hypgraph 0.4.1__tar.gz → 0.4.2__tar.gz

{hypgraph-0.4.1 → hypgraph-0.4.2}/CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## 0.4.2 — 2026-10-25
+
+- **A question is not an assertion.** `language.interrogative` / `asserted` return nothing for a sentence that ends
+  with a question mark, and the report check skips the language and number rules on a sentence that asks or that
+  quotes one of the investigation's own questions (citations are still checked). Found in the first real HypSwarm
+  mixed-swarm run: the lead headed a section "Q2 — Is A significantly more accurate than B…?" and lost its report
+  to `REPORT_SIGNIFICANCE_UNBACKED` three times.
+
 ## 0.4.1 — 2026-10-25
 
 Three rules the hypswarm examples exposed, each with a test:

{hypgraph-0.4.1 → hypgraph-0.4.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: hypgraph
-Version: 0.4.1
+Version: 0.4.2
 Summary: The evidence layer for AI research agents: a hash-chained research record that checks every claim as it is written and derives how well each is known.
 Project-URL: Homepage, https://hypgraph.com
 Project-URL: Documentation, https://hypgraph.com/engineers/

{hypgraph-0.4.1 → hypgraph-0.4.2}/src/hypgraph/__init__.py

@@ -16,7 +16,7 @@ Standing is derived on every read, never stored; every write is an event in a SH
 their numbers; reviews pin versions; independence is counted, not assumed. See https://hypgraph.com.
 """
 
-__version__ = "0.4.1"
+__version__ = "0.4.2"
 
 from .bundle import verify_bundle  # noqa: E402
 from .graph import Author, Graph, GraphError, file_sha256  # noqa: E402

{hypgraph-0.4.1 → hypgraph-0.4.2}/src/hypgraph/language.py

@@ -64,9 +64,21 @@ _ZERO_ON_A_SCALE = re.compile(r"\b(below|above|near|to|from|of|than|at|cross(?:e
                               re.I)
 
 
+# a sentence that asks: it ends with a question mark (markdown emphasis or a closing quote may follow it)
+_INTERROGATIVE = re.compile(r"\?\s*[*_\"'’”)\]]*\s*$")
+
+
+def interrogative(text: str) -> bool:
+    """Does the sentence ask rather than assert? "Is A significantly better than B?" asserts nothing — found when a
+    lead headed a report section with the question it was answering and lost the report to the significance check."""
+    return bool(_INTERROGATIVE.search((text or "").strip()))
+
+
 def asserted(pattern: re.Pattern, text: str) -> list[str]:
     """The distinct matches of a pattern that their sentence actually asserts: not negated in their clause, not an
-    epistemic hedge ("cannot conclude"), and not a number on a scale ("below zero")."""
+    epistemic hedge ("cannot conclude"), not a number on a scale ("below zero"), and nothing in a question."""
+    if interrogative(text):
+        return []
     out = []
     for m in pattern.finditer(text or ""):
         if negated(text, m.start()):

{hypgraph-0.4.1 → hypgraph-0.4.2}/src/hypgraph/report_check.py

@@ -71,6 +71,10 @@ _NEGATOR = re.compile(r"\b(not|cannot|can't|never|no|nor|without|hardly|unlikely
                       r"doesn't|didn't|won't)\b[\w\s-]{0,24}$", re.I)
 
 
+def _norm(text: str) -> str:
+    return re.sub(r"[^a-z0-9]+", " ", (text or "").lower()).strip()
+
+
 def _own_words(sentence: str) -> str:
     """The writer's own words: quoted passages (often the very wording being corrected) blanked out."""
     return _QUOTE.sub(lambda m: " " * len(m.group(0)), sentence)
@@ -152,6 +156,10 @@ def check(g: "Graph", text: str, inv: str = "") -> dict:
         if n["code"] and (n["inv"] in invs or i in invs):
             by_code.setdefault(n["code"], []).append(i)
     plain, stat = _allowed(s, invs)
+    # the investigations' own questions: a report may quote the question it answers without asserting it
+    qtexts = [_norm(n["body"].get("text", "")) for i, n in s.nodes.items()
+              if n["type"] == "question" and (n["inv"] in invs or i in invs)]
+    qtexts = [q for q in qtexts if len(q) >= 20]
     findings: list[dict] = []
     counted = {"sentences": 0, "numbers": 0, "codes": 0}
 
