Process sub-template applicableEntity when it specializes the parent anchor

Author: aothms

.github/workflows/tests.yml

@@ -25,7 +25,7 @@ jobs:
           python-version: "3.10"
 
       - name: Install test dependencies
-        run: python -m pip install "ifcopenshell[advanced]" pytest pyparsing
+        run: python -m pip install "ifcopenshell[advanced]" pytest pyparsing rdflib tabulate xlsxwriter
 
       - name: Install the checked-out MVD modules
         shell: bash

README.md

@@ -28,6 +28,13 @@ concept templates returns a tuple of `template` objects instead. Invalid XML
 and missing template references raise `ValueError` with the relevant document
 or template identifier. Recursive template branches are expanded once.
 
+When a referenced template's `applicableEntity` differs from its anchor entity,
+the parsed tree includes an additional `EntityRule` for that type. Consecutive
+entity rules apply to the same IFC instance; sibling entity rules are alternative
+specializations. Matching roots are expanded without an extra rule. This keeps
+type restrictions visible to traversal, extraction and SPARQL conversion.
+Serialization writes these restrictions as partial-template references again.
+
 ### Graphviz concept templates
 
 The lightweight concept graphs used by the buildingSMART IFC4.x documentation

graphviz.py

@@ -76,15 +76,18 @@ def parse(
     root = min(graph.in_degree(), key=lambda item: item[1])[0]
     resolved_references = _reference_lookup(references or {})
 
-    def build(node: str, top_level_attribute: bool = False) -> tuple[rule, ...]:
+    def build(node: str, top_level_attribute: bool = False, anchor: str = "") -> tuple[rule, ...]:
         node_type = graph.nodes[node].get("type", "EntityRule")
         if node_type == "Reference":
             try:
-                return resolved_references[_normalise_reference(node)].rules
+                referenced = resolved_references[_normalise_reference(node)]
             except KeyError as error:
                 raise ValueError(
                     f"Unknown Graphviz template reference: {node}"
                 ) from error
+            if referenced.entity and referenced.entity.casefold() != anchor.casefold():
+                return (rule("EntityRule", referenced.entity, referenced.rules),)
+            return referenced.rules
 
         if node_type == "Constraint":
             try:
@@ -99,6 +102,7 @@ def build(node: str, top_level_attribute: bool = False) -> tuple[rule, ...]:
             itertools.chain.from_iterable(
                 build(
                     child,
+                    anchor=node.split("_", 1)[0] if node_type == "EntityRule" else anchor,
                     top_level_attribute=(
                         node == root
                         and graph.nodes[child].get("type") == "AttributeRule"

model.py

@@ -167,7 +167,22 @@ def extract(self, ifc_data: ifcopenshell.entity_instance | Any) -> extracted_dat
                 return []
 
             to_combine: list[extracted_data] = []
+            # Sibling entity rules represent alternative specializations of
+            # this instance, just as entity rules under an attribute do.
+            entity_alternatives = [
+                child for child in self.nodes if child.tag == "EntityRule"
+            ]
+            if entity_alternatives:
+                to_combine.append(
+                    [
+                        values
+                        for child in entity_alternatives
+                        for values in child.extract(ifc_data)
+                    ]
+                )
             for child in self.nodes:
+                if child.tag == "EntityRule":
+                    continue
                 if child.tag == "Constraint":
                     on_node = child.attribute[0][0].c.replace("'", "")
                     if isinstance(ifc_data, ifcopenshell.entity_instance):
@@ -249,12 +264,18 @@ def extract(self, ifc_data: ifcopenshell.entity_instance | Any) -> extracted_dat
     def binding_for(self, target: rule) -> str | None:
         binding: str | None = None
 
-        def find(rule: rule, parent: rule | None) -> None:
+        def find(rule: rule, parents: tuple[rule | None, ...]) -> None:
             nonlocal binding
             if rule is target:
-                binding = rule.bind or (parent.bind if parent else None)
-
-        self.traverse(find)
+                binding = rule.bind
+                for parent in reversed(parents):
+                    if binding or parent is None:
+                        break
+                    binding = parent.bind
+                    if parent.tag != "EntityRule":
+                        break
+
+        self.traverse(find, with_parents=True)
         return binding
 
 

mvd.py

@@ -1,5 +1,6 @@
 import ifcopenshell
 import ifcopenshell.geom
+from .model import rule
 
 import itertools
 import os
@@ -36,6 +37,9 @@ def extract_data(mvd_node, ifc_data):
 
 
     """
+    if isinstance(mvd_node, rule):
+        return mvd_node.extract(ifc_data)
+
     to_combine = []
     return_value = []
 

parser.py

@@ -42,6 +42,7 @@ def parse_template(
 
         parsed_rules: list[rule] = []
         next_visited = visited | {template_key}
+        entity = root.getAttribute("applicableEntity")
         # Only the template's own <Rules> are in scope. Nested <SubTemplates>
         # form a hierarchy of separate templates, so they must not contribute
         # rules (and therefore entity scope) to the parent template.
@@ -51,7 +52,7 @@ def parse_template(
             for node in _elements(rules_node):
                 # A <References> child unions one or more <Template>
                 # references, so a single child may expand to several rules.
-                parsed_rules.extend(self._visit_rule(node, "", next_visited))
+                parsed_rules.extend(self._visit_rule(node, "", next_visited, entity))
 
