Move to src/irspack layout & always use relative import (#104)

* move to src/ layout

* modify setup.py

* Fix test

* revert wheels.yml

Author: tohtsky

.github/workflows/run-test.yml

@@ -22,7 +22,7 @@ jobs:
         run: |
           pip install pytest pytest-mock pytest-cov
           pip install lightfm jaxlib jax dm-haiku optax
-          pytest --cov=./irspack tests/
+          pytest --cov=./src/irspack tests/
       - name: Generate coverage (ubuntu)
         run: |
           coverage xml

.github/workflows/wheels.yml

@@ -7,7 +7,7 @@ on:
     types:
       - created
 env:
-  cibuildwheel_version: "2.2.2"
+  cibuildwheel_version: "2.7.0"
 jobs:
   build_sdist:
     name: Build source distribution

.gitignore

@@ -8,7 +8,7 @@ temp/*
 **.log
 **.csv
 **.db
-irspack.egg-info/*
+src/irspack.egg-info/*
 **.so
 .eggs/
 .python-version

setup.py

@@ -79,13 +79,14 @@ def local_scheme(version: Any) -> str:
     },  # https://github.com/pypa/setuptools_scm/issues/342
     author="Tomoki Ohtsuki",
     author_email="tomoki.otsuki129@gmail.com",
-    description="Implicit feedback-based recommender system pack",
+    description="Implicit feedback-based recommender systems, packed for practitioners.",
     long_description=LONG_DESCRIPTION,
     long_description_content_type="text/markdown",
     ext_modules=ext_modules,
     install_requires=install_requires,
     include_package_data=True,
     cmdclass={"build_ext": build_ext},
-    packages=find_packages(),
-    python_requires=">=3.6",
+    packages=find_packages("src"),
+    python_requires=">=3.7",
+    package_dir={"": "src"},
 )

irspack/__init__.py renamed to src/irspack/__init__.py

@@ -6,8 +6,8 @@
     # package is not installed
     pass  # pragma: no cover
 
-from irspack.definitions import DenseScoreArray, InteractionMatrix, UserIndexArray
-from irspack.evaluation import *
-from irspack.recommenders import *
-from irspack.split import *
-from irspack.utils import *
+from .definitions import DenseScoreArray, InteractionMatrix, UserIndexArray
+from .evaluation import *
+from .recommenders import *
+from .split import *
+from .utils import *

irspack/utils/threading.py renamed to src/irspack/_threading.py

@@ -3,7 +3,7 @@
 
 import pandas as pd
 
-from irspack.dataset.downloader import SingleZipDownloader
+from ..downloader import SingleZipDownloader
 
 
 class BaseMovieLenstDataLoader(SingleZipDownloader):

irspack/dataset/__init__.py renamed to src/irspack/dataset/__init__.py

@@ -1,5 +1,11 @@
 from ._core import EvaluatorCore, Metrics
-from .evaluator import METRIC_NAMES, Evaluator, EvaluatorWithColdUser, TargetMetric
+from .evaluator import (
+    METRIC_NAMES,
+    Evaluator,
+    EvaluatorWithColdUser,
+    Metrics,
+    TargetMetric,
+)
 
 __all__ = [
     "Evaluator",

irspack/dataset/amazon_music.py renamed to src/irspack/dataset/amazon_music.py

@@ -2,7 +2,7 @@
 
 import pandas as pd
 
-from ..utils.threading import get_n_threads
+from .._threading import get_n_threads
 from ._core import evaluate_list_vs_list
 
 

irspack/dataset/citeulike.py renamed to src/irspack/dataset/citeulike.py

@@ -6,12 +6,12 @@
 import numpy as np
 from scipy import sparse as sps
 
-from irspack.definitions import DenseScoreArray, InteractionMatrix
-from irspack.evaluation._core import EvaluatorCore, Metrics
-from irspack.utils.threading import get_n_threads
+from .._threading import get_n_threads
+from ..definitions import DenseScoreArray, InteractionMatrix
+from ._core import EvaluatorCore, Metrics
 
 if TYPE_CHECKING:
-    from irspack.recommenders import base as base_recommender
+    from ..recommenders.base import BaseRecommender
 
 
 class TargetMetric(Enum):
@@ -172,7 +172,7 @@ def _get_metrics(
         else:
             raise ValueError("score must be either float32 or float64.")
 
