refactoring: splitting conftest logic into submodules

test/conftest.py

@@ -3,33 +3,20 @@ import logging
 import os
 import random
 import string
-import tempfile
 from functools import partial
 from itertools import starmap
 from operator import itemgetter
-from typing import Iterable
 
 import fpdf
 import numpy as np
 import pytest
-from PIL import Image
-from frozendict import frozendict
 from funcy import rcompose, merge
 
-from image_prediction.classifier.classifier import Classifier
-from image_prediction.classifier.image_classifier import ImageClassifier
-from image_prediction.estimator.adapter.adapter import EstimatorAdapter
-from image_prediction.estimator.preprocessor.preprocessors.basic import BasicPreprocessor
-from image_prediction.estimator.preprocessor.utils import image_to_normalized_tensor
 from image_prediction.exceptions import (
-    UnknownEstimatorAdapter,
-    UnknownImageExtractor,
     UnknownDatabaseType,
     UnknownLabelFormat,
 )
 from image_prediction.image_extractor.extractor import ImageMetadataPair
-from image_prediction.image_extractor.extractors.mock import ImageExtractorMock
-from image_prediction.image_extractor.extractors.parsable import ParsablePDFImageExtractor
 from image_prediction.info import Info
 from image_prediction.label_mapper.mappers.numeric import IndexMapper
 from image_prediction.label_mapper.mappers.probability import ProbabilityMapper, ProbabilityMapperKeys
@@ -39,8 +26,12 @@ from image_prediction.model_loader.loader import ModelLoader
 from image_prediction.model_loader.loaders.mlflow import MlflowConnector
 from image_prediction.pipeline import load_pipeline
 from image_prediction.redai_adapter.mlflow import MlflowModelReader
-from image_prediction.redai_adapter.model import PredictionModelHandle
 from image_prediction.utils import get_logger
+from test.utils.generation.image import array_to_image
+from test.utils.generation.pdf import add_image, pdf_stream
+
+pytest_plugins = ['test.utils.model']
+
 
 
 @pytest.fixture(autouse=True)
@@ -52,46 +43,6 @@ def mute_logger():
     logger.setLevel(level)
 
 
-@pytest.fixture
-def image_extractor(extractor_type):
-    if extractor_type == "mock":
-        return ImageExtractorMock()
-    elif extractor_type == "parsable_pdf":
-        return ParsablePDFImageExtractor()
-    elif extractor_type == "default":
-        return None
-    else:
-        raise UnknownImageExtractor(f"No image extractor for type {extractor_type} was specified.")
-
-
-@pytest.fixture
-def image_classifier(classifier, monkeypatch, batch_of_expected_string_labels):
-    return ImageClassifier(classifier, preprocessor=BasicPreprocessor())
-
-
-@pytest.fixture
-def classifier(estimator_adapter, label_mapper):
-    classifier = Classifier(estimator_adapter, label_mapper)
-    return classifier
-
-
-@pytest.fixture
-def estimator_mock():
-    class EstimatorMock:
-        @staticmethod
-        def predict(batch):
-            return [None for _ in batch]
-
-        @staticmethod
-        def predict_proba(batch):
-            return [None for _ in batch]
-
-        def __call__(self, batch):
-            return self.predict(batch)
-
-    return EstimatorMock()
-
-
 @pytest.fixture
 def label_mapper(label_format, classes):
     if label_format == "index":
@@ -129,51 +80,6 @@ def expected_predictions(label_format, batch_of_expected_numeric_labels, batch_o
         raise UnknownLabelFormat(f"No label mapper for label format {label_format} was specified.")
 
