state_projection: single-source the backlog-hygiene bullet grammar

loopx/state_projection.py

@@ -5,7 +5,10 @@
 import re
 
 from .control_plane.goals.active_state_metadata import todo_role_for_heading
-from .control_plane.status.active_state_projection import SECTION_HEADING_PATTERN
+from .control_plane.status.active_state_projection import (
+    BACKLOG_HYGIENE_BULLET_PATTERN,
+    SECTION_HEADING_PATTERN,
+)
 from .control_plane.todos.contract import (
     TODO_TASK_PATTERN,
     build_todo_id,
@@ -45,7 +48,6 @@
 ACTIVE_STATE_PROJECTION_DIAGNOSTICS_SCHEMA_VERSION = "active_state_projection_diagnostics_v0"
 TODO_ITEM_SCHEMA_VERSION = "todo_item_v0"
 
-BULLET_PATTERN = re.compile(r"^\s*(?:[-*]|\d+[.)])\s+(.+?)\s*$")
 PRIORITY_PATTERN = re.compile(r"^\[(P[0-4])\]\s+(.+)$", re.IGNORECASE)
 NEXT_ACTION_EXECUTABLE_PATTERN = re.compile(
     r"(?i)\b(?:run|repair|fix|implement|add|update|write|record|validate|"
@@ -399,7 +401,7 @@ def append_entry(parts: list[str]) -> None:
     for line in lines:
         if line.strip().startswith("<!--"):
             continue
-        bullet = BULLET_PATTERN.match(line)
+        bullet = BACKLOG_HYGIENE_BULLET_PATTERN.match(line)
         if bullet:
             if current:
                 append_entry(current)

