CU-8699g3ajp v2 beta compatibility (#91)

* CU-8699g3ajp: Update backwards compatibility workflow.

Will now fail upon any individual model failing

* CU-8699g3ajp: Add method to remap module packages when loading model with dill

* CU-8699g3ajp: Using fixed module/class name when deserialising with dill

Author: mart-r

medcat/storage/serialisers.py

@@ -13,6 +13,8 @@
 from medcat.storage.serialisables import get_all_serialisable_members
 from medcat.storage.schema import load_schema, save_schema
 from medcat.storage.schema import DEFAULT_SCHEMA_FILE, IllegalSchemaException
+from medcat.utils.legacy.v2_beta import (
+    fix_module_and_cls_name, RemappingUnpickler)
 
 
 logger = logging.getLogger(__name__)
@@ -146,6 +148,7 @@ def deserialise_manually(cls, folder_path: str, man_cls_path: str,
                              **init_kwargs) -> Serialisable:
         logger.info("Deserialising manually based on %s", man_cls_path)
         module_name, cls_name = man_cls_path.rsplit(".", 1)
+        module_name, cls_name = fix_module_and_cls_name(module_name, cls_name)
         rel_module = importlib.import_module(module_name)
         man_cls = getattr(rel_module, cls_name)
         if not issubclass(man_cls, ManualSerialisable):
@@ -253,7 +256,7 @@ def serialise(self, raw_parts: dict[str, Any], target_file: str) -> None:
 
     def deserialise(self, target_file: str) -> dict[str, Any]:
         with open(target_file, 'rb') as f:
-            return _dill.load(f)
+            return RemappingUnpickler(f).load()
 
 
 _DEF_SER = AvailableSerialisers.dill

medcat/utils/legacy/v2_beta.py

@@ -0,0 +1,16 @@
+import dill
+
+
+class RemappingUnpickler(dill.Unpickler):
+    # TODO: remove at some point - this only serves to load beta models
+    def find_class(self, module: str, name: str):
+        # Remap old module path to new one
+        module, name = fix_module_and_cls_name(module, name)
+        return super().find_class(module, name)
+
+
+def fix_module_and_cls_name(module_name: str, cls_name: str
+                            ) -> tuple[str, str]:
+    if module_name.startswith("medcat2"):
+        module_name = module_name.replace("medcat2", "medcat", 1)
+    return module_name, cls_name

tests/backwards_compatibility/check_backwards_compatibility.sh

@@ -1,3 +1,5 @@
+# fail upon individual failure
+set -euo pipefail
 # CONSTANTs/ shouldn't change
 REGRESSION_MODULE="medcat.utils.regression.regression_checker"
 REGRESSION_OPTIONS="--strictness STRICTEST --require-fully-correct"