TestRunner.py

import inspect
import json
import os
import re
from importlib import import_module
from pathlib import Path
from pkgutil import iter_modules
from test.mapper import map_filter, map_transformation
from typing import Iterable

from interfaces.Operation import Operation
from tasks.TaskTypes import TaskType

disable_tests_for = [
    "negate_strengthen",
    "word_noise",
]  # TODO: Don't disable tests


def load(module, cls):
    my_class = getattr(module, cls.__name__)
    return my_class()


def load_test_cases(test_json):
    try:
        with open(test_json) as f:
            d = json.load(f)
            examples = d["test_cases"]
        return examples
    except FileNotFoundError:
        raise Exception(
            f"\n\n\t\tYou should add a test file at this location!\n\t\t{test_json}"
        )


def convert_to_snake_case(camel_case):
    name = re.sub(r"(?<!^)(?=[A-Z])", "_", camel_case).lower()
    return name


class OperationRuns(object):
    def __init__(self, transformation_name, search="transformations"):
        if transformation_name == "light":
            self._load_all_transformation_test_case(heavy=False, search=search)
        elif transformation_name == "heavy":
            self._load_all_transformation_test_case(heavy=True, search=search)
        else:
            self._load_single_transformation_test_case(
                transformation_name, search
            )

    def _load_single_transformation_test_case(
        self, transformation_name, search="transformations"
    ):
        filters = []
        filter_test_cases = []
        package_dir = Path(__file__).resolve()  # --> TestRunner.py
        filters_dir = package_dir.parent.joinpath(search, transformation_name)

        t_py = import_module(f"{search}.{transformation_name}")
        t_js = os.path.join(filters_dir, "test.json")
        filter_instance = None
        prev_class_args = {}
        for test_case in load_test_cases(t_js):
            class_name = test_case["class"]
            class_args = test_case["args"] if "args" in test_case else {}
            # construct filter class with input args
            cls = getattr(t_py, class_name)
            if (
                filter_instance is None
                or filter_instance.name() != class_name
                or prev_class_args != class_args
            ):
                filter_instance = cls(**class_args)
                prev_class_args = class_args

            filters.append(filter_instance)
            filter_test_cases.append(test_case)

        self.operations = filters
        self.operation_test_cases = filter_test_cases

    def _load_multiple_transformation_test_case(
        self,
        transformation_names: list,
        heavy: bool = False,
        search: str = "transformations",
    ):
        """Load multiple classes within a transforamtion.

        Parameters:
        -----------
        heavy: bool, Default is False,
            heavy or light transformation or filter.
        transformation_names: str,
            list of the transformations or filters.
        search: str, Default is transformations,
            either transformations or filters.

        Returns:
        --------
        None.

        """
        filters = []
        filter_test_cases = []

        for m in transformation_names:
            print(f"Directory = {m}")
            if m in disable_tests_for:
                continue

            # Load only the specified transformation
            t_py = import_module(f"{search}.{m}")
            t_js = os.path.join(
                Path(__file__).resolve().parent.joinpath(search),
                m,
                "test.json",
            )
            filter_instance = None
            prev_class_args = {}

            # Load the test.json for the specified transformation
            for test_case in load_test_cases(t_js):
                class_name = test_case["class"]
                class_args = test_case["args"] if "args" in test_case else {}
                # construct filter class with input args
                cls = getattr(t_py, class_name)
                is_heavy = cls.is_heavy()
                if (not heavy) and is_heavy:
                    continue
                else:
                    # Check if the same instance (i.e. with the same args is already loaded)
                    if (
                        filter_instance is None
                        or filter_instance.name() != class_name
                        or prev_class_args != class_args
                    ):
                        filter_instance = cls(**class_args)
                        prev_class_args = class_args

                filters.append(filter_instance)
                filter_test_cases.append(test_case)

        self.operations = filters
        self.operation_test_cases = filter_test_cases

    def _load_all_transformation_test_case(
        self, heavy=False, search="transformations"
    ):
        if search == "transformations":
            # Load either heavy or light transformations only based on heavy param
            self._load_multiple_transformation_test_case(
                map_transformation["heavy"]
                if heavy
                else map_transformation["light"],
                heavy,
                search,
            )
        elif search == "filters":
            # Load either heavy or light filters only based on heavy param
            self._load_multiple_transformation_test_case(
                map_filter["heavy"] if heavy else map_filter["light"],
                heavy,
                search,
            )

    @staticmethod
    def get_all_folder_names(
        search="transformations", transformation_name="all"
    ) -> Iterable:
        """Get all the folder names.

