Fix the last two M12 bugs and lift the reconciliation skip list

- bound given names join freely in FAMILY_COMMA's post-comma segment
  (v1 reserve_last=False, parser.py:1366): 'salem, abdul salam' ->
  given 'abdul salam'. _group_segment takes the per-segment
  bound_required (3 main / 2 post-comma / 0 family segment, which v1
  never bound-joined).
- middle_as_family renders in v1's prepend order: folded tokens carry
  vocab:folded-middle and the family view orders them before the
  original family tokens (spans cannot reorder, anti-#100) --
  'Hassan, Mohamad Ahmad Ali' -> family 'Ahmad Ali Hassan', matching
  the no-comma and rotated forms. The stage-ownership contract gains
  tags for post_rules accordingly.

With these, every v1 test file is reconciled: the conftest
collect_ignore_glob is GONE (the mypy exclude went earlier). The full
v1 suite runs against the facade: 1160 passed, 12 xfailed.

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

Author: derek73

nameparser/_pipeline/_group.py

@@ -74,6 +74,7 @@ def _is_rootname(piece: Sequence[int], ptags: Set[str],
 
 def _group_segment(seg: tuple[int, ...], additional: int,
                    tokens: Sequence[WorkToken],
+                   bound_required: int = 3,
                    ) -> tuple[list[Piece], list[set[str]]]:
     pieces: list[Piece] = [[i] for i in seg]
     ptags: list[set[str]] = [set() for _ in seg]
@@ -156,18 +157,23 @@ def merge(lo: int, hi: int, add: Set[str] = frozenset(),
                 j += 1
             merge(k, j, drop={"prefix"})
             k += 1
-        # bound given names: first non-title piece joins the next when
-        # enough rootname pieces remain (v1 reserve_last)
+        # bound given names: first non-title piece joins the next.
+        # bound_required is v1's reserve_last: 3 keeps a family piece in
+        # reserve (main segments); 2 joins freely (FAMILY_COMMA's
+        # post-comma segment, v1 reserve_last=False -- 'salem, abdul
+        # salam' -> given 'abdul salam'); 0 disables (family segment,
+        # which v1 never bound-joined).
         first_name_k = next(
             (k for k in range(len(pieces)) if not title(k)), None)
-        if (first_name_k is not None
+        if (bound_required
+                and first_name_k is not None
                 and first_name_k + 1 < len(pieces)
                 and len(pieces[first_name_k]) == 1
                 and "vocab:bound-given"
                 in tokens[pieces[first_name_k][0]].tags):
             non_suffix = sum(1 for k in range(len(pieces))
                              if not title(k) and not suffix(k))
-            if non_suffix >= 3:
+            if non_suffix >= bound_required:
                 merge(first_name_k, first_name_k + 2)
     return pieces, ptags
 
@@ -186,8 +192,14 @@ def group(state: ParseState) -> ParseState:
     cores = delimiter_cores(state.policy.extra_suffix_delimiters)
     tail_start = {Structure.SUFFIX_COMMA: 1,
                   Structure.FAMILY_COMMA: 2}.get(state.structure)
+    family_comma = state.structure is Structure.FAMILY_COMMA
     for seg_idx, seg in enumerate(state.segments):
-        pieces, ptags = _group_segment(seg, additional, tokens)
+        if family_comma:
+            bound_required = 2 if seg_idx == 1 else 0
+        else:
+            bound_required = 3
+        pieces, ptags = _group_segment(seg, additional, tokens,
+                                       bound_required)
         if tail_start is not None and seg_idx >= tail_start:
             # v1 renders each tail COMMA SEGMENT as one suffix entry
             # ('Smith, V MD' -> suffix 'V MD'); a delimiter core inside

nameparser/_pipeline/_post_rules.py

@@ -118,8 +118,12 @@ def post_rules(state: ParseState) -> ParseState:
             _retag(tokens, f, Role.MIDDLE)
             _retag(tokens, g, Role.FAMILY)
     # rule 4: opt-in fold of middles into family (v1
-    # handle_middle_name_as_last; span order reproduces v1's prepend)
+    # handle_middle_name_as_last). v1 PREPENDED middle_list to
+    # last_list; spans cannot reorder (anti-#100), so folded tokens
+    # carry a tag and the family views order them first.
     if state.policy.middle_as_family:
         for i in _idx(tokens, Role.MIDDLE):
-            _retag(tokens, i, Role.FAMILY)
+            tokens[i] = dataclasses.replace(
+                tokens[i], role=Role.FAMILY,
+                tags=tokens[i].tags | {"vocab:folded-middle"})
     return dataclasses.replace(state, tokens=tuple(tokens))

