Merge pull request #576 from lenskit/feature/common-tests

Implement and use common component test suites

Author: mdekstrand

lenskit-funksvd/lenskit/funksvd.py

@@ -238,7 +238,7 @@ class FunkSVDScorer(Component, Trainable):
 
     def __init__(
         self,
-        features: int,
+        features: int = 50,
         iterations: int = 100,
         *,
         lrate: float = 0.001,
@@ -314,6 +314,7 @@ def __call__(self, query: QueryInput, items: ItemList) -> ItemList:
         query = RecQuery.create(query)
 
         user_id = query.user_id
+        user_num = None
         if user_id is not None:
             user_num = self.users_.number(user_id, missing=None)
         if user_num is None:

lenskit-funksvd/tests/test_funksvd.py

@@ -17,7 +17,7 @@
 from lenskit.data.bulk import dict_to_df, iter_item_lists
 from lenskit.funksvd import FunkSVDScorer
 from lenskit.metrics import call_metric, quick_measure_model
-from lenskit.testing import ml_100k, ml_ds, wantjit  # noqa: F401
+from lenskit.testing import BasicComponentTests, ScorerTests, wantjit
 
 _log = logging.getLogger(__name__)
 
@@ -27,6 +27,10 @@
 simple_ds = from_interactions_df(simple_df)
 
 
+class TestFunkSVD(BasicComponentTests, ScorerTests):
+    component = FunkSVDScorer
+
+
 def test_fsvd_basic_build():
     algo = FunkSVDScorer(20, iterations=20)
     algo.train(simple_ds)

lenskit-hpf/lenskit/hpf.py

@@ -39,7 +39,7 @@ class HPFScorer(Component, Trainable):
     items_: Vocabulary
     item_features_: np.ndarray[tuple[int, int], np.dtype[np.float64]]
 
-    def __init__(self, features: int, **kwargs):
+    def __init__(self, features: int = 50, **kwargs):
         self.features = features
         self._kwargs = kwargs
 
@@ -78,6 +78,7 @@ def __call__(self, query: QueryInput, items: ItemList) -> ItemList:
         query = RecQuery.create(query)
 
         user_id = query.user_id
+        user_num = None
         if user_id is not None:
             user_num = self.users_.number(user_id, missing=None)
         if user_num is None:

lenskit-hpf/tests/test_hpf.py

@@ -15,12 +15,17 @@
 from lenskit.data import ItemList, from_interactions_df
 from lenskit.metrics import quick_measure_model
 from lenskit.pipeline import topn_pipeline
+from lenskit.testing import BasicComponentTests, ScorerTests
 
 hpf = importorskip("lenskit.hpf")
 
 _log = logging.getLogger(__name__)
 
 
+class TestHPF(BasicComponentTests, ScorerTests):
+    component = hpf.HPFScorer
+
+
 @mark.slow
 def test_hpf_train_large(tmp_path, ml_ratings):
     algo = hpf.HPFScorer(20)

lenskit-implicit/tests/test_implicit.py

@@ -15,10 +15,19 @@
 from lenskit.data import ItemList, from_interactions_df
 from lenskit.implicit import ALS, BPR
 from lenskit.metrics import quick_measure_model
+from lenskit.testing import BasicComponentTests, ScorerTests
 
 _log = logging.getLogger(__name__)
 
 
+class TestImplicitALS(BasicComponentTests, ScorerTests):
+    component = ALS
+
+
+class TestImplicitBPR(BasicComponentTests, ScorerTests):
+    component = BPR
+
+
 @mark.slow
 def test_implicit_als_train_rec(ml_ds):
     algo = ALS(25)

lenskit-sklearn/lenskit/sklearn/svd.py

@@ -53,7 +53,7 @@ class BiasedSVDScorer(Component, Trainable):
 
     def __init__(
         self,
-        features: int,
+        features: int = 50,
         *,
         damping: UITuple[float] | float | tuple[float, float] = 5,
         algorithm: Literal["arpack", "randomized"] = "randomized",

lenskit-sklearn/tests/test_svd.py

@@ -15,6 +15,7 @@
 from lenskit.data import Dataset, ItemList, from_interactions_df
 from lenskit.metrics import call_metric, quick_measure_model
 from lenskit.sklearn import svd
+from lenskit.testing import BasicComponentTests, ScorerTests
 
