fix(eval): strip query strings when extracting Wikipedia titles from URLs

eval/lib/retrieval.py

@@ -900,7 +900,7 @@ def _extract_wiki_title(self, url: str) -> str | None:
         # Match patterns like:
         # https://en.wikipedia.org/wiki/Python_(programming_language)
         # https://zh.wikipedia.org/wiki/Artificial_intelligence
-        pattern = r"https?://[a-z]{2,3}\.wikipedia\.org/wiki/(.+?)(?:#.*)?$"
+        pattern = r"https?://[a-z]{2,3}\.wikipedia\.org/wiki/(.+?)(?:[?#].*)?$"
         match = re.match(pattern, url)
         if match:
             title = unquote(match.group(1))

tests/test_wiki_title_extraction.py

@@ -0,0 +1,35 @@
+"""Regression tests for Wikipedia title extraction from URLs."""
+
+from pathlib import Path
+from runpy import run_path
+
+RETRIEVAL = run_path(str(Path(__file__).parents[1] / "eval" / "lib" / "retrieval.py"))
+extract = RETRIEVAL["WikipediaAPIRetriever"]._extract_wiki_title
+
+
+def test_plain_url_title_unchanged():
+    assert extract(None, "https://en.wikipedia.org/wiki/Albert_Einstein") == (
+        "Albert Einstein"
+    )
+
+
+def test_query_string_is_stripped():
+    assert extract(
+        None, "https://en.wikipedia.org/wiki/Albert_Einstein?wprov=rarw1"
+    ) == ("Albert Einstein")
+
+
+def test_fragment_is_stripped():
+    assert extract(
+        None, "https://en.wikipedia.org/wiki/Albert_Einstein#Early_life"
+    ) == ("Albert Einstein")
+
+
+def test_percent_encoded_title_is_decoded():
+    assert extract(
+        None, "https://en.wikipedia.org/wiki/Python_%28programming_language%29"
+    ) == ("Python (programming language)")
+
+
+def test_non_wiki_url_returns_none():
+    assert extract(None, "https://example.com/wiki/Albert_Einstein") is None