Merge pull request #469 from linkml/schemaview_cycle_check

schemaview.py: add cycle detection function

Author: kevinschaper

linkml_runtime/utils/schemaview.py

@@ -8,7 +8,6 @@
 import uuid
 import warnings
 from collections import defaultdict, deque
-from collections.abc import Callable, Mapping
 from copy import copy, deepcopy
 from dataclasses import dataclass
 from enum import Enum
@@ -45,7 +44,7 @@
 from linkml_runtime.utils.pattern import PatternResolver
 
 if TYPE_CHECKING:
-    from collections.abc import Mapping
+    from collections.abc import Callable, Iterable, Mapping
     from types import NotImplementedType
 
     from linkml_runtime.utils.metamodelcore import URI, URIorCURIE
@@ -92,13 +91,90 @@ class OrderedBy(Enum):
     """
 
 
+WHITE = 0
+GREY = 1
+BLACK = 2
+
+
+def detect_cycles(f: Callable[[Any], Iterable[Any] | None], x: Any) -> None:
+    """Detect cycles in a graph, using function `f` to walk the graph, starting at node `x`.
+
+    Uses the classic white/grey/black colour coding algorithm to track which nodes have been explored. In this
+    case, "node" refers to any element in a schema and "neighbours" are elements that can be reached from that
+    node by executing function `f`.
+
+    WHITE: unexplored
+    GREY:  node is being processed; processing includes exploring all neighbours reachable via f(node)
+    BLACK: node and all of its neighbours (and their neighbours, etc.) have been processed
+
+    A directed cycle reachable from node `x` raises a ValueError.
+
+    :param f: function that returns an iterable of neighbouring nodes (parents or children)
+    :type f: Callable[[Any], Iterable[Any] | None]
+    :param x: graph node
+    :type x: Any
+    :raises ValueError: if a cycle is discovered through repeated calls to f(x)
+    """
+    # keep track of the processing state of nodes in the graph
+    processing_state: dict[Any, int] = {}
+
+    # Stack entries are (node, processed_flag).
+    # processed_flag == True means all neighbours (nodes generated by running `f(node)`)
+    # have been added to the todo stack and the node can be marked BLACK.
+    todo: list[tuple[Any, bool]] = [(x, False)]
+
+    while todo:
+        node, processed_flag = todo.pop()
+
+        if processed_flag:
+            # all neighbours have been processed
+            processing_state[node] = BLACK
+            continue
+
+        # check the state of this node
+        node_state = processing_state.get(node, WHITE)
+
+        if node_state == GREY:
+            # this node was already being processed
+            # we have discovered an edge back to that node - i.e. a cycle
+            err_msg = f"Cycle detected at node {node!r}"
+            raise ValueError(err_msg)
+
+        if node_state == BLACK:
+            # already fully explored - nothing to do
+            continue
+
+        # mark the node as being processed (GREY) and set the processed_flag to True
+        processing_state[node] = GREY
+        todo.append((node, True))
+
+        # push the neighbours on to the processing stack
+        todo.extend((child, False) for child in f(node) or [])
+
+
 def _closure(
-    f: Callable,
-    x,
+    f: Callable[[Any], Iterable[Any] | None],
+    x: Any,
     reflexive: bool = True,
     depth_first: bool = True,
-    **kwargs: dict[str, Any] | None,  # noqa: ARG001
+    **kwargs: dict[str, Any] | None,
 ) -> list[str | ElementName | ClassDefinitionName | EnumDefinitionName | SlotDefinitionName | TypeDefinitionName]:
+    """Walk the graph using function `f` and generate the closure.
+
+    :param f: function that returns an iterable of neighbouring nodes (parents or children)
+    :type f: Callable[[Any], Iterable[Any]  |  None]
+    :param x: start node
+    :type x: Any
+    :param reflexive: assume the graph is reflexive, defaults to True
+    :type reflexive: bool, optional
+    :param depth_first: depth first traversal, defaults to True
+    :type depth_first: bool, optional
+    :return: list of nodes
+    :rtype: list[str | ElementName | ClassDefinitionName | EnumDefinitionName | SlotDefinitionName | TypeDefinitionName]
+    """
+    if kwargs and kwargs.get("detect_cycles"):
+        detect_cycles(f, x)
+
     rv = [x] if reflexive else []
     visited = []
     todo = [x]