tests/architecture/test_state_projection_bullet_pattern_owner.py

@@ -0,0 +1,229 @@
+"""Guard one owner for the backlog-hygiene bullet grammar (#4447).
+
+``BACKLOG_HYGIENE_BULLET_PATTERN`` in ``control_plane/status/active_state_projection.py``
+is the compiled grammar for a Next Action / Operating Lessons bullet
+(``^\\s*(?:[-*]|\\d+[.)])\\s+(.+?)\\s*$``). ``state_projection.py`` restated the *same*
+compiled regex under a different name (``BULLET_PATTERN``) to split its Next Action
+section into entries. ``state_projection`` *already imports* ``SECTION_HEADING_PATTERN``
+from that very owner module, so borrowing one more compiled pattern adds no new
+import edge and cannot introduce a cycle. The restated regex is retired and
+``_section_entries`` now matches against the single owner.
+
+Two existing machines miss this. The by-name semantic inventory only catalogues
+``UPPER_NAME`` carriers and would not pair two differently named bindings; the
+status-facade vocabulary-owners machine is also name-keyed, so a copy renamed to
+``BULLET_PATTERN`` slips past it (that machine is extended separately to record the
+consumer as an importer of the owner name). This guard keys on the **compiled
+pattern literal** instead, so a renamed restatement is caught, while the genuinely
+different neighbours -- the owner's own ``SECTION_HEADING_PATTERN`` and the
+consumer's ``PRIORITY_PATTERN`` -- are never swept.
+
+Two layers: a whole-tree compiled-pattern census (only the owner compiles it), and
+behavioural parity through the consumer's public ``active_state_next_action_entries``
+gate -- marker stripping, non-bullet passthrough and continuation grouping -- so a
+migration that silently changed which lines count as bullets fails here.
+"""
+
+from __future__ import annotations
+
+import ast
+import pathlib
+import re
+
+import pytest
+
+from loopx import state_projection as consumer
+from loopx.control_plane.status import active_state_projection as owner
+from loopx.control_plane.todos.contract import compact_todo_text
+
+REPO_ROOT = pathlib.Path(__file__).resolve().parents[2]
+PACKAGE_ROOT = REPO_ROOT / "loopx"
+OWNER_PATH = "loopx/control_plane/status/active_state_projection.py"
+CONSUMER_PATH = "loopx/state_projection.py"
+
+PATTERN = r"^\s*(?:[-*]|\d+[.)])\s+(.+?)\s*$"
+SECTION_HEADING_PATTERN = r"^##+\s+(.+?)\s*$"
+PRIORITY_PATTERN = r"^\[(P[0-4])\]\s+(.+)$"
+
+
+def _compiled_pattern_sites(source: str, pattern: str) -> list[int]:
+    """Every ``re.compile(<literal>)`` call in the source whose pattern matches.
+
+    Keys on the pattern literal, so a restatement under any name (or with extra
+    flags) is still one site; a bare pattern string that is never compiled is not
+    a compiled-regex restatement and is left alone.
+    """
+
+    tree = ast.parse(source)
+    sites: list[int] = []
+    for node in ast.walk(tree):
+        if not isinstance(node, ast.Call):
+            continue
+        fname = getattr(node.func, "id", None) or getattr(node.func, "attr", None)
+        if fname != "compile":
+            continue
+        if (
+            node.args
+            and isinstance(node.args[0], ast.Constant)
+            and node.args[0].value == pattern
+        ):
+            sites.append(node.lineno)
+    return sites
+
+
+@pytest.fixture(scope="module")
+def tree_offenders() -> dict[str, list[int]]:
+    offenders: dict[str, list[int]] = {}
+    for path in sorted(PACKAGE_ROOT.rglob("*.py")):
+        relative = path.relative_to(REPO_ROOT).as_posix()
+        if relative == OWNER_PATH:
+            continue
+        sites = _compiled_pattern_sites(path.read_text(encoding="utf-8"), PATTERN)
+        if sites:
+            offenders[relative] = sites
+    return offenders
+
+
+def test_only_the_owner_compiles_the_bullet_pattern(tree_offenders) -> None:
+    assert tree_offenders == {}, (
+        f"a restatement of the backlog-hygiene bullet grammar remains: {tree_offenders}"
+    )
+
+
+def test_the_owner_compiles_the_recorded_pattern_once() -> None:
+    assert owner.BACKLOG_HYGIENE_BULLET_PATTERN.pattern == PATTERN
+    source = (REPO_ROOT / OWNER_PATH).read_text(encoding="utf-8")
+    assert len(_compiled_pattern_sites(source, PATTERN)) == 1
+
+
+def test_state_projection_borrows_the_owner_object() -> None:
+    assert (
+        consumer.BACKLOG_HYGIENE_BULLET_PATTERN is owner.BACKLOG_HYGIENE_BULLET_PATTERN
+    )
+
+
+def test_state_projection_retired_its_restated_regex() -> None:
+    source = (REPO_ROOT / CONSUMER_PATH).read_text(encoding="utf-8")
+    assert _compiled_pattern_sites(source, PATTERN) == []
+    # The retired binding was the standalone name BULLET_PATTERN; the borrowed
