Merge pull request #296 from emmo-repo/parse-manchester-syntax

Added support for evaluating Manchester expression to owlready2

README.md

@@ -120,6 +120,7 @@ pip install EMMOntoPy
   - [semver]: Required for `ontoversion`-tool.
   - [pydot]: Used for generating graphs.
     Will be deprecated.
+  - [pyparsing](https://github.com/pyparsing/pyparsing): Used for parsing Manchester syntax
 
 See [docs/docker-instructions.md](docs/docker-instructions.md) for how to build a docker image.
 

demo/horizontal/step1_generate_metadata.py

@@ -52,4 +52,4 @@
 
 print("Generated metadata for the usercase ontology:")
 print(f"  {e.coll.count()} instances")
-print(f"  {len(list(e.coll.relations()))} relations")
+print(f"  {len(list(e.coll.get_relations()))} relations")

docs/api_reference/ontopy/manchester.md

@@ -0,0 +1,3 @@
+# manchester
+
+::: ontopy.manchester

docs/index.md

@@ -120,6 +120,7 @@ pip install EMMOntoPy
   - [semver]: Required for `ontoversion`-tool.
   - [pydot]: Used for generating graphs.
     Will be deprecated.
+  - [pyparsing](https://github.com/pyparsing/pyparsing): Used for parsing Manchester syntax
 
 See [docker-instructions.md](docker-instructions.md) for how to build a docker image.
 

emmopy/__init__.py

@@ -10,7 +10,7 @@
     raise RuntimeError("emmopy requires Python 3.6 or later")
 
 # Ensure emmopy is imported before owlready2...
-if "owlready2" in sys.modules and "ontopy" not in sys.modules:
+if "owlready2" in sys.modules and "emmopy" not in sys.modules:
     raise RuntimeError("emmopy must be imported before owlready2")
 
 # Import functions from emmopy

ontopy/__init__.py

@@ -9,7 +9,7 @@
     raise RuntimeError("ontopy requires Python 3.6 or later")
 
 # Ensure ontopy is imported before owlready2...
-if "owlready2" in sys.modules and "emmopy" not in sys.modules:
+if "owlready2" in sys.modules and "ontopy" not in sys.modules:
     raise RuntimeError("ontopy must be imported before owlready2")
 
