Tighten the compound rule to names the order flip can reach

Review found the delimiter-only regex claiming more than its comment
said: 「」『』 sit in CJK Symbols and Punctuation, OUTSIDE every
classified span, so 'John 「Jack」 Kennedy' matched and a bare
first/last regression on a Latin name classified as intended. The
rule now requires a classified codepoint alongside the delimiter
(both lookaheads), drops the unearned middle field (all three claimed
diffs touch exactly first/last/nickname), and its comment owns the
one absorption subset-matching cannot avoid -- a nickname-only diff
on a nakaguro name -- instead of claiming 'cannot absorb'.

The second lookahead is the toml's second hand copy of the script
spans, so it gets the same sync pin as the first, which also pins the
delimiter set and records the #271/#272-substring taboo where a rule
author will look. Plus the smaller review items: the provenance
comment now credits the parity rows (the corpus members that pin
NON-diffs), the positional note within the name_regex tier is stated,
test_regex_sync's module docstring counts its outside-the-package
tests correctly, and the README comma splice is mended.

Noted, not changed: the Latin probes the review used do not fall to
UNEXPLAINED without the compound rule -- the fields-only
fix(suffix-routing) rule absorbs any bare first/last diff on any
name, and always has. Pre-existing fields-tier looseness, outside
this change.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: derek73

tests/v2/test_regex_sync.py

@@ -12,8 +12,9 @@
 Layering is the usual reason for a copy but not the only one, so this
 module's scope is the PROMISE rather than that one pair of packages:
 the comma-set pin below reads _pipeline._state instead of config, and
-the last test reaches outside the package altogether, to a TOML file
-that could not import a Python constant if it wanted to.
+the last three tests reach outside the package altogether -- two to a
+TOML file that could not import a Python constant if it wanted to,
+one to a generated corpus whose generator can, and must stay run.
 """
 import importlib.util
 import json
@@ -236,6 +237,42 @@ def test_differential_cjk_rule_matches_the_script_ranges() -> None:
         f"_SCRIPT_RANGES' BMP spans are {sorted(expected)}")
 
 
+def test_differential_compound_rule_matches_the_script_ranges() -> None:
+    """The fix(cjk-delimited-nickname) rule (#295) carries the SECOND
+    hand copy of the script spans in the toml: its require-a-classified-
+    codepoint lookahead exists so the delimiter set alone cannot claim a
+    Latin name's first/last regression ('John 「Jack」 Kennedy' -- the
+    corner brackets sit outside every classified span). Pin that copy to
+    the table exactly as the canonical CJK rule's is, and pin the
+    delimiter set itself, so widening either (to ASCII quotes, say) is
+    an explicit decision here rather than a silent absorption change.
+
+    Selection note for future rule authors: the canonical-CJK-rule pin
+    above selects by the literal '#271'/'#272' substrings and asserts
+    uniqueness -- this rule's slug avoids them on purpose, and any new
+    rule's must too.
+    """
+    toml_path = (Path(__file__).parents[2] / "tools" / "differential"
+                 / "expected_changes.toml")
+    rules = tomllib.loads(toml_path.read_text())["change"]
+    matched = [r for r in rules if "cjk-delimited-nickname" in r["issue"]]
+    assert len(matched) == 1
+    regex = matched[0]["name_regex"]
+    assert "[「」『』・･]" in regex, (
+        "the compound rule's delimiter set changed; decide deliberately")
+    declared = {
+        (int(lo, 16), int(hi, 16))
+        for lo, hi in re.findall(r"\\u([0-9A-Fa-f]{4})-\\u([0-9A-Fa-f]{4})",
+                                 regex)}
+    expected = {span
+                for spans in _policy._SCRIPT_RANGES.values()
+                for span in spans
+                if span[1] <= 0xFFFF} | {(0xFF65, 0xFF65)}
+    assert declared == expected, (
+        f"compound rule's codepoint lookahead declares {sorted(declared)}; "
+        f"_SCRIPT_RANGES' BMP spans are {sorted(expected)}")
+
+
 def test_cjk_corpus_matches_the_case_table() -> None:
     """corpus_cjk.jsonl is GENERATED, not curated (#295): every
     distinct case-table text bearing a codepoint the script table

tools/differential/README.md

@@ -51,10 +51,10 @@ that stops being run. Pass `--corpus PATH` (repeatable) to narrow it.
 | `corpus_cjk.jsonl` | the CJK-bearing rows of `tests/v2/cases.py`, via `build_cjk_corpus.py` (#295) | anything the case table itself missed — it re-witnesses reviewed expectations at the 1.4 boundary rather than discovering new shapes |
 
