Move DiskCache to diskcache-community and read its index with SafeDisk

iceframe/cache.py

@@ -152,10 +152,10 @@ class DiskCache(QueryCache):
     def __init__(self, cache_dir: str = ".iceframe_cache", max_size_gb: float = 1.0):
         super().__init__(max_size=10**9)  # disk capacity bound is enforced by diskcache
         try:
-            import diskcache  # noqa: F401
+            import diskcache_community  # noqa: F401
         except ImportError:
             raise ImportError(
-                "diskcache required for disk caching. Install with: pip install 'iceframe[cache]'"
+                "diskcache-community required for disk caching. Install with: pip install 'iceframe[cache]'"
             ) from None
         import os
 
@@ -165,11 +165,12 @@ def __init__(self, cache_dir: str = ".iceframe_cache", max_size_gb: float = 1.0)
         # diskcache index in <cache_dir>/index/ holds metadata pointers.
         self._data_dir = os.path.join(self.cache_dir, "data")
         os.makedirs(self._data_dir, exist_ok=True)
-        import diskcache
+        from diskcache_community import Cache, SafeDisk
 
-        self.cache = diskcache.Cache(
+        self.cache = Cache(
             os.path.join(self.cache_dir, "index"),
             size_limit=int(max_size_gb * 1024**3),
+            disk=SafeDisk.with_options(on_error="miss"),
         )
 
     def _payload_path(self, key: str) -> str:

pyproject.toml

@@ -88,6 +88,7 @@ dev = [
     "typer>=0.9.0",
     "rich>=13.0.0",
     "pandas>=2.0.0",
+    "diskcache-community>=5.8.4",
 ]
 pdf = [
     "fpdf2>=2.7.0",
@@ -116,7 +117,7 @@ agent = [
     "rich>=13.0.0",
 ]
 cache = [
-    "diskcache>=5.6.0",
+    "diskcache-community>=5.8.4",
 ]
 streaming = [
     "kafka-python>=2.0.0",

tests/test_disk_cache.py

@@ -0,0 +1,39 @@
+from pathlib import Path
+
+import polars as pl
+import pytest
+
+diskcache_community = pytest.importorskip("diskcache_community")
+
+from iceframe.cache import DiskCache
+
+PARAMS = {"q": 1}
+
+
+def _write_marker(path: str) -> None:
+    Path(path).write_text("ran")
+
+
+class _PlantedEntry:
+    def __init__(self, marker: Path):
+        self.marker = str(marker)
+
+    def __reduce__(self):
+        return (_write_marker, (self.marker,))
+
+
+def test_disk_cache_planted_pickle_is_a_miss(tmp_path):
+    cache_dir = tmp_path / "cache"
+    marker = tmp_path / "planted-code-ran"
+    df = pl.DataFrame({"id": [2, 3], "g": ["a", "b"]})
+    cache = DiskCache(cache_dir=str(cache_dir))
+    cache.put("default.t", PARAMS, df, ttl=300)
+
+    with diskcache_community.Cache(
+        str(cache_dir / "index"), disk=diskcache_community.Disk
+    ) as writer:
+        (key,) = list(writer)
+        writer.set(key, _PlantedEntry(marker))
+
+    assert cache.get("default.t", PARAMS) is None
+    assert not marker.exists()