 # Monkey patch Owlready2 by injecting some methods

ontopy/manchester.py

@@ -0,0 +1,192 @@
+"""Evaluate Manchester syntax
+
+This module compiles restrictions and logical constructs in Manchester
+syntax into Owlready2 classes. The main function in this module is
+`manchester.evaluate()`, see its docstring for usage example.
+
+Pyparsing is used under the hood for parsing.
+"""
+# pylint: disable=unused-import,wrong-import-order
+import pyparsing as pp
+import ontopy  # noqa F401 -- ontopy must be imported before owlready2
+import owlready2
+
+
+GRAMMAR = None  # Global cache
+
+
+def manchester_expression():
+    """Returns pyparsing grammar for a Manchester expression.
+
+    This function is mostly for internal use.
+
+    See also: https://www.w3.org/TR/owl2-manchester-syntax/
+    """
+    # pylint: disable=global-statement,invalid-name,too-many-locals
+    global GRAMMAR
+    if GRAMMAR:
+        return GRAMMAR
+
+    # Subset of the Manchester grammar for expressions
+    # It is based on https://www.w3.org/TR/owl2-manchester-syntax/
+    # but allows logical constructs within restrictions (like Protege)
+    ident = pp.Word(pp.alphas + "_", pp.alphanums + "_", asKeyword=True)
+    uint = pp.Word(pp.nums)
+    alphas = pp.Word(pp.alphas)
+    string = pp.Word(pp.alphanums + ":")
+    quotedString = (
+        pp.QuotedString('"""', multiline=True) | pp.QuotedString('"')
+    )("string")
+    typedLiteral = pp.Combine(quotedString + "^^" + string("datatype"))
+    stringLanguageLiteral = pp.Combine(quotedString + "@" + alphas("language"))
+    stringLiteral = quotedString
+    numberLiteral = pp.pyparsing_common.number("number")
+    literal = (
+        typedLiteral | stringLanguageLiteral | stringLiteral | numberLiteral
+    )
+    logOp = pp.oneOf(["and", "or"], asKeyword=True)
+    expr = pp.Forward()
+    restriction = pp.Forward()
+    primary = pp.Keyword("not")[...] + (
+        restriction | ident("cls") | pp.nestedExpr("(", ")", expr)
+    )
+    objPropExpr = (
+        pp.Literal("inverse")
+        + pp.Suppress("(")
+        + ident("objProp")
+        + pp.Suppress(")")
+        | pp.Literal("inverse") + ident("objProp")
+        | ident("objProp")
+    )
+    dataPropExpr = ident("dataProp")
+    restriction <<= (
+        objPropExpr + pp.Keyword("some") + expr
+        | objPropExpr + pp.Keyword("only") + expr
+        | objPropExpr + pp.Keyword("Self")
+        | objPropExpr + pp.Keyword("value") + ident("individual")
+        | objPropExpr + pp.Keyword("min") + uint + expr
+        | objPropExpr + pp.Keyword("max") + uint + expr
+        | objPropExpr + pp.Keyword("exactly") + uint + expr
+        | dataPropExpr + pp.Keyword("value") + literal
+    )
+    expr <<= primary + (logOp("op") + expr)[...]
+
+    GRAMMAR = expr
+    return expr
+
+
+class ManchesterError(Exception):
+    """Raised on invalid Manchester notation."""
+
+
+# pylint: disable=too-many-statements
+def evaluate(ontology: owlready2.Ontology, expr: str) -> owlready2.Construct:
+    """Evaluate expression in Manchester syntax.
+
+    Args:
+        ontology: The ontology within which the expression will be evaluated.
+        expr: Manchester expression to be evaluated.
+
+    Returns:
+        An Owlready2 construct that corresponds to the expression.
+
+    Example:
+    >>> from ontopy.manchester import evaluate
+    >>> from ontopy import get_ontology
+    >>> emmo = get_ontology.load()
+
+    >>> restriction = evaluate(emmo, 'hasPart some Atom')
+    >>> cls = evaluate(emmo, 'Atom')
+    >>> expr = evaluate(emmo, 'Atom or Molecule')
+
+    Note:
+        Logical expressions (with `not`, `and` and `or`) are supported as
+        well as object property restrictions.  For data properterties are
+        only value restrictions supported so far.
+    """
+
+    # pylint: disable=invalid-name
+    def _parse_literal(r):
+        """Compiles literal to Owlready2 type."""
+        if r.language:
+            v = owlready2.locstr(r.string, r.language)
+        elif r.number:
+            v = r.number
+        else:
+            v = r.string
+        return v
+
+    # pylint: disable=invalid-name,no-else-return,too-many-return-statements
+    # pylint: disable=too-many-branches
+    def _eval(r):
+        """Recursively evaluate expression produced by pyparsing into an
+        Owlready2 construct."""
+
+        def fneg(x):
+            """Negates the argument if `neg` is true."""
+            return owlready2.Not(x) if neg else x
+
+        if isinstance(r, str):  # r is atomic, returns its owlready2 repr
+            return ontology[r]
+
+        neg = False  # whether the expression starts with "not"
+        while r[0] == "not":
+            r.pop(0)  # strip off the "not" and proceed
+            neg = not neg
+
+        if len(r) == 1:  # r is either a atomic or a parenthesised
+            # subexpression that should be further evaluated
+            if isinstance(r[0], str):
+                return fneg(ontology[r[0]])
+            else:
+                return fneg(_eval(r[0]))
+        elif r.op:  # r contains a logical operator: and/or
+            ops = {"and": owlready2.And, "or": owlready2.Or}
+            op = ops[r.op]
+            if len(r) == 3:
+                return op([fneg(_eval(r[0])), _eval(r[2])])
+            else:
+                arg1 = fneg(_eval(r[0]))
+                r.pop(0)
+                r.pop(0)
+                return op([arg1, _eval(r)])
+        elif r.objProp:  # r is a restriction
+            if r[0] == "inverse":
+                r.pop(0)
+                prop = owlready2.Inverse(ontology[r[0]])
+            else:
+                prop = ontology[r[0]]
+            rtype = r[1]
+            if rtype == "Self":
+                return fneg(prop.has_self())
+            r.pop(0)
+            r.pop(0)
+            f = getattr(prop, rtype)
+            if rtype == "value":
+                return fneg(f(_eval(r)))
+            elif rtype in ("some", "only"):
+                return fneg(f(_eval(r)))
+            elif rtype in ("min", "max", "exactly"):
+                cardinality = r.pop()
+                return fneg(f(cardinality, _eval(r)))
+            else:
+                raise ManchesterError(f"invalid restriction type: {rtype}")
+        elif r.dataProp:  # r is a data property restriction
+            prop = ontology[r[0]]
+            rtype = r[1]
+            r.pop(0)
+            r.pop(0)
+            f = getattr(prop, rtype)
+            if rtype == "value":
+                print("===", _parse_literal(r))
+                return f(_parse_literal(r))
+            else:
+                raise ManchesterError(
+                    f"unimplemented data property restriction: "
+                    f"{prop} {rtype} {r}"
+                )
+        else:
+            raise ManchesterError(f"invalid expression: {r}")
+
+    grammar = manchester_expression()
+    return _eval(grammar.parseString(expr, parseAll=True))