 
-@pytest.fixture
-def estimator_adapter(
-    estimator_type, estimator_mock, keras_model, model_handle_mock, output_batch_generator, monkeypatch
-):
-    if estimator_type == "mock":
-        estimator_adapter = EstimatorAdapter(estimator_mock)
-    elif estimator_type == "keras":
-        estimator_adapter = EstimatorAdapter(keras_model)
-    elif estimator_type == "redai":
-        estimator_adapter = EstimatorAdapter(PredictionModelHandle(model_handle_mock))
-    else:
-        raise UnknownEstimatorAdapter(f"No adapter for estimator type {estimator_type} was specified.")
-
-    def mock_predict(batch):
-        # Run real predict function to test for mechanical issues, but return externally defined
-        # predictions to test the callers of the estimator adapter against the expected predictions
-        return [next(output_batch_generator) for _ in _predict(batch)]
-
-    _predict = estimator_adapter.predict
-    monkeypatch.setattr(estimator_adapter, "predict", mock_predict)
-
-    return estimator_adapter
-
-
-@pytest.fixture
-def keras_model(input_size):
-    import os
-
-    os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
-
-    import tensorflow as tf
-
-    tf.keras.backend.set_image_data_format("channels_last")
-
-    inputs = tf.keras.Input(shape=input_size)
-    conv = tf.keras.layers.Conv2D(3, 3)
-    dense = tf.keras.layers.Dense(10)
-
-    outputs = tf.keras.layers.Dense(10)(dense(conv(inputs)))
-    model = tf.keras.Model(inputs=inputs, outputs=outputs)
-    model.compile()
-
-    return model
-
-
 @pytest.fixture
 def images(input_batch):
     return list(map(array_to_image, input_batch))
@@ -205,21 +111,6 @@ def __input_size(request):
     return itemgetter("width", "height", "depth")(request.param)
 
 
-def array_to_image(array):
-    assert np.all(array <= 1)
-    assert np.all(array >= 0)
-
-    if array.shape[-1] == 3:
-        mode = "RGB"
-    elif array.shape[-1] == 4:
-        mode = "RGBA"
-    else:
-        raise ValueError(f"Unexpected number of channels {array.shape[-1]}. Expected 3 or 4.")
-
-    # noinspection PyTypeChecker
-    return Image.fromarray(np.uint8(array * 255), mode=mode)
-
-
 @pytest.fixture
 def batch_of_expected_string_labels(batch_of_expected_numeric_labels, classes):
     return map_labels(batch_of_expected_numeric_labels, classes)
@@ -340,44 +231,6 @@ def pdf(image_metadata_pairs):
     return pdf_stream(pdf)
 
 
-def add_image(pdf, image_metadata_pair, suffix="png"):
-    while fewer_pages_then_required(image_metadata_pair.metadata[Info.PAGE_IDX], pdf):
-        pdf.add_page()
-
-    add_image_to_last_page(pdf, image_metadata_pair, suffix=suffix)
-
-
-def fewer_pages_then_required(page_idx, pdf):
-    return page_idx > pdf.page - 1
-
-
-def pdf_stream(pdf: fpdf.fpdf.FPDF):
-    return pdf.output(dest="S").encode("latin1")
-
-
-def add_image_to_last_page(pdf: fpdf.fpdf.FPDF, image_metadata_pair, suffix):
-    image, metadata = image_metadata_pair
-    x, y, w, h = itemgetter(Info.X1, Info.Y1, Info.WIDTH, Info.HEIGHT)(metadata)
-
-    with tempfile.NamedTemporaryFile(suffix=f".{suffix}") as temp_image:
-        image.save(temp_image.name)
-        pdf.image(temp_image.name, x=x, y=y, w=w, h=h, type=suffix)
-
-
-@pytest.fixture
-def model():
-    class Model:
-        @staticmethod
-        def predict(*args):
-            return True
-
-        @staticmethod
-        def predict_proba(*args):
-            return True
-
-    return Model()
-
-
 @pytest.fixture
 def model_database_record_identifier():
     return "".join(random.sample(string.ascii_letters, k=10))
@@ -429,24 +282,6 @@ def mlflow_reader(mlruns_dir):
     return MlflowModelReader(mlruns_dir)
 
