fix: correct suffix boundary lookup for prefixed last names (#100) (#179)

* fix: correct suffix boundary lookup for prefixed last names (#100)

The prefix-joining loop located the suffix stop boundary with a
value-based pieces.index() that searched from position 0. When a token
value repeated (a trailing title that is also a suffix acronym, e.g.
the second 'dr' in 'dr Vincent van Gogh dr'), it matched the leading
occurrence, producing an empty slice that duplicated pieces and
corrupted the middle name. Constrain the lookup to start at i + 1,
consistent with the sibling next_prefix lookup.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* test: guard against #108 exponential blow-up on repeated prefixes

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* review: improve comment accuracy and test coverage

- Fix inline comment in join_on_conjunctions: clarify that filter()
  finds the value in pieces[i+1:] but index() searches from 0 by
  default, and drop the misleading "title" framing (the token only
  needs to satisfy is_suffix, not is_title)
- Add test for two-word prefix collision ("van der") — different loop
  iteration count than the single-word case
- Add test with a genuine middle name alongside the repeated token,
  since the pre-fix bug corrupted the middle field specifically
- Add @pytest.mark.timeout(2) to the #108 guard so the timeout is
  enforced locally and in CI, not just by CI job limits
- Assert hn.last contains "Berg" in the #108 guard to catch silent
  last-name corruption
- Add pytest-timeout dev dependency
- Resolve pre-existing stash conflict in docs/resources.rst (keep upstream)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: derek73

nameparser/parser.py

@@ -980,7 +980,12 @@ def join_on_conjunctions(self, pieces: list[str], additional_parts_count: int =
                     try:
                         # if there are no more prefixes, look for a suffix to stop at
                         stop_at = next(iter(filter(self.is_suffix, pieces[i + 1:])))
-                        j = pieces.index(stop_at)
+                        # search from i + 1: filter() finds the value of stop_at
+                        # in pieces[i+1:] but pieces.index() without a start
+                        # argument searches from 0, so an earlier occurrence of
+                        # the same token (e.g. a suffix token that also appears
+                        # before the prefix) would be matched instead.
+                        j = pieces.index(stop_at, i + 1)
                         new_piece = ' '.join(pieces[i:j])
                         pieces = pieces[:i] + [new_piece] + pieces[j:]
                     except StopIteration:

pyproject.toml

@@ -43,7 +43,8 @@ dev = [
   "dill (>=0.2.5)",
   "sphinx (>=8)",
   "mypy (>=2.1)",
-  "ruff (>=0.15)"
+  "ruff (>=0.15)",
+  "pytest-timeout>=2.4.0",
 ]
 
 [tool.mypy]

tests/test_prefixes.py

@@ -1,3 +1,5 @@
+import pytest
+
 from nameparser import HumanName
 
 from tests.base import HumanNameTestBase
@@ -43,6 +45,50 @@ def test_prefix_before_two_part_last_name_with_acronym_suffix(self) -> None:
         self.m(hn.last, "von bergen wessels", hn)
         self.m(hn.suffix, "M.D.", hn)
 
+    def test_title_before_and_after_prefixed_last_name(self) -> None:
+        # Issue #100: a repeated title/suffix token ("dr") before AND after a
+        # prefixed last name used to corrupt the middle name into
+        # " dr Vincent van" because the suffix-boundary lookup matched the
+        # LEADING "dr" instead of the trailing one.
+        hn = HumanName("dr Vincent van Gogh dr")
+        self.m(hn.title, "dr", hn)
+        self.m(hn.first, "Vincent", hn)
+        self.m(hn.middle, "", hn)
+        self.m(hn.last, "van Gogh", hn)
+        self.m(hn.suffix, "dr", hn)
+
+    def test_suffix_token_collision_with_two_word_prefix(self) -> None:
+        # Same fix as #100 but with a two-word prefix ("van der"). Exercises a
+        # different iteration count through the prefix-joining loop.
+        hn = HumanName("dr Vincent van der Gogh dr")
+        self.m(hn.title, "dr", hn)
+        self.m(hn.first, "Vincent", hn)
+        self.m(hn.middle, "", hn)
+        self.m(hn.last, "van der Gogh", hn)
+        self.m(hn.suffix, "dr", hn)
+
+    def test_title_before_and_after_prefixed_last_name_with_middle(self) -> None:
+        # The pre-fix bug corrupted the middle field; verify it is not disturbed
+        # when a genuine middle name is present alongside the repeated token.
+        hn = HumanName("dr Vincent James van Gogh dr")
+        self.m(hn.title, "dr", hn)
+        self.m(hn.first, "Vincent", hn)
+        self.m(hn.middle, "James", hn)
+        self.m(hn.last, "van Gogh", hn)
+        self.m(hn.suffix, "dr", hn)
+
+    @pytest.mark.timeout(2)
+    def test_many_repeated_prefixes_does_not_blow_up(self) -> None:
+        # Issue #108: a name with a long run of repeated prefixes used to grow
+        # the pieces list exponentially and exhaust memory. The 2-second timeout
+        # enforces this locally and in CI — if the test hangs, an exponential
+        # regression has been reintroduced.
+        name = "Jan " + "van der " * 30 + "Berg"
+        hn = HumanName(name)
+        self.assertFalse(hn.unparsable)
+        self.m(hn.first, "Jan", hn)
+        self.assertIn("Berg", hn.last)
+
     def test_two_part_last_name_with_suffix_comma(self) -> None:
         hn = HumanName("pennie von bergen wessels, III")
         self.m(hn.first, "pennie", hn)