fix(corpus): preserve case-sensitive exact identifiers

src/benchmark_radar/corpus.py

@@ -75,7 +75,7 @@ def _exact_candidates(item: dict[str, Any]) -> list[tuple[int, str]]:
         if host == "openreview.net":
             forum = (parse_qs(parsed.query).get("id") or [None])[0]
             if forum:
-                candidates.add((3, f"artifact:openreview:{str(forum).casefold()}"))
+                candidates.add((3, f"artifact:openreview:{str(forum)}"))
         if host == "github.com" and len(segments) >= 2:
             candidates.add(
                 (4, f"artifact:github:{segments[0].casefold()}/{segments[1].casefold()}")
@@ -103,17 +103,23 @@ def _exact_candidates(item: dict[str, Any]) -> list[tuple[int, str]]:
 
     # No recognizable URL identifier, so fall back to the source's own id.
     source = str(item.get("source") or "").casefold()
-    source_id = str(item.get("source_id") or "").strip().casefold()
+    source_id = str(item.get("source_id") or "").strip()
     if source == "arxiv":
-        base_id = re.sub(r"v\d+$", "", source_id)
+        base_id = re.sub(r"v\d+$", "", source_id.casefold())
         return [(2, f"artifact:arxiv:{base_id}")]
     if source == "openreview":
         return [(3, f"artifact:openreview:{source_id}")]
     if source in {"github", "github release"}:
-        return [(4, f"artifact:github:{source_id.split('@', 1)[0]}")]
+        return [(4, f"artifact:github:{source_id.casefold().split('@', 1)[0]}")]
     if source == "hugging face":
-        return [(5, f"artifact:huggingface:datasets:{source_id}")]
-    return [(9, _stable_id("artifact:url", str(item.get("url") or source_id).casefold()))]
+        return [(5, f"artifact:huggingface:datasets:{source_id.casefold()}")]
+    # Schemes and hostnames are case-insensitive; paths and queries are not.
+    url = str(item.get("url") or source_id)
+    parsed = urlsplit(url)
+    normalized = parsed._replace(
+        scheme=parsed.scheme.lower(), netloc=parsed.netloc.lower()
+    ).geturl()
+    return [(9, _stable_id("artifact:url", normalized))]
 
 
 def exact_artifact_keys(item: dict[str, Any]) -> list[str]:

tests/test_corpus.py

@@ -135,3 +135,44 @@ def test_validation_rejects_edges_to_unknown_entities():
 
     with pytest.raises(CorpusError, match="unknown entity"):
         validate_corpus(corpus)
+
+
+@pytest.mark.parametrize(
+    "first,second",
+    [
+        (
+            item(
+                source="OpenReview",
+                source_id="AbC123",
+                url="https://openreview.net/forum?id=AbC123",
+            ),
+            item(
+                source="OpenReview",
+                source_id="abc123",
+                url="https://openreview.net/forum?id=abc123",
+            ),
+        ),
+        (
+            item(source="OpenReview", source_id="AbC123", url="https://openreview.net/pdf/AbC123"),
+            item(source="OpenReview", source_id="abc123", url="https://openreview.net/pdf/abc123"),
+        ),
+        (
+            item(url="https://example.com/Benchmark?version=A"),
+            item(source_id="paper-2", url="https://example.com/benchmark?version=a"),
+        ),
+    ],
+)
+def test_corpus_keeps_case_sensitive_identifiers_distinct(first, second):
+    corpus = build_corpus([snapshot(first, second)])
+    artifacts = [entity for entity in corpus["entities"] if entity["type"] == "artifact"]
+    assert len(artifacts) == 2
+    assert all(entity["observation_count"] == 1 for entity in artifacts)
+
+
+def test_generic_url_hostname_is_case_insensitive():
+    first = item(url="https://EXAMPLE.com/Benchmark?version=A")
+    second = item(source_id="paper-2", url="https://example.COM/Benchmark?version=A")
+    corpus = build_corpus([snapshot(first, second)])
+    artifacts = [entity for entity in corpus["entities"] if entity["type"] == "artifact"]
+    assert len(artifacts) == 1
+    assert artifacts[0]["observation_count"] == 2