 
-@pytest.fixture
-def model_handle_mock(estimator_mock):
-    class ModelHandleMock:
-        def __init__(self):
-            self.model = estimator_mock
-
-        def prep_images(self, batch):
-            return [None for _ in batch]
-
-        def predict(self, batch):
-            return [None for _ in batch]
-
-        def predict_proba(self, batch):
-            return [None for _ in batch]
-
-    return ModelHandleMock()
-
-
 @pytest.fixture
 def real_pdf():
     with open(os.path.join(TEST_DATA_DIR, "f2dc689ca794fccb8cd38b95f2bf6ba9.pdf"), "rb") as f:
@@ -465,10 +300,6 @@ def pipeline():
     return pipeline
 
 
-def transform_equal(a, b):
-    return (list(a) if isinstance(a, map) else a) == b
-
-
 def get_base_position_metadata(width, height, page_width, page_height):
     return {
         Info.WIDTH: width,
@@ -504,23 +335,3 @@ def page_height(request):
 @pytest.fixture(params=[100, 310])
 def page_width(request):
     return request.param
-
-
-def random_single_color_image_from_metadata(metadata):
-    image = Image.new(
-        "RGB", (metadata[Info.WIDTH], metadata[Info.HEIGHT]), color=tuple(map(int, np.random.uniform(size=3) * 255))
-    )
-    return image
-
-
-def gray_image_from_metadata(metadata):
-    image = Image.new("RGB", (metadata[Info.WIDTH], metadata[Info.HEIGHT]), color=(100, 100, 100))
-    return image
-
-
-def images_equal(im1: Image, im2: Image, **kwargs):
-    return np.allclose(image_to_normalized_tensor(im1), image_to_normalized_tensor(im2), **kwargs)
-
-
-def metadata_equal(mdat1: Iterable, mdat2: Iterable):
-    return set(map(frozendict, mdat1)) == set(map(frozendict, mdat2))

test/unit_tests/compositor_test.py

@@ -4,7 +4,7 @@ from image_prediction.compositor.compositor import TransformerCompositor
 from image_prediction.formatter.formatters.camel_case import Snake2CamelCaseKeyFormatter
 from image_prediction.formatter.formatters.enum import EnumFormatter
 from image_prediction.formatter.formatters.identity import IdentityFormatter
-from test.conftest import transform_equal
+from test.utils.comparison import transform_equal
 
 
 def test_identity(metadata):

test/unit_tests/coordinate_transformer_test.py

@@ -12,7 +12,10 @@ from image_prediction.info import Info
 from image_prediction.transformer.transformers.coordinate.fitz import FitzCoordinateTransformer
 from image_prediction.transformer.transformers.coordinate.fpdf import FPDFCoordinateTransformer
 from image_prediction.transformer.transformers.coordinate.pdfnet import PDFNetCoordinateTransformer
-from test.conftest import array_to_image, add_image, transform_equal, get_base_position_metadata
+from test.conftest import get_base_position_metadata
+from test.utils.generation.image import array_to_image
+from test.utils.generation.pdf import add_image
+from test.utils.comparison import transform_equal
 
 
 @pytest.mark.parametrize("coordinate_system", ["fpdf"])

test/unit_tests/image_extractor_test.py

@@ -11,7 +11,8 @@ from image_prediction.extraction import extract_images_from_pdf
 from image_prediction.image_extractor.extractor import ImageMetadataPair
 from image_prediction.image_extractor.extractors.parsable import extract_pages, get_image_infos, has_alpha_channel
 from image_prediction.info import Info
-from test.conftest import add_image, pdf_stream, images_equal, metadata_equal
+from test.utils.generation.pdf import add_image, pdf_stream
+from test.utils.comparison import images_equal, metadata_equal, image_sets_equal
 
 
 @pytest.mark.parametrize("extractor_type", ["mock"])
@@ -27,7 +28,7 @@ def test_image_extractor_mock(image_extractor, images):
 def test_parsable_pdf_image_extractor(image_extractor, pdf, images, metadata, input_size, alpha):
     images_extracted, metadata_extracted = map(list, extract_images_from_pdf(pdf, image_extractor))
     if not alpha:
-        all(any(images_equal(imex, im) for im in images) for imex in images_extracted)
+        assert image_sets_equal(images_extracted, images)
     assert metadata_equal(metadata_extracted, metadata)
 