tests/test_utils/input/cycles.yaml

@@ -0,0 +1,189 @@
+# yaml-language-server: $schema=https://linkml.io/linkml-model/linkml_model/jsonschema/meta.schema.json
+id: https://example.org/test-cycle-schema
+name: TestCycleSchema
+description: |
+  Schema with intentional cycles in classes, mixins, slot ranges, and types.
+prefixes:
+  ex:     https://example.org/schema/
+
+default_prefix: ex
+default_range: string
+
+slots:
+  identifier_slot:
+    range: string
+    identifier: true
+  noncycle_slot:
+  slot_a:
+  slot_b:
+  slot_c:
+  mixed_in_slot_a:
+  mixed_in_slot_b:
+
+classes:
+  # Non‑cyclic base class with some children
+  BaseClass:
+    description: Simple base class with no cycles.
+    slots:
+      - noncycle_slot
+      - identifier_slot
+
+  MixinA:
+    mixin: true
+    slots:
+    - mixed_in_slot_a
+
+  MixinB:
+    mixin: true
+    slots:
+    - mixed_in_slot_b
+
+  NonCycleClassA:
+    is_a: BaseClass
+
+  NonCycleClassB:
+    mixins:
+    - MixinA
+    is_a: NonCycleClassA
+
+  NonCycleClassC:
+    mixins:
+    - MixinB
+    is_a: NonCycleClassA
+
+  # cycle in the range of a slot
+  IdentifierCycleClassA:
+    description: Class with a cycle in the range for the identifier slot.
+    attributes:
+      id_slot:
+        range: IdentifierCycleClassB
+        identifier: true
+
+  IdentifierCycleClassB:
+    description: Class with a cycle in the range for the identifier slot.
+    attributes:
+      id_slot:
+        range: IdentifierCycleClassC
+        identifier: true
+
+  IdentifierCycleClassC:
+    description: Class with a cycle in the range for the identifier slot.
+    attributes:
+      id_slot:
+        range: IdentifierCycleClassA
+        identifier: true
+
+  IdentifierCycleClassD:
+    description: Class with itself as the range for the identifier slot.
+    attributes:
+      id_slot:
+        range: IdentifierCycleClassD
+        identifier: true
+
+  # Cycle: ClassA -> ClassB -> ClassC -> ClassA
+  # ClassD and ClassE have the misfortune of inheriting all this nonsense.
+  ClassA:
+    is_a: ClassB
+    description: Part of a subclass inheritance cycle (A -> B).
+    slots:
+      - slot_a
+
+  ClassB:
+    is_a: ClassC
+    description: Part of a subclass inheritance cycle (B -> C).
+    slots:
+      - slot_b
+
+  ClassC:
+    is_a: ClassA
+    description: Part of a subclass inheritance cycle (C -> A).
+    slots:
+      - slot_c
+
+  ClassD:
+    is_a: ClassA
+
+  ClassE:
+    is_a: ClassD
+
+  # Cycle: ClassF -> ClassF
+  # ClassG is_a ClassF so inherits the cycle
+  ClassF:
+    is_a: ClassF
+
+  ClassG:
+    is_a: ClassF
+
+  # Mixin cycle (mixins reference each other)
+  Mixin1:
+    description: Mixin that pulls in Mixin2, forming a mixin cycle.
+    mixin: true
+    mixins:
+      - Mixin2
+
+  Mixin2:
+    description: Mixin that pulls in Mixin1, forming a mixin cycle.
+    mixin: true
+    mixins:
+      - Mixin1
+
+  MixedClass:
+    description: Class that applies both Mixin1 and Mixin2, thus inheriting the mixin cycle.
+    mixins:
+      - Mixin1
+      - Mixin2
+
+
+types:
+  # string and subtypes
+  string:
+    uri: ex:string
+    base: str
+    description: A character string
+
+  super_string:
+    typeof: string
+    description: Type with ancestors, no cycles
+
+  supreme_string:
+    typeof: super_string
+    description: Type with ancestors, no cycles
+
+  integer:
+    uri: ex:integer
+    base: int
+    description: An integer
+
+  boolean:
+    uri: ex:boolean
+    base: Bool
+    repr: bool
+    description: Your basic bool.
+
+  # in a cycle with itself!
+  circle:
+    typeof: circle
+    uri: ex:circle
+
+  circle_of_life:
+    typeof: circle
+
+  # cycle between type_circular and circular_type
+  circular_type:
+    typeof: type_circular
+    uri: ex:circ
+    description: Type in a cycle
+
+  type_circular:
+    typeof: circular_type
+    uri: ex:circ
+    description: Type in a cycle
+
+  # inherit the type_circular/circular_type confusion
+  semi_circular_type:
+    typeof: circular_type
+    description: Type with cyclic ancestors
+
+  curve_type:
+    typeof: semi_circular_type
+    description: Type with cyclic ancestors