 They are deliberately separate rather than merged: `corpus.jsonl` is
-reproducible forever from an immutable git ref, the issue tracker is
-mutable, so regenerating `corpus_issues.jsonl` is an explicit,
-reviewable act that can only add names — and the CJK corpus exists
-because BOTH of those are structurally blind to unspaced CJK (v1's
+reproducible forever from an immutable git ref, while the issue
+tracker is mutable, so regenerating `corpus_issues.jsonl` is an
+explicit, reviewable act that can only add names — and the CJK corpus
+exists because BOTH of those are structurally blind to unspaced CJK (v1's
 banks never tested it; `build_issues_corpus.py` requires an internal
 space, which unspaced names never have). It regenerates from the case
 table, and `tests/v2/test_regex_sync.py` pins the checked-in file

tools/differential/expected_changes.toml

@@ -45,12 +45,13 @@ issue = "fix(#271/#272) native-script CJK: family-first order, hangul segmentati
 # span.
 #
 # corpus_cjk.jsonl (#295) exists so this rule fires in every real
-# run: it derives from the fix(#271)/fix(#272)/fix(shime-mark) rows of
-# tests/v2/cases.py (build_cjk_corpus.py; the 〆 span rides in on the
-# same HAN entry), closing the blind spot the other two corpora have
-# by construction -- v1's banks had no reason to test CJK, and
-# build_issues_corpus.py requires an internal space, which unspaced
-# names never have.
+# run: build_cjk_corpus.py harvests every CJK-bearing text in
+# tests/v2/cases.py -- the fix(#271)/fix(#272)/fix(shime-mark) rows
+# AND the parity rows, whose value here is pinning NON-diffs (マイケル
+# must not change against 1.4) -- closing the blind spot the other
+# two corpora have by construction: v1's banks had no reason to test
+# CJK, and build_issues_corpus.py requires an internal space, which
+# unspaced names never have.
 name_regex = "[\\u3005-\\u3006\\u3040-\\u309F\\u30A0-\\u30FF\\u3400-\\u4DBF\\u4E00-\\u9FFF\\uF900-\\uFAFF\\uAC00-\\uD7A3\\uFF65-\\uFF65]"
 fields = ["first", "middle", "last"]
 
@@ -131,14 +132,25 @@ issue = "fix(cjk-delimited-nickname) delimiter recognition compounds with the CJ
 # first/last. Neither single-change rule may claim the union (the
 # delimiter rule's fields exclude first/last, the CJK rule's exclude
 # nickname -- each on purpose, so a lone regression in the other's
-# fields stays loud). Scoped to the CJK-typographic delimiters and
-# the nakaguro only: a name containing 「」『』 or ・ is CJK-adjacent by
-# construction, so this cannot absorb a Latin delimiter diff. The
-# issue slug deliberately avoids the literal #271/#272 strings --
-# test_regex_sync's differential pin selects THE CJK rule by those
-# substrings and asserts it is unique.
-name_regex = "[「」『』・･]"
-fields = ["first", "middle", "last", "nickname"]
+# fields stays loud). Both lookaheads are required: the delimiters
+# alone would match 'John 「Jack」 Kennedy' -- the brackets sit in CJK
+# Symbols and Punctuation, OUTSIDE every classified span, so a Latin
+# name can carry them -- and would then absorb a bare first/last
+# regression on it. Requiring a classified codepoint too confines the
+# rule to names the order flip can actually reach. The second
+# lookahead's class is the same hand copy of _SCRIPT_RANGES the rule
+# above carries, pinned by the same sync test. One absorption is
+# inherent and accepted: a nickname-ONLY diff on a nakaguro name
+# ('マイケル・ジャクソン') classifies here, because subset-matching
+# cannot REQUIRE a field to have changed. Positional note: this rule
+# overlaps the CJK rule above and the delimiter rule below; within
+# the name_regex tier file order decides which label a diff reports
+# under, so it sits after the tighter single-change rules on purpose.
+# The issue slug deliberately avoids the literal #271/#272 strings --
+# test_regex_sync's differential pin selects the canonical CJK rule
+# by those substrings and asserts it is unique.
+name_regex = "(?s)(?=.*[「」『』・･])(?=.*[\\u3005-\\u3006\\u3040-\\u309F\\u30A0-\\u30FF\\u3400-\\u4DBF\\u4E00-\\u9FFF\\uF900-\\uFAFF\\uAC00-\\uD7A3\\uFF65-\\uFF65])"
+fields = ["first", "last", "nickname"]
 
 [[change]]
 issue = "feat(#269) non-Latin titles/conjunctions recognized"