 _log = logging.getLogger(__name__)
 
@@ -26,6 +27,10 @@
 need_skl = mark.skipif(not svd.SKL_AVAILABLE, reason="scikit-learn not installed")
 
 
+class TestBiasedSVD(BasicComponentTests, ScorerTests):
+    component = svd.BiasedSVDScorer
+
+
 @need_skl
 def test_svd_basic_build():
     algo = svd.BiasedSVDScorer(2)

lenskit/lenskit/als/_explicit.py

@@ -58,7 +58,7 @@ class BiasedMFScorer(ALSBase):
 
     def __init__(
         self,
-        features: int,
+        features: int = 50,
         *,
         epochs: int = 10,
         reg: float | tuple[float, float] = 0.1,

lenskit/lenskit/als/_implicit.py

@@ -81,7 +81,7 @@ class ImplicitMFScorer(ALSBase):
 
     def __init__(
         self,
-        features: int,
+        features: int = 50,
         *,
         epochs: int = 20,
         reg: float | tuple[float, float] = 0.1,

lenskit/lenskit/data/__init__.py

@@ -13,6 +13,8 @@
 from .convert import from_interactions_df
 from .dataset import Dataset, FieldError
 from .items import ItemList
+from .lazy import LazyDataset
+from .matrix import MatrixDataset
 from .movielens import load_movielens, load_movielens_df
 from .mtarray import MTArray, MTFloatArray, MTGenericArray, MTIntArray
 from .query import QueryInput, RecQuery
@@ -23,6 +25,8 @@
     "Dataset",
     "FieldError",
     "from_interactions_df",
+    "LazyDataset",
+    "MatrixDataset",
     "ID",
     "NPID",
     "UITuple",

lenskit/lenskit/knn/item.py

@@ -79,7 +79,7 @@ class ItemKNNScorer(Component, Trainable):
 
     def __init__(
         self,
-        nnbrs: int,
+        nnbrs: int = 20,
         min_nbrs: int = 1,
         min_sim: float = 1.0e-6,
         save_nbrs: int | None = None,
@@ -202,7 +202,10 @@ def __call__(self, query: QueryInput, items: ItemList) -> ItemList:
         ratings = query.user_items
         if ratings is None:
             if query.user_id is None:
-                raise ValueError("cannot recommend without without either user ID or items")
+                warnings.warn(
+                    "cannot recommend without without either user ID or items", DataWarning
+                )
+                return ItemList(items, scores=np.nan)
 
             upos = self.users_.number(query.user_id, missing=None)
             if upos is None:

lenskit/lenskit/knn/user.py

@@ -79,7 +79,7 @@ class UserKNNScorer(Component, Trainable):
 
     def __init__(
         self,
-        nnbrs: int,
+        nnbrs: int = 20,
         min_nbrs: int = 1,
         min_sim: float = 1.0e-6,
         feedback: FeedbackType = "explicit",
@@ -155,6 +155,9 @@ def __call__(self, query: QueryInput, items: ItemList) -> ItemList:
         query = RecQuery.create(query)
         watch = util.Stopwatch()
         log = _log.bind(user_id=query.user_id, n_items=len(items))
+        if len(items) == 0:
+            log.debug("no candidate items, skipping")
+            return ItemList(items, scores=np.nan)
 
         udata = self._get_user_data(query)
         if udata is None:

lenskit/lenskit/testing/__init__.py

@@ -8,9 +8,9 @@
 import os
 from contextlib import contextmanager
 
-import pytest
-
 from ._arrays import coo_arrays, scored_lists, sparse_arrays, sparse_tensors
+from ._components import BasicComponentTests, ScorerTests
+from ._markers import jit_enabled, wantjit
 from ._movielens import (
     demo_recs,
     ml_100k,
@@ -36,16 +36,10 @@
     "wantjit",
     "jit_enabled",
     "set_env_var",
+    "BasicComponentTests",
+    "ScorerTests",
 ]
 
-jit_enabled = True
-if "NUMBA_DISABLE_JIT" in os.environ:
-    jit_enabled = False
-if os.environ.get("PYTORCH_JIT", None) == "0":
-    jit_enabled = False
-
-wantjit = pytest.mark.skipif(not jit_enabled, reason="JIT required")
-
 
 @contextmanager
 def set_env_var(var, val):