Overhaul as_dict generation to avoid endless nesting of objects.

[plomtask] / plomtask / db.py

diff --git a/plomtask/db.py b/plomtask/db.py
index cce2630cd58bfb8bf7283c2eb0d2f45006d8ba26..b3f1db00986b1142f5f31be34060864840ab5bdc 100644 (file)
--- a/plomtask/db.py
+++ b/plomtask/db.py
@@ -274,24 +274,58 @@ class BaseModel(Generic[BaseModelId]):
 
     @property
     def as_dict(self) -> dict[str, object]:
-        """Return self as (json.dumps-coompatible) dict."""
-        d: dict[str, object] = {'id': self.id_}
+        """Return self as (json.dumps-compatible) dict."""
+        library: dict[str, dict[str | int, object]] = {}
+        d: dict[str, object] = {'id': self.id_, '_library': library}
+        for to_save in self.to_save:
+            attr = getattr(self, to_save)
+            if hasattr(attr, 'as_dict_into_reference'):
+                d[to_save] = attr.as_dict_into_reference(library)
+            else:
+                d[to_save] = attr
         if len(self.to_save_versioned) > 0:
             d['_versioned'] = {}
-        for k in self.to_save:
-            attr = getattr(self, k)
-            if hasattr(attr, 'as_dict'):
-                d[k] = attr.as_dict
-            d[k] = attr
         for k in self.to_save_versioned:
             attr = getattr(self, k)
             assert isinstance(d['_versioned'], dict)
             d['_versioned'][k] = attr.history
         for r in self.to_save_relations:
             attr_name = r[2]
-            d[attr_name] = [x.as_dict for x in getattr(self, attr_name)]
+            l: list[int | str] = []
+            for rel in getattr(self, attr_name):
+                l += [rel.as_dict_into_reference(library)]
+            d[attr_name] = l
         return d
 
+    def as_dict_into_reference(self,
+                               library: dict[str, dict[str | int, object]]
+                               ) -> int | str:
+        """Return self.id_ while writing .as_dict into library."""
+        def into_library(library: dict[str, dict[str | int, object]],
+                         cls_name: str,
+                         id_: str | int,
+                         d: dict[str, object]
+                         ) -> None:
+            if cls_name not in library:
+                library[cls_name] = {}
+            if id_ in library[cls_name]:
+                if library[cls_name][id_] != d:
+                    msg = 'Unexpected inequality of entries for ' +\
+                            f'_library at: {cls_name}/{id_}'
+                    raise HandledException(msg)
+            else:
+                library[cls_name][id_] = d
+        as_dict = self.as_dict
+        assert isinstance(as_dict['_library'], dict)
+        for cls_name, dict_of_objs in as_dict['_library'].items():
+            for id_, obj in dict_of_objs.items():
+                into_library(library, cls_name, id_, obj)
+        del as_dict['_library']
+        assert self.id_ is not None
+        into_library(library, self.__class__.__name__, self.id_, as_dict)
+        assert isinstance(as_dict['id'], (int, str))
+        return as_dict['id']
+
     # cache management
     # (we primarily use the cache to ensure we work on the same object in
     # memory no matter where and how we retrieve it, e.g. we don't want