        Parameters:
        ----------
        search: str, default "transformations"
            value can be either transformations or filters.
        transformation_name: str, default "all"
            value can be either all (both light and heavy transformations) or light.

        Returns:
        -------
        list of folder names.
        """
        # iterate through the modules in the current package
        package_dir = Path(__file__).resolve()  # --> TestRunner.py
        transformations_dir = package_dir.parent.joinpath(search)
        if search == "transformations" and transformation_name == "light":
            for entry in map_transformation["light"]:
                yield entry  # only light transformations
        elif search == "transformations" and transformation_name == "heavy":
            for entry in map_transformation["heavy"]:
                yield entry  # only heavy transformations
        elif search == "filters" and transformation_name == "light":
            for entry in map_filter["light"]:
                yield entry  # only light filters
        elif search == "filters" and transformation_name == "heavy":
            for entry in map_filter["heavy"]:
                yield entry  # only heavy filters
        else:
            for (_, folder, _) in iter_modules(
                [transformations_dir]
            ):  # ---> ["back_translation", ...]
                yield folder

    @staticmethod
    def get_all_operations(search="transformations") -> Iterable:
        # iterate through the modules in the current package
        package_dir = Path(__file__).resolve()  # --> TestRunner.py
        transformations_dir = package_dir.parent.joinpath(search)
        for (_, folder, _) in iter_modules(
            [transformations_dir]
        ):  # ---> ["back_translation", ...]
            t_py = import_module(f"{search}.{folder}")
            for name, obj in inspect.getmembers(t_py):
                if (
                    inspect.isclass(obj)
                    and issubclass(obj, Operation)
                    and not obj.__module__.startswith("interfaces")
                ):
                    yield obj

    @staticmethod
    def get_all_operations_for_task(
        query_task_type: TaskType, search="transformations"
    ) -> Iterable:
        # iterate through the modules in the current package
        for operation in OperationRuns.get_all_operations(search):
            if query_task_type in operation.tasks:
                yield operation


def get_implementation(clazz: str, search="transformations"):
    for operation in OperationRuns.get_all_operations(search):
        if operation.name() == clazz:
            return operation
    raise ValueError(
        f"No class called {clazz} found in the {search} folder. Check if you've spelled it right!"
    )


if __name__ == "__main__":
    for x in OperationRuns.get_all_folder_names("transformations", "heavy"):
        print(x)
    for x in OperationRuns.get_all_folder_names("transformations"):
        print(x)
    for x in OperationRuns.get_all_folder_names("filters", "heavy"):
        print(x)
    for x in OperationRuns.get_all_folder_names("filters"):
        print(x)
    for x in OperationRuns.get_all_folder_names():
        print(x)
    for x in OperationRuns.get_all_operations():
        print(x)
    for x in OperationRuns.get_all_operations("filters"):
        print(x)
    print()
    for transformation in OperationRuns.get_all_operations_for_task(
        TaskType.QUESTION_ANSWERING
    ):
        print(transformation.name())
        impl = transformation()
        print(impl.generate("context", "question", ["answer1", "answerN"]))
    for transformation in OperationRuns.get_all_operations_for_task(
        TaskType.TEXT_CLASSIFICATION
    ):
        print(transformation.name())
        impl = transformation()
        for sentence in [
            "Mahendra Dhoni finally travelled to Australia with 5 suitcases. "
            "He wanted to prepare for the biggest game of the season!!!"
        ]:
            for p in impl.generate(sentence):
                print(p)