Fix cache key mismatch (#257)

* Fix cache key mismatch (#256)
* Fix code quality for CI
* Add #256, #257 to `doc/whatsnew.rst`
* Expand unit test for caching behavior
* Add dummy response for new unit test
* Fix missing commas
* Apply `ruff format`

Author: benfrankel

doc/whatsnew.rst

@@ -14,6 +14,7 @@ Next release
   with both :py:`version=...` and :py:`urn=...` would raise :class:`ValueError`
   even if the two were in agreement.
 - Fix two regressions in :func:`.to_pandas` introduced in v2.23.0 (:issue:`251`, :pull:`252`).
+- Fix false cache hits and misses (thanks :gh-user:`benfrankel` for :issue:`256`, :pull:`257`).
 - Fix a bug where supplying `references=...` to ESTAT or EMPL would raise :class:`ValueError` (thanks :gh-user:`benfrankel` for :issue:`259`, :pull:`260`).
 
 v2.23.1 (2025-10-01)

sdmx/client.py

@@ -339,7 +339,7 @@ def get(
         use_cache: bool = False,
         dry_run: bool = False,
         **kwargs,
-    ) -> "sdmx.message.Message":
+    ) -> "sdmx.message.Message | requests.Request":
         """Retrieve SDMX data or metadata.
 
         (Meta)data is retrieved from the :attr:`source` of the current Client. The
@@ -517,8 +517,8 @@ def get(
         msg = self.source.finish_message(msg, self, **kwargs)
 
         # store in memory cache if needed
-        if use_cache:
-            self.cache[req.url] = msg
+        if use_cache and req_prepared.url:
+            self.cache[req_prepared.url] = msg
 
         return msg
 

sdmx/testing/data.py

@@ -263,8 +263,17 @@ def add_responses(session: "Session", file_cache_path: Path, source: "Source") -
         "userdefinedoperatorscheme",
         "vtlmappingscheme",
     ):
-        url = f"{source.url}/{endpoint}/{source.id}/all/latest"
-        save_response(session, method="GET", url=url, content=content, headers=headers)
+        for url in (
+            f"{source.url}/{endpoint}/{source.id}/all/latest",
+            f"{source.url}/{endpoint}/{source.id}/all/latest?references=children",
+        ):
+            save_response(
+                session,
+                method="GET",
+                url=url,
+                content=content,
+                headers=headers,
+            )
 
     for url in (
         f"{source.url}/availableconstraint",

sdmx/tests/test_client.py

@@ -70,8 +70,26 @@ def test_init(self) -> None:
             sdmx.Client("noagency")
 
     # Regular methods
-    def test_clear_cache(self, client: "Client") -> None:
+    def test_cache(self, client: "Client") -> None:
+        # Cache starts empty
+        assert not client.cache
+
+        # Response gets cached by the correct URL on cache miss
+        req = client.get("dataflow", references="children", dry_run=True)
+        assert isinstance(req, PreparedRequest)
+        msg0 = client.get("dataflow", references="children", use_cache=True)
+        assert len(client.cache) == 1
+        assert req.url is not None and client.cache[req.url] is msg0
+
+        # Cached response gets returned on cache hit
+        msg1 = client.get("dataflow", references="children", use_cache=True)
+        assert msg1 is msg0
+        assert len(client.cache) == 1
+        assert req.url is not None and client.cache[req.url] is msg0
+
+        # Clearing the cache works
         client.clear_cache()
+        assert not client.cache
 
     def test_session_attrs0(
         self, caplog: "pytest.LogCaptureFixture", client: "Client"