nameparser/_types.py

@@ -306,6 +306,7 @@ def _text_for(self, *roles: Role, tag: str | None = None,
                   without_tag: str | None = None) -> str:
         suffix_join = roles == (Role.SUFFIX,)
         parts: list[str] = []
+        folded: list[str] = []
         for tok in self.tokens:
             if tok.role not in roles:
                 continue
@@ -318,9 +319,13 @@ def _text_for(self, *roles: Role, tag: str | None = None,
             # view's ", " join does not split one credential in two
             if suffix_join and "joined" in tok.tags and parts:
                 parts[-1] += " " + tok.text
+            elif "vocab:folded-middle" in tok.tags:
+                # middle_as_family fold: v1 PREPENDED middle_list to
+                # last_list; spans cannot reorder, so the view does
+                folded.append(tok.text)
             else:
                 parts.append(tok.text)
-        return (", " if suffix_join else " ").join(parts)
+        return (", " if suffix_join else " ").join(folded + parts)
 
     @property
     def title(self) -> str:

tests/conftest.py

@@ -6,33 +6,6 @@
 
 from nameparser.config import CONSTANTS
 
-# TEMPORARY (Task M12, final remnant): the whole v1 suite is reconciled
-# against the 2.0 facade and mypy-clean (the [tool.mypy] exclude twin in
-# pyproject.toml is gone). Only the two files below remain skipped, each
-# held on one confirmed, repro'd parity bug; delete this list (and the
-# entries' tests will run) once those land. Must be GONE before this
-# branch leaves draft.
-collect_ignore_glob = [
-    # test_bound_first_names.py is reconciled (bucket A: the two
-    # is_bound_first_name predicate tests died with the v1 hooks) but 1 test
-    # fails on a join-guard bug: the FAMILY_COMMA bound-given join does not
-    # fire when the join would consume the whole post-comma segment. Repro:
-    #   HumanName('salem, abdul salam') -> first='abdul', middle='salam'
-    # (v1: first='abdul salam'; three-token 'salem, abdul salam ahmed'
-    # joins fine). Remove once fixed.
-    "test_bound_first_names.py",
-    # test_middle_name_as_last.py needs NO bucket edits but 4 tests fail on
-    # a fold-order bug: v1's middle_name_as_last fold PREPENDED middle_list
-    # to last_list, so comma and rotated forms converge on the no-comma
-    # result; v2 renders the folded family in token order. Repros:
-    #   HumanName('Hassan, Mohamad Ahmad Ali', constants=
-    #     Constants(middle_name_as_last=True)) -> last='Hassan Ahmad Ali'
-    #   (v1: 'Ahmad Ali Hassan');
-    #   with patronymic_name_order too: 'Ivanov Petr Sergeyevich' ->
-    #   last='Ivanov Sergeyevich' (v1: 'Sergeyevich Ivanov').
-    # Remove once fixed.
-    "test_middle_name_as_last.py",
-]
 
 # Scalar (non-collection) config attributes that individual tests mutate on the
 # global CONSTANTS singleton. Several tests change these without restoring them;

tests/v2/cases.py

@@ -41,6 +41,17 @@ class Case:
          {"given": "John", "family": "Smith"}),
     Case("suffix_comma", "John Smith, PhD",
          {"given": "John", "family": "Smith", "suffix": "PhD"}),
+    Case("bound_given_whole_segment", "salem, abdul salam",
+         {"given": "abdul salam", "family": "salem"},
+         notes="v1 joins bound given names freely in the post-comma "
+               "segment (reserve_last=False, parser.py:1366) -- even "
+               "when the join consumes the whole segment"),
+    Case("middle_as_family_fold_order", "Hassan, Mohamad Ahmad Ali",
+         {"given": "Mohamad", "family": "Ahmad Ali Hassan"},
+         policy=Policy(middle_as_family=True),
+         notes="v1 PREPENDED middle_list to last_list; folded tokens "
+               "carry vocab:folded-middle and the family view orders "
+               "them first (spans cannot reorder)"),
     Case("ambiguous_surname_acronyms", "Jack Ma",
          {"given": "Jack", "family": "Ma"},
          notes="'ma'/'do' joined suffix_acronyms_ambiguous: common "

tests/v2/pipeline/test_state.py

@@ -63,7 +63,9 @@ def test_stage_field_ownership() -> None:
         "classify": {"tags"},
         "group": {"tags", "role"},
         "assign": {"role"},
-        "post_rules": {"role"},
+        # post_rules also tags: the middle_as_family fold marks folded
+        # tokens vocab:folded-middle for the family view's prepend order
+        "post_rules": {"role", "tags"},
     }
     for case in CASES:
         state = ParseState(original=case.text, lexicon=_Lexicon.default(),