+    # owner name BACKLOG_HYGIENE_BULLET_PATTERN contains it as a substring, so a
+    # word boundary is needed to avoid matching the import.
+    assert re.search(r"\bBULLET_PATTERN\b", source) is None
+    assert "BACKLOG_HYGIENE_BULLET_PATTERN" in source
+
+
+def test_the_section_heading_neighbour_is_a_different_pattern() -> None:
+    # SECTION_HEADING_PATTERN shares the owner module and is also borrowed, but it
+    # is a different grammar; the bullet census must not sweep it.
+    assert owner.SECTION_HEADING_PATTERN.pattern == SECTION_HEADING_PATTERN
+    assert SECTION_HEADING_PATTERN != PATTERN
+
+
+def test_the_consumer_keeps_its_own_priority_neighbour() -> None:
+    assert consumer.PRIORITY_PATTERN.pattern == PRIORITY_PATTERN
+    assert PRIORITY_PATTERN != PATTERN
+
+
+# --- behavioural parity through the consumer's public Next Action gate ------------------
+
+BULLET_LINES = (
+    "- foo",
+    "* foo",
+    "1. foo",
+    "1) foo",
+    "12) multi word task",
+    "  - leading spaces",
+    "- trailing spaces  ",
+    "\t- tabbed",
+    "- [P1] do thing",
+    "- 日本語のタスク",
+)
+
+NON_BULLET_LINES = (
+    "plain prose line",
+    "-foo",
+    "-",
+    "1.foo",
+    "> a quote",
+    "a - b",
+    "   ",
+)
+
+
+def _next_action(line: str) -> list[str]:
+    state = "## Next Action\n" + line + "\n"
+    return consumer.active_state_next_action_entries(state, limit=None, text_limit=None)
+
+
+@pytest.mark.parametrize("line", BULLET_LINES)
+def test_bullet_marker_is_stripped_exactly_as_the_owner(line: str) -> None:
+    match = owner.BACKLOG_HYGIENE_BULLET_PATTERN.match(line)
+    assert match is not None
+    expected = compact_todo_text(match.group(1))
+    assert _next_action(line) == ([expected] if expected else [])
+
+
+@pytest.mark.parametrize("line", NON_BULLET_LINES)
+def test_non_bullet_line_keeps_its_text(line: str) -> None:
+    assert owner.BACKLOG_HYGIENE_BULLET_PATTERN.match(line) is None
+    stripped = line.strip()
+    expected = compact_todo_text(stripped) if stripped else ""
+    assert _next_action(line) == ([expected] if expected else [])
+
+
+def test_bullet_groups_an_indented_continuation() -> None:
+    state = "## Next Action\n- first task\n  continued detail\n"
+    got = consumer.active_state_next_action_entries(state, limit=None, text_limit=None)
+    match = owner.BACKLOG_HYGIENE_BULLET_PATTERN.match("- first task")
+    assert match is not None
+    expected = compact_todo_text(match.group(1) + " continued detail")
+    assert got == [expected]
+
+
+# --- the census bites every restatement form and leaves the others ----------------------
+
+POISON_CORPUS = (
+    ("re.compile of the pattern", "X = re.compile(" + repr(PATTERN) + ")\n", 1),
+    (
+        "bare compile from re",
+        "from re import compile\nX = compile(" + repr(PATTERN) + ")\n",
+        1,
+    ),
+    (
+        "nested in a function",
+        "def f():\n    return re.compile(" + repr(PATTERN) + ")\n",
+        1,
+    ),
+    (
+        "flags do not hide the restated pattern",
+        "X = re.compile(" + repr(PATTERN) + ", re.IGNORECASE)\n",
+        1,
+    ),
+    (
+        "the section-heading grammar is different",
+        "X = re.compile(" + repr(SECTION_HEADING_PATTERN) + ")\n",
+        0,
+    ),
+    (
+        "the priority grammar is different",
+        "X = re.compile(" + repr(PRIORITY_PATTERN) + ")\n",
+        0,
+    ),
+    ("a bare pattern string is not a compiled regex", "X = " + repr(PATTERN) + "\n", 0),
+    (
+        "importing the owner is not a restatement",
+        "from .control_plane.status.active_state_projection import "
+        "BACKLOG_HYGIENE_BULLET_PATTERN\n",
+        0,
+    ),
+    ("alias rebind is not a restatement", "BULLET_PATTERN = BACKLOG_HYGIENE_BULLET_PATTERN\n", 0),
+)
+
+
+@pytest.mark.parametrize(
+    "label,source,expected", POISON_CORPUS, ids=[case[0] for case in POISON_CORPUS]
+)
+def test_the_census_names_every_form_and_leaves_the_others(
+    label: str, source: str, expected: int
+) -> None:
+    assert len(_compiled_pattern_sites(source, PATTERN)) == expected, label

tests/control_plane/test_status_facade_vocabulary_owners.py

@@ -54,7 +54,10 @@
         (status,),
     ),
     "AUTONOMOUS_REPLAN_SCHEMA_VERSION": (autonomous_replan_projection, (status,)),
-    "BACKLOG_HYGIENE_BULLET_PATTERN": (active_state_projection, (status,)),
+    "BACKLOG_HYGIENE_BULLET_PATTERN": (
+        active_state_projection,
+        (state_projection, status),
+    ),
     "BACKLOG_HYGIENE_HINT_PATTERN": (active_state_projection, (status,)),
     "BACKLOG_HYGIENE_SECTION_HEADINGS": (active_state_projection, (status,)),
     "DEAD_MONITOR_REPEAT_SCHEMA_VERSION": (autonomous_replan_projection, (status,)),