Update license, plugin interface, dataset interface (#28)

* Update all configurations

* Set up data format and validation (#2)

* Implement library config

* Isort docs/conf.py

* Improve configuration

* Add logging

* Add .print to logger

* Set up way of updating things on config changes

* Simplify and use get_config()

* Tidy multi-line logger messages

* Add logger.raise_ convenience method

* Move logging into a package

* Add diagnose, backtrace as configurable options

* Make default config be defined in yaml file only

* Add validator design

* Add extra config test

* Define global data settings

* Clean up validator

* Use builtin types in DataSettings

* Minor refactor validator

* Set up DataRequirements

* Set up validation implementation

* Decouple tests, tidy imports

* Imports to Google style where possible

* Implement some data validators (root_validate)

* Use decorator in val. impl. method registry

* Improve exception logging

* Update some validation_implementation methods

* Reorganize dir structure

* Clean up tests

* Allow for inheritance in val. impl.

* Add validator > df tests

* Separate out SupportsContainer interface

* Shorten package names

* Factor out SupportsImplementations

* Factor out RegisterMethodDecorator

* Add dispatch_to_implementation()

* Separate out interface

* Set up framework for *Samples objects

* Add data utils

* Implement as_array for TimeSeriesSamples

* Introduce data container def.

* Update data container defs, separate out tests

* Set up default container flavor mechanism

* Add check_untyped_defs = True in mypy.ini

* Implement EventSamples

* Implement as_array for EventSamples

* Update setup requirements

* Set up docs (#4)

* Remove unneeded BAK file

* Small additions to README

* Update image display

* Major update README

* Table width

* Change image align

* Prepare README for docs

* Add logo

* Update docs

* Fix issue in README

* Dr shushen/model setup (#23)

* Remove unneeded BAK file

* Change abc import to be Google code format style

* Introduce TemporBaseModel

* Introduce fit()

* Bugfix, add typing overloads for fit()

* Add core requirements

* Introduce RequirementCategory

* Set up requirement validator concept

* Reorganize data/

* Introduce DataBundle

* Add DataBundle requirements

* Set up from_data_containers static method

* Update install_requires

* Introduce RequirementsConfig

* Reorganize packages to improve naming

* Develop RequirementsConfig further

* Improve RequirementsConfig repr

* Deal with _validate_method_config

* Minor bugfixes

* Add some int. tests for base model fit config

* Add _fit_called flag

* Introduce transform method

* Tidy imports

* Update LICENSE (#24)

* [Feat] Basic plugin interface and loader (#27)

* Rename model dir to plugins

* Set up plugins/core dir

* Simplify estimator, transformer

* Add predictor

* Fix circular import

* Add core plugin methods like name()

* Implement a Plugin interface

* Factorize out test utility "patch_module"

* Add test for plugin infrastructure

* Add test for loaded plugins

* add fit_{predict,transform} methods

* Add hyperparameter methods

* Add Base* to indicate base models

* Keep only estimator init with params as kwargs

* Rename test file

* Add parent constructor calls for clarity

* Set up Dataset (#36)

* Remove old data format

* Initial interface for Dataset

* Remove unnecessary decorator

* Add StaticSamples validation and tests

* Add TimeSeriesSamples validation

* Add EventSamples validation

* Add tests for samples basics

* Test EventSamples.split method

* Add  time/sample_index helper methods

* Add @validate_arguments and some note comments

* Add repr's

* Implement from_numpy for {Static,Event}Samples

* Implement .numpy for {Static,Event}Samples

* Add utils for array/df manipulation

* Implement TimeSeriesSamples .numpy()

* Fix some typing definitions

* Write utils for array -> TS df conversion

* Check in register_plugin if re-imported (no exc.)

