Merge pull request #76 from apdavison/issue73

Fix for #73 - Infinite recursion in validate()

Author: Raphael-Gazzotti

pipeline/src/base.py

@@ -161,11 +161,21 @@ def validate(self, ignore=None):
 
         Returns a dict containing information about any validation failures.
         """
+        return self._validate(ignore=ignore)
+
+    def _validate(self, ignore=None, seen=None):
+        # this is implemented as an internal method so that the
+        # "seen" set, needed to avoid possible infinite recursion,
+        # can be hidden from the public interface.
+        if seen is None:
+            seen = set()
         failures = defaultdict(list)
         for property in self.properties:
             value = getattr(self, property.name, None)
-            for key, values in property.validate(value, ignore=ignore).items():
-                failures[key] += values
+            if (id(self), property.name) not in seen:
+                seen.add((id(self), property.name))
+                for key, values in property.validate(value, ignore=ignore, seen=seen).items():
+                    failures[key] += values
         return failures
 
     @property
@@ -298,7 +308,7 @@ def __str__(self):
     def to_jsonld(self):
         return self.value
 
-    def validate(self, ignore=None):
+    def _validate(self, ignore=None, seen=None):
         if ignore is None:
             ignore = []
         failures = defaultdict(list)

pipeline/src/properties.py

@@ -100,14 +100,16 @@ def types(self):
     def is_link(self) -> bool:
         return issubclass(self.types[0], Node)
 
-    def validate(self, value, ignore=None):
+    def validate(self, value, ignore=None, seen=None):
         """
         Check whether `value` satisfies all constraints.
 
         Arguments:
             value: the value to be checked
             ignore: an optional list of check types that should be ignored
                     ("required", "type", "multiplicity")
+            seen: for internal use: contains a set with Python object ids that have
+                  already been encountered in the validation tree.
 
         Returns a dict containing information about any validation failures.
         """
@@ -131,11 +133,10 @@ def validate(self, value, ignore=None):
                             else:
                                 item_type = f"value contains {type(item)}"
                             failures["type"].append(
-                                f"{self.name}: Expected {', '.join(t.__name__ for t in self.types)}, " +
-                                item_type
+                                f"{self.name}: Expected {', '.join(t.__name__ for t in self.types)}, " + item_type
                             )
                     elif isinstance(item, (Node, IRI)):
-                        failures.update(item.validate(ignore=ignore))
+                        failures.update(item._validate(ignore=ignore, seen=seen))
                 if self.min_items:
                     if len(value) < self.min_items and "multiplicity" not in ignore:
                         failures["multiplicity"].append(
@@ -167,11 +168,10 @@ def validate(self, value, ignore=None):
                     else:
                         value_type = f"value contains {type(value)}"
                     failures["type"].append(
-                        f"{self.name}: Expected {', '.join(t.__name__ for t in self.types)}, " +
-                        value_type
+                        f"{self.name}: Expected {', '.join(t.__name__ for t in self.types)}, " + value_type
                     )
             elif isinstance(value, (Node, IRI)):
-                failures.update(value.validate(ignore=ignore))
+                failures.update(value._validate(ignore=ignore, seen=seen))
         # todo: check formatting, multiline
         return failures
 

pipeline/tests/test_instantiation.py

@@ -69,7 +69,7 @@ def test_IRI():
     for value in valid_iris:
         iri = IRI(value)
         assert iri.value == value
-        failures = iri.validate()
+        failures = iri._validate()
         if value.startswith("http"):
             assert not failures
         else:

pipeline/tests/test_regressions.py

@@ -254,3 +254,19 @@ def test_issue0056():
     assert failures["multiplicity"] == ['digital_identifier does not accept multiple values, but contains 2']
     data = dataset.to_jsonld()
     json.dumps(data)  # this should not raise an Exception
+
+
+def test_issue0073():
+    # https://github.com/openMetadataInitiative/openMINDS_Python/issues/73
+    # Infinite recursion in validate()
+    ds1 = omcore.DatasetVersion(
+        short_name="ds1",
+        is_variant_of=None
+    )
+    ds2 = omcore.DatasetVersion(
+        short_name="ds2",
+        is_variant_of=ds1
+    )
+    ds1.is_variant_of = ds2
+
+    failures = ds1.validate()