 

test/unit_tests/image_stitching_test.py

@@ -30,12 +30,9 @@ from image_prediction.stitching.utils import (
     make_coord_getter,
     make_length_getter,
 )
-from test.conftest import (
-    add_image,
-    random_single_color_image_from_metadata,
-    gray_image_from_metadata,
-    images_equal,
-)
+from test.utils.generation.pdf import add_image
+from test.utils.generation.image import random_single_color_image_from_metadata, gray_image_from_metadata
+from test.utils.comparison import images_equal
 from test.utils.stitching import BoxSplitter
 
 x1_getter, y1_getter, x2_getter, y2_getter = map(make_coord_getter, ("x1", "y1", "x2", "y2"))

test/utils/comparison.py

@@ -0,0 +1,23 @@
+from typing import Iterable
+
+import numpy as np
+from PIL import Image
+from frozendict import frozendict
+
+from image_prediction.estimator.preprocessor.utils import image_to_normalized_tensor
+
+
+def transform_equal(a, b):
+    return (list(a) if isinstance(a, map) else a) == b
+
+
+def images_equal(im1: Image, im2: Image, **kwargs):
+    return np.allclose(image_to_normalized_tensor(im1), image_to_normalized_tensor(im2), **kwargs)
+
+
+def metadata_equal(mdat1: Iterable, mdat2: Iterable):
+    return set(map(frozendict, mdat1)) == set(map(frozendict, mdat2))
+
+
+def image_sets_equal(ims1, ims2):
+    return all(any(images_equal(im1, im2) for im2 in ims2) for im1 in ims1)

test/utils/generation/image.py

@@ -0,0 +1,31 @@
+import numpy as np
+from PIL import Image
+
+from image_prediction.info import Info
+
+
+def random_single_color_image_from_metadata(metadata):
+    image = Image.new(
+        "RGB", (metadata[Info.WIDTH], metadata[Info.HEIGHT]), color=tuple(map(int, np.random.uniform(size=3) * 255))
+    )
+    return image
+
+
+def gray_image_from_metadata(metadata):
+    image = Image.new("RGB", (metadata[Info.WIDTH], metadata[Info.HEIGHT]), color=(100, 100, 100))
+    return image
+
+
+def array_to_image(array):
+    assert np.all(array <= 1)
+    assert np.all(array >= 0)
+
+    if array.shape[-1] == 3:
+        mode = "RGB"
+    elif array.shape[-1] == 4:
+        mode = "RGBA"
+    else:
+        raise ValueError(f"Unexpected number of channels {array.shape[-1]}. Expected 3 or 4.")
+
+    # noinspection PyTypeChecker
+    return Image.fromarray(np.uint8(array * 255), mode=mode)

test/utils/generation/pdf.py

@@ -0,0 +1,30 @@
+import tempfile
+from operator import itemgetter
+
+import fpdf
+
+from image_prediction.info import Info
+
+
+def add_image(pdf, image_metadata_pair, suffix="png"):
+    while fewer_pages_then_required(image_metadata_pair.metadata[Info.PAGE_IDX], pdf):
+        pdf.add_page()
+
+    add_image_to_last_page(pdf, image_metadata_pair, suffix=suffix)
+
+
+def fewer_pages_then_required(page_idx, pdf):
+    return page_idx > pdf.page - 1
+
+
+def pdf_stream(pdf: fpdf.fpdf.FPDF):
+    return pdf.output(dest="S").encode("latin1")
+
+
+def add_image_to_last_page(pdf: fpdf.fpdf.FPDF, image_metadata_pair, suffix):
+    image, metadata = image_metadata_pair
+    x, y, w, h = itemgetter(Info.X1, Info.Y1, Info.WIDTH, Info.HEIGHT)(metadata)
+
+    with tempfile.NamedTemporaryFile(suffix=f".{suffix}") as temp_image:
+        image.save(temp_image.name)
+        pdf.image(temp_image.name, x=x, y=y, w=w, h=h, type=suffix)