* Add docstrings in utils

* Add pydantic.validate_arguments in data utils

* Implement TimeSeriesSamples from_numpy()

* Update docstring in TimeSeriesSamples.__init__

* Add unit tests for Dataset

* Set default debug level to INFO

* Update tests

* Add data format tutorial notebook

* Add reprs

* Simplify _check_same_class

* Add docstring to Dataset classes

setup.cfg

@@ -8,7 +8,7 @@ name = temporai
 description = TemporAI: ML-centric Toolkit for Medical Time Series
 author = Evgeny Saveliev
 author_email = [email protected]
-license = GPL-3.0-or-later
+license = Apache-2.0 license
 license_files = LICENSE.txt
 long_description = file: README.md
 long_description_content_type = text/markdown; charset=UTF-8; variant=GFM
@@ -55,6 +55,7 @@ install_requires =
     pandas >=1
     pandera >=0.13
     pydantic >=1, <2
+    rich
 
 
 [options.packages.find]
@@ -74,9 +75,11 @@ tempor =
 
 # Add here test requirements (semicolon/line-separated)
 testing =
-    setuptools
+    jupyter
+    notebook
     pytest
     pytest-cov
+    setuptools
 
 [options.entry_points]
 # Add here console scripts like:
@@ -118,6 +121,7 @@ exclude =
 max-line-length = 120
 disable = R, C
 generated-members = torch.*
+extension-pkg-whitelist = pydantic
 
 [pylint.messages_control]
 disable = C0330, C0326, fixme, c-extension-no-member

src/tempor/config/conf/tempor/config.yaml

@@ -1,6 +1,6 @@
 logging:
   mode: LIBRARY  # {LIBRARY, SCRIPT}
-  level: DEBUG
+  level: INFO
   diagnose: false
   backtrace: true
   file_log: true

src/tempor/core/__init__.py

@@ -1,4 +1,4 @@
-from . import _supports_impl as supports_impl
+from . import supports_impl
 from ._register_method_decorator import RegisterMethodDecorator
 
 __all__ = [

src/tempor/core/_supports_impl.py

@@ -1,11 +1,11 @@
-from abc import ABC, abstractmethod
+import abc
 from typing import Dict, Generic, Tuple, TypeVar
 
 SupportsT = TypeVar("SupportsT")
 ImplementationT = TypeVar("ImplementationT")
 
 
-class SupportsImplementations(Generic[SupportsT, ImplementationT], ABC):
+class SupportsImplementations(Generic[SupportsT, ImplementationT], abc.ABC):
     def __init__(self) -> None:
         super().__init__()
         self._implementations: Dict[SupportsT, ImplementationT] = self._register_implementations()
@@ -26,11 +26,11 @@ def dispatch_to_implementation(self, key: SupportsT) -> ImplementationT:
         return self._implementations[key]
 
     @property
-    @abstractmethod
+    @abc.abstractmethod
     def supports_implementations_for(self) -> Tuple[SupportsT, ...]:  # pragma: no cover
         ...
 
-    @abstractmethod
+    @abc.abstractmethod
     def _register_implementations(self) -> Dict[SupportsT, ImplementationT]:  # pragma: no cover
         ...
 

src/tempor/core/pydantic_utils.py

@@ -0,0 +1,16 @@
+from typing import Dict, Optional
+
+
+def exclusive_args(
+    values: Dict,
+    arg1: str,
+    arg2: str,
+    arg1_friendly_name: Optional[str] = None,
+    arg2_friendly_name: Optional[str] = None,
+) -> None:
+    arg1_value = values.get(arg1, None)
+    arg2_value = values.get(arg2, None)
+    arg1_name = arg1_friendly_name if arg1_friendly_name else f"`{arg1}`"
+    arg2_name = arg2_friendly_name if arg2_friendly_name else f"`{arg2}`"
+    if arg1_value is not None and arg2_value is not None:
+        raise ValueError(f"Must provide either {arg1_name} or {arg2_name} but not both")

