[7.x] notation3.py: don't normalize float representation (#3221)

* 7.1.4 pre-release

* fix namespace prefixes in longturtle serialization (#3134)

Co-authored-by: Daan de Schepper <[email protected]>

* Fix failing webtest (#3194)

* test: fix failing webtest

Fixes #3192

* Revert "remove old hacks against 2to3 (#3076)" (#3195)

This reverts commit b74c657.

* Specify `Optional` parameters in `Graph.triples_choices` (#3075)

* Specify `Optional` parameters in `Graph.triples_choices`

The two non-list parameters can be `None`, but this is not reflected in
the type hint.

Also introduces a type alias to simplify method signatures.

* style: remove unused imports

---------

Co-authored-by: Nicholas Car <[email protected]>
Co-authored-by: Edmond Chuc <[email protected]>
Co-authored-by: Edmond Chuc <[email protected]>

* feat: canonicalization with longturtle serializer now optional (#3197)

* feat: canonicalization with longturtle serializer now optional

Fixes #3196

* docs: fix docs build error by removing py obj reference to canon

* added n3 test to check for internal float normalization

made as a separate commit to illustrate the old broken behavior priro to
the fix in the next commit

* notation3.py: don't normalize float representation

fix behavior of the n3 parser family to avoid normalizing raw float
string representation which makes it impossible to roundtrip the exact
original string representation of e.g. 1e10

* test: refactor test_float_no_norm to use pytest parametrization

* style: add noqa to sfloat class

* chore: remove unused mypy type ignore

* docs: fix sfloat reference

* fix: sphinx nitpicky reference

---------

Co-authored-by: Nicholas Car <[email protected]>
Co-authored-by: Daan de Schepper <[email protected]>
Co-authored-by: Sigmund Lahn <[email protected]>
Co-authored-by: Nicholas Car <[email protected]>
Co-authored-by: Tom Gillespie <[email protected]>

Author: 5 people

rdflib/plugins/parsers/notation3.py

@@ -83,6 +83,7 @@
     "Formula",
     "RDFSink",
     "SinkParser",
+    "sfloat",
 ]
 
 from rdflib.parser import Parser
@@ -380,6 +381,10 @@ def unicodeExpand(m: Match) -> str:
 langcode = re.compile(r"[a-zA-Z0-9]+(-[a-zA-Z0-9]+)*")
 
 
+class sfloat(str):  # noqa: N801
+    """don't normalize raw XSD.double string representation"""
+
+
 class SinkParser:
     def __init__(
         self,
@@ -1528,7 +1533,7 @@ def nodeOrLiteral(self, argstr: str, i: int, res: MutableSequence[Any]) -> int:
                 m = exponent_syntax.match(argstr, i)
                 if m:
                     j = m.end()
-                    res.append(float(argstr[i:j]))
+                    res.append(sfloat(argstr[i:j]))
                     return j
 
                 m = decimal_syntax.match(argstr, i)
@@ -1911,7 +1916,7 @@ def makeStatement(
     def normalise(
         self,
         f: Optional[Formula],
-        n: Union[Tuple[int, str], bool, int, Decimal, float, _AnyT],
+        n: Union[Tuple[int, str], bool, int, Decimal, sfloat, _AnyT],
     ) -> Union[URIRef, Literal, BNode, _AnyT]:
         if isinstance(n, tuple):
             return URIRef(str(n[1]))
@@ -1931,7 +1936,7 @@ def normalise(
             s = Literal(value, datatype=DECIMAL_DATATYPE)
             return s
 
-        if isinstance(n, float):
+        if isinstance(n, sfloat):
             s = Literal(str(n), datatype=DOUBLE_DATATYPE)
             return s
 
@@ -1947,7 +1952,7 @@ def normalise(
         #    f.universals[n] = f.newBlankNode()
         #    return f.universals[n]
         # type error: Incompatible return value type (got "Union[int, _AnyT]", expected "Union[URIRef, Literal, BNode, _AnyT]")  [return-value]
-        return n  # type: ignore[return-value]
+        return n
 
     def intern(self, something: _AnyT) -> _AnyT:
         return something

test/test_n3.py

@@ -251,6 +251,23 @@ def test_empty_prefix(self):
             g2
         ), "Document with declared empty prefix must match default #"
 
+    @pytest.mark.parametrize(
+        "do_normalize_literal, expected_result",
+        [(True, {"1.0", "10000000000.0"}), (False, {"1e10", "1e0"})],
+    )
+    def test_float_no_norm(self, do_normalize_literal, expected_result):
+        import rdflib
+
+        original_normalize_literal = rdflib.NORMALIZE_LITERALS
+        try:
+            rdflib.NORMALIZE_LITERALS = do_normalize_literal
+            g1 = Graph()
+            g1.parse(data=":a :b 1e10, 1e0 .", format="n3")
+            values = set(str(o) for o in g1.objects())
+            assert values == expected_result
+        finally:
+            rdflib.NORMALIZE_LITERALS = original_normalize_literal
+
 
 class TestRegularExpressions:
     def test_exponents(self):