@@ -170,6 +178,7 @@ def check(g: "Graph", text: str, inv: str = "") -> dict:
                 continue
             counted["sentences"] += 1
             own = _own_words(sent)
+            asks = L.interrogative(sent) or any(q in _norm(sent) for q in qtexts)
             qualified = bool(_QUALIFIED.search(sent))
             settled = _asserted(_SETTLED, own)
             cited = []
@@ -208,6 +217,8 @@ def check(g: "Graph", text: str, inv: str = "") -> dict:
                     add(ln, "warning", "REPORT_OMITS_DISPUTE", sent,
                         f"{code} is {'contested' if review == 'contested' else 'at risk'} in the record — say so where "
                         "you cite it.")
+            if asks:                     # a question, or the question's own words: citations checked, nothing asserted
+                continue
             # numbers
             for p in numbers(sent):
                 if _skip_number(p, sent) or re.search(rf"(?<![\w-])[A-Z]{{1,3}}{re.escape(p[3])}(?![\w-])", sent):

hypgraph-0.4.2/tests/test_questions_not_assertions.py

@@ -0,0 +1,43 @@
+"""0.4.2: a question is not an assertion. From the first real HypSwarm mixed-swarm run: the lead headed a report
+section with the question it was answering and the report was refused three times for 'significantly'."""
+from hypgraph import Graph, Policy
+from hypgraph import language as L
+
+
+def test_interrogatives_assert_nothing():
+    assert L.asserted(L.SIGNIFICANCE, "Is A significantly more accurate than B?") == []
+    assert L.asserted(L.SIGNIFICANCE, "**Q2 — Is A significantly more accurate than B at the two-sided 0.05 level?**") == []
+    assert L.asserted(L.ABSOLUTE, "Does it never fail?") == []
+    assert L.asserted(L.SIGNIFICANCE, "A is significantly more accurate than B.") == ["significantly"]
+    assert L.interrogative("Is it so?\"") and not L.interrogative("It is so.")
+
+
+def _record():
+    g = Graph(policy=Policy(reviewers={"critic"}))
+    inv = g.create(author="h", type="investigation", fields={"title": "Does A beat B.", "objective": "Compare A and B on the same items."})["id"]
+    q = g.create(author="h", type="question", inv=inv, code="Q2",
+                 fields={"text": "Is A significantly more accurate than B (two-sided, 0.05)?"})["id"]
+    r = g.create(author="stats", type="run", inv=inv, fields={"kind": "agent", "name": "mcnemar", "state": "succeeded"})
+    rs = g.create(author="stats", type="result", inv=inv, fields={"summary": "Compared A and B on the same 1,000 items.",
+                                                                  "values": {"a": {"k": 900, "n": 1000}, "b": {"k": 800, "n": 1000}}},
+                  links=[{"rel": "produced", "from": r["id"]}])
+    c = g.create(author="stats", type="claim", inv=inv, fields={"statement": "A is right on 900 of 1000 items and B on 800 of 1000.",
+                                                                "scope": "The 1,000 shared items."},
+                 links=[{"rel": "answers", "to": q}, {"rel": "derived_from", "to": rs["id"]}])
+    g.review(author="critic", target=c["id"], verdict="checked", method="Recounted both columns from the file.", notes="The counts hold and the claim says what the result says.")
+    return g, inv, c["code"]
+
+
+def test_a_report_may_quote_the_question_it_answers():
+    g, inv, cl = _record()
+    text = ("## Q2 — Is A significantly more accurate than B at the two-sided 0.05 level?\n\n"
+            f"A is right on 900 of 1000 items and B on 800 of 1000 ({cl}).\n\n"
+            "The question 'Is A significantly more accurate than B (two-sided, 0.05)?' is what this section answers.\n")
+    r = g.check_text(text, inv)
+    assert not [f for f in r["findings"] if f["code"] == "REPORT_SIGNIFICANCE_UNBACKED"], r["findings"]
+    assert r["ok"]
+    # the assertion itself is still caught, and an unknown code in a question heading is still an unknown code
+    bad = g.check_text("A is significantly more accurate than B.", inv)
+    assert any(f["code"] == "REPORT_SIGNIFICANCE_UNBACKED" and f["level"] == "error" for f in bad["findings"])
+    unk = g.check_text("## Is CL99 significantly better?", inv)
+    assert any(f["code"] == "REPORT_UNKNOWN_CODE" for f in unk["findings"])