scopes.
 
@@ -151,7 +151,7 @@ def _has_field_auth_access(self, field: DatasetFieldSchema) -> Permission:
         else:
             return Permission.none
 
-    @methodtools.lru_cache()
+    @cached_method()
     def _has_dataset_profile_access(self, dataset_id: str) -> Permission:
         """Give the permission access level for a dataset, as defined by the profile."""
         return max(
@@ -162,7 +162,7 @@ def _has_dataset_profile_access(self, dataset_id: str) -> Permission:
             default=Permission.none,
         )
 
-    @methodtools.lru_cache()
+    @cached_method()
     def _has_table_profile_access(self, table: DatasetTableSchema) -> Permission:
         """Give the permission level for a table.
 
@@ -239,7 +239,7 @@ def _has_field_profile_access(self, field: DatasetFieldSchema) -> Permission:
 
         return max_permission
 
-    @methodtools.lru_cache()
+    @cached_method()
     def get_active_profile_datasets(self, dataset_id: str) -> list[ProfileDatasetSchema]:
         """Find all profiles that mention a dataset and match the scopes.
 
@@ -260,7 +260,7 @@ def get_active_profile_datasets(self, dataset_id: str) -> list[ProfileDatasetSch
             and (profile_dataset := profile.datasets.get(dataset_id)) is not None
         ]
 
-    @methodtools.lru_cache()
+    @cached_method()
     def get_active_profile_tables(
         self, dataset_id: str, table_id: str
     ) -> list[ProfileTableSchema]:

src/schematools/types.py

@@ -29,9 +29,8 @@
     cast,
 )
 
-from methodtools import lru_cache
-
 from schematools import MAX_TABLE_NAME_LENGTH
+from schematools._utils import cached_method
 from schematools.exceptions import (
     DatasetFieldNotFound,
     DatasetTableNotFound,
@@ -492,7 +491,7 @@ def _get_tables(self, include_nested: bool = False, include_through: bool = Fals
         if include_through:
             yield from self.through_tables
 
-    @lru_cache()  # type: ignore[misc]
+    @cached_method()  # type: ignore[misc]
     def get_table_by_id(
         self, table_id: str, include_nested: bool = True, include_through: bool = True
     ) -> DatasetTableSchema:
@@ -931,7 +930,7 @@ def get_db_fields(self) -> Iterator[DatasetFieldSchema]:
                 continue
             yield field
 
-    @lru_cache()  # type: ignore[misc]
+    @cached_method()  # type: ignore[misc]
     def get_field_by_id(self, field_id: str) -> DatasetFieldSchema:
         """Get a fields based on the ids of the field."""
         for field_schema in self.fields:
@@ -940,7 +939,7 @@ def get_field_by_id(self, field_id: str) -> DatasetFieldSchema:
 
         raise DatasetFieldNotFound(f"Field '{field_id}' does not exist in table '{self.id}'.")
 
-    @lru_cache()  # type: ignore[misc]
+    @cached_method()  # type: ignore[misc]
     def get_additional_relation_by_id(self, relation_id: str) -> AdditionalRelationSchema:
         """Get the reverse relation based on the ids of the relation."""
         for additional_relation in self.additional_relations:
@@ -1631,7 +1630,7 @@ def field_items(self) -> Json | None:
         """Return the item definition for an array type."""
         return self.get("items", {}) if self.is_array else None
 
-    @lru_cache()  # type: ignore[misc]
+    @cached_method()  # type: ignore[misc]
     def get_field_by_id(self, field_id: str) -> DatasetFieldSchema:
         """Finds and returns the subfield with the given id.