         return template(
             entity=(
@@ -70,12 +71,14 @@ def _visit_rule(
         node: Element,
         prefix: str,
         visited: frozenset[str],
+        anchor: str,
     ) -> tuple[rule, ...]:
         attribute: Any = None
         bind: str | None = None
         optional = False
         target = node
         child_prefix = prefix
+        tag = node.localName
 
         if node.localName == "AttributeRule":
             attribute = node.attributes["AttributeName"].value
@@ -89,6 +92,7 @@ def _visit_rule(
                 )
         elif node.localName == "EntityRule":
             attribute = node.attributes["EntityName"].value
+            anchor = attribute
         elif node.localName == "Template":
             reference = node.attributes["ref"].value
             reference_key = reference.casefold()
@@ -103,6 +107,13 @@ def _visit_rule(
             if "IdPrefix" in node.attributes:
                 child_prefix += node.attributes["IdPrefix"].value
             visited = visited | {reference_key}
+            entity = target.getAttribute("applicableEntity")
+            if entity and entity.casefold() != anchor.casefold():
+                # A reference may specialize its anchor without following an
+                # attribute. Keep that type restriction in the rule tree.
+                tag = "EntityRule"
+                attribute = entity
+                anchor = entity
         elif node.localName == "Constraint":
             attribute = mvdxml_expression.parse(node.attributes["Expression"].value)
         elif node.localName in {
@@ -121,13 +132,13 @@ def _visit_rule(
         children = tuple(
             parsed
             for child in _elements(target)
-            for parsed in self._visit_rule(child, child_prefix, visited)
+            for parsed in self._visit_rule(child, child_prefix, visited, anchor)
         )
 
         if attribute is not None:
             return (
                 rule(
-                    tag=node.localName,
+                    tag=tag,
                     attribute=attribute,
                     nodes=children,
                     bind=child_prefix + bind if bind else None,

serializer.py

@@ -52,7 +52,7 @@ def token(value: mvdxml_expression.node | str) -> str:
 }
 
 
-def _rule(parent: ET.Element, value: rule) -> None:
+def _rule(parent: ET.Element, value: rule, referenced: list[tuple[str, template]]) -> None:
     attributes: dict[str, str] = {}
     if value.tag == "AttributeRule":
         attributes = {"AttributeName": str(value.attribute)}
@@ -67,12 +67,30 @@ def _rule(parent: ET.Element, value: rule) -> None:
 
     element = _add(parent, value.tag, **attributes)
     for tag, children in groupby(value.nodes, key=lambda child: child.tag):
+        if value.tag == "EntityRule" and tag == "EntityRule":
+            # Consecutive entity rules are an in-memory type restriction.
+            # Write them as partial-template references in mvdXML.
+            container = _add(element, "References")
+            for child in children:
+                identifier = str(uuid.uuid4())
+                _add(container, "Template", ref=identifier)
+                referenced.append(
+                    (identifier, template(child.attribute, None, child.nodes))
+                )
+            continue
         container = _add(element, _CONTAINERS[tag])
         for child in children:
-            _rule(container, child)
+            _rule(container, child, referenced)
 
 
-def _template(parent: ET.Element, value: template, identifier: str) -> None:
+def _template(
+    parent: ET.Element,
+    value: template,
+    identifier: str,
+    referenced: list[tuple[str, template]],
+    *,
+    partial: bool = False,
+) -> None:
     element = _add(
         parent,
         "ConceptTemplate",
@@ -81,11 +99,12 @@ def _template(parent: ET.Element, value: template, identifier: str) -> None:
         status="sample",
         applicableSchema="IFC4",
         applicableEntity=value.entity,
+        isPartial="true" if partial else None,
     )
     if value.rules:
         rules = _add(element, "Rules")
         for item in value.rules:
-            _rule(rules, item)
+            _rule(rules, item, referenced)
 
 
 def _key(value: template) -> str | int:
@@ -138,8 +157,15 @@ def serialize(items: Iterable[mvd_item], destination: str | Path) -> None:
         {"uuid": str(uuid.uuid4()), "name": "", "status": "sample"},
     )
     template_elements = _add(document, "Templates")
+    referenced: list[tuple[str, template]] = []
     for key, value in templates.items():
-        _template(template_elements, value, template_ids[key])
+        _template(template_elements, value, template_ids[key], referenced)
+    # Serializing a partial template can discover further specializations.
+    index = 0
+    while index < len(referenced):
+        identifier, value = referenced[index]
+        _template(template_elements, value, identifier, referenced, partial=True)
+        index += 1
 
     roots = [item for item in parsed if isinstance(item, concept_root)]
     if roots:

sparql.py

@@ -365,6 +365,18 @@ def build(rule, parents):
                 return_value = lambda: bld.append(INDENT + "}")
 
             if rule.tag == "EntityRule":
+                siblings = [
+                    child
+                    for child in getattr(parents[-1], "nodes", ())
+                    if child.tag == "EntityRule"
+                ]
+                if len(siblings) > 1:
+                    bld.append(INDENT + "{")
+
+                    def close_alternative():
+                        bld.append(INDENT + ("} UNION" if rule is not siblings[-1] else "}"))
+
+                    return_value = close_alternative
                 # G.nm = G.next_nm
 
                 if not ifcOwl.is_select(rule.attribute):