-    def get_target_score(self, model: "base_recommender.BaseRecommender") -> float:
+    def get_target_score(self, model: "BaseRecommender") -> float:
         r"""Compute the optimization target score (self.target_metric) with the cutoff being ``self.cutoff``.
 
         Args:
@@ -183,7 +183,7 @@ def get_target_score(self, model: "base_recommender.BaseRecommender") -> float:
         """
         return self.get_score(model)[self.target_metric.name]
 
-    def get_score(self, model: "base_recommender.BaseRecommender") -> Dict[str, float]:
+    def get_score(self, model: "BaseRecommender") -> Dict[str, float]:
         r"""Compute the score with the cutoff being ``self.cutoff``.
 
         Args:
@@ -195,7 +195,7 @@ def get_score(self, model: "base_recommender.BaseRecommender") -> Dict[str, floa
         return self._get_scores_as_list(model, [self.cutoff])[0]
 
     def get_scores(
-        self, model: "base_recommender.BaseRecommender", cutoffs: List[int]
+        self, model: "BaseRecommender", cutoffs: List[int]
     ) -> Dict[str, float]:
         r"""Compute the score with the specified cutoffs.
 
@@ -217,7 +217,7 @@ def get_scores(
         return result
 
     def _get_scores_as_list(
-        self, model: "base_recommender.BaseRecommender", cutoffs: List[int]
+        self, model: "BaseRecommender", cutoffs: List[int]
     ) -> List[Dict[str, float]]:
         if self.offset + self.n_users > model.n_users:
             raise ValueError("evaluator offset + n_users exceeds the model's n_users.")
@@ -343,7 +343,7 @@ def __init__(
 
     def _get_scores_as_list(
         self,
-        model: "base_recommender.BaseRecommender",
+        model: "BaseRecommender",
         cutoffs: List[int],
     ) -> List[Dict[str, float]]:
 

irspack/dataset/downloader.py renamed to src/irspack/dataset/downloader.py

@@ -5,15 +5,14 @@
 
 import pandas as pd
 
-from irspack.evaluation import Evaluator
-from irspack.utils.default_logger import get_default_logger
-
+from ..default_logger import get_default_logger
 from .parameter_range import is_valid_param_name
 
 if TYPE_CHECKING:
     from optuna import Study, Trial
 
-    from irspack.recommenders.base import BaseRecommender, InteractionMatrix
+    from ..evaluation import Evaluator
+    from ..recommenders.base import BaseRecommender, InteractionMatrix
 
 SparseMatrixSuggestFunction = Callable[["Trial"], "InteractionMatrix"]
 ParameterSuggestFunction = Callable[["Trial"], Dict[str, Any]]
@@ -60,7 +59,7 @@ class Optimizer:
     Args:
         data_suggest_function:
             The train data.
-        val_evaluator (Evaluator):
+        val_evaluator:
             The validation evaluator that measures the performance of the recommenders.
         logger:
             The logger used during the optimization steps. Defaults to `None`.
@@ -85,14 +84,15 @@ def __init__(
         data_suggest_function: SparseMatrixSuggestFunction,
         parameter_suggest_function: ParameterSuggestFunction,
         fixed_params: Dict[str, Any],
-        val_evaluator: Evaluator,
+        val_evaluator: "Evaluator",
         logger: Optional[logging.Logger] = None,
         max_epoch: int = 128,
         validate_epoch: int = 5,
         score_degradation_max: int = 5,
     ):
 
         if logger is None:
+
             logger = get_default_logger()
 
         self.logger = logger

irspack/dataset/movielens/ML100K.py renamed to src/irspack/dataset/movielens/ML100K.py