test/utils/model.py

@@ -0,0 +1,127 @@
+import pytest
+
+from image_prediction.classifier.classifier import Classifier
+from image_prediction.classifier.image_classifier import ImageClassifier
+from image_prediction.estimator.adapter.adapter import EstimatorAdapter
+from image_prediction.estimator.preprocessor.preprocessors.basic import BasicPreprocessor
+from image_prediction.exceptions import UnknownImageExtractor, UnknownEstimatorAdapter
+from image_prediction.image_extractor.extractors.mock import ImageExtractorMock
+from image_prediction.image_extractor.extractors.parsable import ParsablePDFImageExtractor
+from image_prediction.redai_adapter.model import PredictionModelHandle
+
+
+@pytest.fixture
+def estimator_mock():
+    class EstimatorMock:
+        @staticmethod
+        def predict(batch):
+            return [None for _ in batch]
+
+        @staticmethod
+        def predict_proba(batch):
+            return [None for _ in batch]
+
+        def __call__(self, batch):
+            return self.predict(batch)
+
+    return EstimatorMock()
+
+
+@pytest.fixture
+def image_extractor(extractor_type):
+    if extractor_type == "mock":
+        return ImageExtractorMock()
+    elif extractor_type == "parsable_pdf":
+        return ParsablePDFImageExtractor()
+    elif extractor_type == "default":
+        return None
+    else:
+        raise UnknownImageExtractor(f"No image extractor for type {extractor_type} was specified.")
+
+
+@pytest.fixture
+def image_classifier(classifier, monkeypatch, batch_of_expected_string_labels):
+    return ImageClassifier(classifier, preprocessor=BasicPreprocessor())
+
+
+@pytest.fixture
+def classifier(estimator_adapter, label_mapper):
+    classifier = Classifier(estimator_adapter, label_mapper)
+    return classifier
+
+
+@pytest.fixture
+def estimator_adapter(
+    estimator_type, estimator_mock, keras_model, model_handle_mock, output_batch_generator, monkeypatch
+):
+    if estimator_type == "mock":
+        estimator_adapter = EstimatorAdapter(estimator_mock)
+    elif estimator_type == "keras":
+        estimator_adapter = EstimatorAdapter(keras_model)
+    elif estimator_type == "redai":
+        estimator_adapter = EstimatorAdapter(PredictionModelHandle(model_handle_mock))
+    else:
+        raise UnknownEstimatorAdapter(f"No adapter for estimator type {estimator_type} was specified.")
+
+    def mock_predict(batch):
+        # Run real predict function to test for mechanical issues, but return externally defined
+        # predictions to test the callers of the estimator adapter against the expected predictions
+        return [next(output_batch_generator) for _ in _predict(batch)]
+
+    _predict = estimator_adapter.predict
+    monkeypatch.setattr(estimator_adapter, "predict", mock_predict)
+
+    return estimator_adapter
+
+
+@pytest.fixture
+def keras_model(input_size):
+    import os
+
+    os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
+
+    import tensorflow as tf
+
+    tf.keras.backend.set_image_data_format("channels_last")
+
+    inputs = tf.keras.Input(shape=input_size)
+    conv = tf.keras.layers.Conv2D(3, 3)
+    dense = tf.keras.layers.Dense(10)
+
+    outputs = tf.keras.layers.Dense(10)(dense(conv(inputs)))
+    model = tf.keras.Model(inputs=inputs, outputs=outputs)
+    model.compile()
+
+    return model
+
+
+@pytest.fixture
+def model():
+    class Model:
+        @staticmethod
+        def predict(*args):
+            return True
+
+        @staticmethod
+        def predict_proba(*args):
+            return True
+
+    return Model()
+
+
+@pytest.fixture
+def model_handle_mock(estimator_mock):
+    class ModelHandleMock:
+        def __init__(self):
+            self.model = estimator_mock
+
+        def prep_images(self, batch):
+            return [None for _ in batch]
+
+        def predict(self, batch):
+            return [None for _ in batch]
+
+        def predict_proba(self, batch):
+            return [None for _ in batch]
+
+    return ModelHandleMock()