src/tempor/core/requirements.py

@@ -0,0 +1,59 @@
+import abc
+import enum
+from typing import Any, ClassVar, Sequence, get_type_hints
+
+import pydantic
+import rich.pretty
+
+
+class RequirementCategory(enum.Enum):
+    DATA_CONTAINER = enum.auto()
+    DATA_BUNDLE = enum.auto()
+    MODEL = enum.auto()
+
+
+class Requirement(abc.ABC):
+    definition: Any
+
+    name: ClassVar[str] = None  # type: ignore
+    validator_methods: ClassVar[Sequence[Any]] = tuple()
+
+    def __init__(self, definition) -> None:
+        validator = pydantic.create_model(
+            f"{self.__class__.__name__}Validator",
+            __validators__={clsmtd.__func__.__name__: clsmtd for clsmtd in self.validator_methods},  # type: ignore
+            name=(str, ...),
+            definition=(get_type_hints(self)["definition"], ...),
+        )
+        validator(name=self.name, definition=definition)
+        self.definition = definition
+
+    @property
+    @abc.abstractmethod
+    def requirement_category(self) -> RequirementCategory:  # pragma: no cover
+        ...
+
+    def __rich_repr__(self):
+        yield "definition", self.definition
+
+    def __repr__(self) -> str:
+        return rich.pretty.pretty_repr(self)
+
+
+class RequirementValidator(abc.ABC):
+    @property
+    @abc.abstractmethod
+    def supported_requirement_category(self) -> RequirementCategory:  # pragma: no cover
+        ...
+
+    def validate(self, target: Any, *, requirements: Sequence[Requirement], **kwargs):  # pragma: no cover
+        for r in requirements:
+            if r.requirement_category != self.supported_requirement_category:
+                raise ValueError(
+                    f"{self.__class__.__name__} does not support requirement category {r.requirement_category}"
+                )
+        self._validate(target=target, requirements=requirements, **kwargs)
+
+    @abc.abstractmethod
+    def _validate(self, *args, **kwargs):  # pragma: no cover
+        ...

src/tempor/core/supports_impl.py

@@ -0,0 +1 @@
+from ._supports_impl import *  # noqa  # pylint: disable=W

src/tempor/core/utils.py

@@ -0,0 +1,22 @@
+import enum
+
+
+def get_class_full_name(o: object):
+    # See: https://stackoverflow.com/a/2020083
+    class_ = o.__class__
+    module = class_.__module__
+    if module == "builtins":
+        return class_.__qualname__  # avoid outputs like "builtins.str"
+    return module + "." + class_.__qualname__
+
+
+def get_enum_name(enum_: enum.Enum) -> str:
+    return enum_.name.lower()
+
+
+class RichReprStrPassthrough:
+    def __init__(self, string: str) -> None:
+        self.string = string
+
+    def __repr__(self) -> str:
+        return self.string

src/tempor/data/__init__.py

@@ -1,34 +0,0 @@
-from ._check_data_container_def import CheckDataContainerDefinition
-from ._settings import (
-    CONTAINER_CLASS_TO_CONTAINER_FLAVORS,
-    DATA_CATEGORY_TO_CONTAINER_FLAVORS,
-    DATA_SETTINGS,
-    DEFAULT_CONTAINER_FLAVOR,
-    DataSettings,
-)
-from ._supports_container import (
-    DataContainerDef,
-    EventSupportsContainer,
-    StaticSupportsContainer,
-    SupportsContainer,
-    TimeSeriesSupportsContainer,
-)
-from ._types import ContainerFlavor, DataCategory, DataContainer, Dtype
-
-__all__ = [
-    "CheckDataContainerDefinition",
-    "CONTAINER_CLASS_TO_CONTAINER_FLAVORS",
-    "ContainerFlavor",
-    "DEFAULT_CONTAINER_FLAVOR",
-    "DATA_CATEGORY_TO_CONTAINER_FLAVORS",
-    "DATA_SETTINGS",
-    "DataCategory",
-    "DataContainer",
-    "DataContainerDef",
-    "DataSettings",
-    "Dtype",
-    "EventSupportsContainer",
-    "StaticSupportsContainer",
-    "SupportsContainer",
-    "TimeSeriesSupportsContainer",
-]