@@ -1,13 +1,9 @@
-from irspack.recommenders.base import (
-    BaseRecommender,
-    BaseSimilarityRecommender,
-    get_recommender_class,
-)
-from irspack.recommenders.base_earlystop import BaseRecommenderWithEarlyStopping
-from irspack.recommenders.dense_slim import DenseSLIMConfig, DenseSLIMRecommender
-from irspack.recommenders.edlae import EDLAEConfig, EDLAERecommender
-from irspack.recommenders.ials import IALSConfig, IALSRecommender
-from irspack.recommenders.knn import (
+from .base import BaseRecommender, BaseSimilarityRecommender, get_recommender_class
+from .base_earlystop import BaseRecommenderWithEarlyStopping
+from .dense_slim import DenseSLIMConfig, DenseSLIMRecommender
+from .edlae import EDLAEConfig, EDLAERecommender
+from .ials import IALSConfig, IALSRecommender
+from .knn import (
     AsymmetricCosineKNNConfig,
     AsymmetricCosineKNNRecommender,
     CosineKNNConfig,
@@ -17,7 +13,6 @@
     TverskyIndexKNNConfig,
     TverskyIndexKNNRecommender,
 )
-
 from .p3 import P3alphaConfig, P3alphaRecommender
 from .rp3 import RP3betaConfig, RP3betaRecommender
 from .slim import SLIMConfig, SLIMRecommender
@@ -63,17 +58,14 @@
 ]
 
 try:
-    from irspack.recommenders.truncsvd import (
-        TruncatedSVDConfig,
-        TruncatedSVDRecommender,
-    )
+    from .truncsvd import TruncatedSVDConfig, TruncatedSVDRecommender
 
     __all__.extend(["TruncatedSVDRecommender", "TruncatedSVDConfig"])
 except ImportError:  # pragma: no cover
     pass  # pragma: no cover
 
 try:
-    from irspack.recommenders.nmf import NMFConfig, NMFRecommender
+    from .nmf import NMFConfig, NMFRecommender
 
     __all__.extend(["NMFRecommender", "NMFConfig"])
 except ImportError:  # pragma: no cover

irspack/dataset/movielens/ML1M.py renamed to src/irspack/dataset/movielens/ML1M.py

@@ -33,7 +33,7 @@
     InteractionMatrix,
     UserIndexArray,
 )
-from .optimization.parameter_range import ParameterRange
+from ..optimization.parameter_range import ParameterRange
 
 R = TypeVar("R", bound="BaseRecommender")
 
@@ -268,7 +268,7 @@ def tune_with_study(
         score_degradation_max: int = 5,
         logger: Optional[logging.Logger] = None,
     ) -> Tuple[Dict[str, Any], pd.DataFrame]:
-        from .optimization.optimizer import Optimizer
+        from ..optimization.optimizer import Optimizer
 
         if data is None:
             if data_suggest_function is None:

irspack/dataset/movielens/ML20M.py renamed to src/irspack/dataset/movielens/ML20M.py

@@ -4,8 +4,8 @@
 
 from fastprogress import progress_bar
 
-from irspack.definitions import InteractionMatrix
-from irspack.recommenders.base import BaseRecommender, RecommenderConfig
+from ..definitions import InteractionMatrix
+from .base import BaseRecommender, RecommenderConfig
 
 if TYPE_CHECKING:
     from optuna import Trial

irspack/dataset/movielens/__init__.py renamed to src/irspack/dataset/movielens/__init__.py

@@ -4,21 +4,20 @@
 import numpy as np
 from lightfm import LightFM
 
-from irspack.utils import get_n_threads
-
+from .._threading import get_n_threads
 from ..definitions import (
     DenseMatrix,
     DenseScoreArray,
     InteractionMatrix,
     UserIndexArray,
 )
+from ..optimization import CategoricalRange, LogUniformFloatRange, UniformIntegerRange
 from .base import BaseRecommenderWithItemEmbedding, BaseRecommenderWithUserEmbedding
 from .base_earlystop import (
     BaseEarlyStoppingRecommenderConfig,
     BaseRecommenderWithEarlyStopping,
     TrainerBase,
 )
-from .optimization import CategoricalRange, LogUniformFloatRange, UniformIntegerRange
 
 
 class BPRFMTrainer(TrainerBase):

irspack/dataset/movielens/base.py renamed to src/irspack/dataset/movielens/base.py