requirements.txt

@@ -6,6 +6,7 @@ Owlready2>=0.28,<0.36,!=0.32,!=0.34
 packaging>=21.0<22
 pydot>=1.4.1,<2
 Pygments>=2.7.4,<3
+pyparsing>=2.4.7
 PyYAML>=5.4.1,<7
 rdflib>=4.2.1,<7
 semver>=2.8.1,<3

tests/test_manchester.py

@@ -0,0 +1,75 @@
+from ontopy import get_ontology
+from ontopy.manchester import evaluate
+from owlready2 import And, Or, Not, Inverse, locstr
+
+
+emmo = get_ontology().load()
+
+
+def check(s, expected):
+    r = evaluate(emmo, s)
+    print(s, "-->", r)
+    assert repr(r) == repr(expected)
+
+
+def test_manchester():
+    check("Item", emmo.Item)
+    check("not Item", Not(emmo.Item))
+    check("not not Item", emmo.Item)
+    check("not (not Item)", Not(Not(emmo.Item)))
+    check("hasPart some Atom", emmo.hasPart.some(emmo.Atom))
+    check("Atom and not Molecule", emmo.Atom & Not(emmo.Molecule))
+    check("Atom and (not Molecule)", emmo.Atom & Not(emmo.Molecule))
+    check("not Atom and Molecule", Not(emmo.Atom) & emmo.Molecule)
+    check("(not Atom) and Molecule", Not(emmo.Atom) & emmo.Molecule)
+    check("inverse hasPart some Atom", Inverse(emmo.hasPart).some(emmo.Atom))
+    check("inverse(hasPart) some Atom", Inverse(emmo.hasPart).some(emmo.Atom))
+    check("not hasPart some Atom", Not(emmo.hasPart.some(emmo.Atom)))
+    check("not (hasPart some Atom)", Not(emmo.hasPart.some(emmo.Atom)))
+    check("hasPart some (not Atom)", emmo.hasPart.some(Not(emmo.Atom)))
+    check("hasPart some not Atom", emmo.hasPart.some(Not(emmo.Atom)))
+    check("not hasPart some not Atom", Not(emmo.hasPart.some(Not(emmo.Atom))))
+    check(
+        "hasPart only (inverse hasPart some not Atom)",
+        emmo.hasPart.only(Inverse(emmo.hasPart).some(Not(emmo.Atom))),
+    )
+    check(
+        "hasPart only inverse hasPart some not Atom",
+        emmo.hasPart.only(Inverse(emmo.hasPart).some(Not(emmo.Atom))),
+    )
+    check(
+        "Atom and Molecule and Proton",
+        emmo.Atom & (emmo.Molecule & emmo.Proton),
+    )
+    check(
+        "Atom and (Molecule and Proton)",
+        emmo.Atom & (emmo.Molecule & emmo.Proton),
+    )
+    check(
+        "(Atom and Molecule) or Proton",
+        (emmo.Atom & emmo.Molecule) | emmo.Proton,
+    )
+    check(
+        "(Atom and Molecule) or Proton",
+        (emmo.Atom & emmo.Molecule) | emmo.Proton,
+    )
+    check(
+        "inverse(hasPart) value Universe",
+        Inverse(emmo.hasPart).value(emmo.Universe),
+    )
+    # literal data restriction
+    check('hasSymbolData value "hello"', emmo.hasSymbolData.value("hello"))
+    check("hasSymbolData value 42", emmo.hasSymbolData.value(42))
+    check("hasSymbolData value 3.14", emmo.hasSymbolData.value(3.14))
+    check(
+        'hasSymbolData value "abc"^^xsd:string',
+        emmo.hasSymbolData.value("abc"),
+    )
+    check(
+        'hasSymbolData value "hello"@en',
+        emmo.hasSymbolData.value(locstr("hello", "en")),
+    )
+
+
+if __name__ == "__main__":
+    test_manchester()