@@ -4,8 +4,8 @@
 from scipy import linalg
 
 from ..definitions import InteractionMatrix
+from ..optimization.parameter_range import LogUniformFloatRange
 from .base import BaseSimilarityRecommender, RecommenderConfig
-from .optimization.parameter_range import LogUniformFloatRange
 
 
 class DenseSLIMConfig(RecommenderConfig):

irspack/dataset/neu_mf.py renamed to src/irspack/dataset/neu_mf.py

@@ -4,8 +4,8 @@
 from scipy import linalg
 
 from ..definitions import InteractionMatrix
+from ..optimization.parameter_range import LogUniformFloatRange, UniformFloatRange
 from .base import BaseSimilarityRecommender, RecommenderConfig
-from .optimization.parameter_range import LogUniformFloatRange, UniformFloatRange
 
 
 class EDLAEConfig(RecommenderConfig):

irspack/utils/default_logger.py renamed to src/irspack/default_logger.py

@@ -10,14 +10,19 @@
 from typing_extensions import Literal  # pragma: no cover, type: ignore
 
 from .. import evaluation
+from .._threading import get_n_threads
 from ..definitions import (
     DenseMatrix,
     DenseScoreArray,
     InteractionMatrix,
     UserIndexArray,
 )
 from ..evaluation.evaluator import Evaluator
-from ..utils import get_n_threads
+from ..optimization.parameter_range import (
+    LogUniformFloatRange,
+    ParameterRange,
+    UniformIntegerRange,
+)
 from ._ials import IALSModelConfigBuilder, IALSSolverConfigBuilder
 from ._ials import IALSTrainer as CoreTrainer
 from ._ials import LossType, SolverType
@@ -31,11 +36,6 @@
     BaseRecommenderWithEarlyStopping,
     TrainerBase,
 )
-from .optimization.parameter_range import (
-    LogUniformFloatRange,
-    ParameterRange,
-    UniformIntegerRange,
-)
 
 if TYPE_CHECKING:
     from optuna import Study, Trial

irspack/definitions.py renamed to src/irspack/definitions.py

@@ -2,21 +2,22 @@
 from abc import abstractmethod
 from typing import Optional, Union
 
+from .._threading import get_n_threads
 from ..definitions import InteractionMatrix
-from ..utils import get_n_threads, okapi_BM_25_weight, remove_diagonal, tf_idf_weight
+from ..optimization.parameter_range import (
+    CategoricalRange,
+    UniformFloatRange,
+    default_tune_range_knn,
+    default_tune_range_knn_with_weighting,
+)
+from ..utils import okapi_BM_25_weight, remove_diagonal, tf_idf_weight
 from ._knn import (
     AsymmetricSimilarityComputer,
     CosineSimilarityComputer,
     JaccardSimilarityComputer,
     TverskyIndexComputer,
 )
 from .base import BaseSimilarityRecommender, RecommenderConfig
-from .optimization.parameter_range import (
-    CategoricalRange,
-    UniformFloatRange,
-    default_tune_range_knn,
-    default_tune_range_knn_with_weighting,
-)
 
 
 class FeatureWeightingScheme(str, enum.Enum):

irspack/evaluation/__init__.py renamed to src/irspack/evaluation/__init__.py

@@ -11,12 +11,12 @@
 from scipy import sparse as sps
 
 from ..definitions import DenseScoreArray, InteractionMatrix, UserIndexArray
+from ..optimization.parameter_range import CategoricalRange
 from .base_earlystop import (
     BaseEarlyStoppingRecommenderConfig,
     BaseRecommenderWithEarlyStopping,
     TrainerBase,
 )
-from .optimization.parameter_range import CategoricalRange
 
 
 class BaseMLP: