Refactor asa tests for better isolation (#690)

* rev docs, module strings; refactor tests for better isolation (skip if extras not installed)

compiler/quilt/test/test_asa_plot.py

@@ -0,0 +1,114 @@
+"""test class against quilt.asa.plot"""
+import os
+
+import numpy as np
+import pytest
+
+from quilt.tools import command
+from .utils import QuiltTestCase, try_require
+
+if not try_require('quilt[img]'):
+    # pylint: disable=unexpected-keyword-arg
+    pytest.skip(
+        "only test if [img] extras installed",
+        allow_module_level=True)
+
+# pylint: disable=no-self-use
+class ImportTest(QuiltTestCase):
+    # the following two lines must happen first
+    import matplotlib as mpl
+    mpl.use('Agg') # specify a backend so headless unit tests don't barf
+
+    def test_asa_plot(self):
+        from quilt.asa.img import plot
+
+        mydir = os.path.dirname(__file__)
+        build_path = os.path.join(mydir, './build_img.yml')
+        command.build('foo/imgtest', build_path)
+        pkg = command.load('foo/imgtest')
+        # expect no exceptions on root
+        pkg(asa=plot())
+        # pylint: disable=no-member
+        # expect no exceptions on GroupNode with only DF children
+        pkg.dataframes(asa=plot())
+        # expect no exceptions on GroupNode with mixed children
+        pkg.mixed(asa=plot())
+        # expect no exceptions on dir of images
+        pkg.mixed.img(asa=plot())
+        pkg.mixed.img(asa=plot(formats=['jpg', 'png']))
+        # assert images != filtered, 'Expected only .jpg and .png images'
+        # expect no exceptions on single images
+        pkg.mixed.img.sf(asa=plot())
+        pkg.mixed.img.portal(asa=plot())
+
+    def _are_similar(self, ima, imb, error=0.01):
+        """predicate to see if images differ by less than
+        the given error; uses mean squared error; see also
+        https://www.pyimagesearch.com/2014/09/15/python-compare-two-images/
+
+        ima, imb: PIL.Image instances
+        """
+        ima_ = np.array(ima).astype('float')
+        imb_ = np.array(imb).astype('float')
+        assert ima_.shape == imb_.shape, 'ima and imb must have same shape'
+        # pylint: disable=invalid-name
+        for x, y, _ in (ima_.shape, imb_.shape):
+            assert x > 0 and y > 0, \
+                'unexpected image dimension: {}'.format((x, y))
+        # sum of normalized channel differences squared
+        error_ = np.sum(((ima_ - imb_)/255) ** 2)
+        # normalize by total number of samples
+        error_ /= float(ima_.shape[0] * imb_.shape[1])
+
+        return error_ < error
+
+    def test_asa_plot_output(self):
+        from PIL import Image
+        from matplotlib import pyplot as plt
+
+        from quilt.asa.img import plot
+
+        mydir = os.path.dirname(__file__)
+        build_path = os.path.join(mydir, 'build_img.yml')
+        command.build('foo/imgtest', build_path)
+        pkg = command.load('foo/imgtest')
+
+        outfile = os.path.join('.', 'temp-plot.png')
+        # pylint: disable=no-member
+        pkg.mixed.img(asa=plot(figsize=(10, 10)))
+        # size * dpi = 1000 x 1000 pixels
+        plt.savefig(outfile, dpi=100, format='png', transparent=False)
+
+        ref_path = os.path.join(mydir, 'data', 'ref-asa-plot.png')
+
+        ref_img = Image.open(ref_path)
+        tst_img = Image.open(outfile)
+
+        assert self._are_similar(ref_img, tst_img), \
+            'render differs from reference: {}'.format(ref_img)
+
+    def test_asa_plot_formats_output(self):
+        from PIL import Image
+        from matplotlib import pyplot as plt
+
+        from quilt.asa.img import plot
+
+        mydir = os.path.dirname(__file__)
+        build_path = os.path.join(mydir, 'build_img.yml')
+        command.build('foo/imgtest', build_path)
+        pkg = command.load('foo/imgtest')
+
+        outfile = os.path.join('.', 'temp-formats-plot.png')
+
+        # pylint: disable=no-member
+        pkg.mixed.img(asa=plot(figsize=(10, 10), formats=['png']))
+        # size * dpi = 1000 x 1000 pixels
+        plt.savefig(outfile, dpi=100, format='png', transparent=False)
+
+        ref_path = os.path.join(mydir, 'data', 'ref-asa-formats.png')
+
+        ref_img = Image.open(ref_path)
+        tst_img = Image.open(outfile)
+
+        assert self._are_similar(ref_img, tst_img), \
+            'render differs from reference: {}'.format(ref_img)

compiler/quilt/test/test_asa_torch.py

@@ -0,0 +1,89 @@
+"""test class against quilt.asa.torch"""
+import os
+
+import pytest
+from six import string_types
+
+from quilt.tools import command
+from quilt.nodes import DataNode
+from .utils import QuiltTestCase, try_require
+
+if not try_require('quilt[img,pytorch,torchvision]'):
+    # pylint: disable=unexpected-keyword-arg
+    pytest.skip("only test if [img,pytorch,torchvision] extras installed",
+        allow_module_level=True)
+
+# pylint: disable=no-self-use
+class ImportTest(QuiltTestCase):
+    def test_asa_pytorch(self):
+        """test asa.torch interface by converting a GroupNode with asa="""
+        from torchvision.transforms import Compose, CenterCrop, ToTensor, Resize
+        from torch.utils.data import Dataset
+        from PIL import Image
+        from torch import Tensor
+
+        from quilt.asa.pytorch import dataset
+        # pylint: disable=missing-docstring
+        # helper functions to simulate real pytorch dataset usage
+        def calculate_valid_crop_size(crop_size, upscale_factor):
+            return crop_size - (crop_size % upscale_factor)
+
+        def node_parser(node):
+            path = node()
+            if isinstance(path, string_types):
+                img = Image.open(path).convert('YCbCr')
+                chan, _, _ = img.split()
+                return chan
+            else:
+                raise TypeError('Expected string path to an image fragment')
+
+        def input_transform(crop_size, upscale_factor):
+            return Compose([
+                CenterCrop(crop_size),
+                Resize(crop_size // upscale_factor),
+                ToTensor(),
+            ])
+
+        def target_transform(crop_size):
+            def _inner(img):
+                img_ = img.copy()
+                return Compose([
+                    CenterCrop(crop_size),
+                    ToTensor(),
+                ])(img_)
+            return _inner
+        # pylint: disable=protected-access
+        def is_image(node):
+            """file extension introspection on Quilt nodes"""
+            if isinstance(node, DataNode):
+                filepath = node._meta.get('_system', {}).get('filepath')
+                if filepath:
+                    return any(
+                        filepath.endswith(extension)
+                        for extension in [".png", ".jpg", ".jpeg"])
+        # end helper functions
+
+        mydir = os.path.dirname(__file__)
+        build_path = os.path.join(mydir, 'build_img.yml')
+        command.build('foo/torchtest', build_path)
+        pkg = command.load('foo/torchtest')
+
+        upscale_factor = 3
+        crop_size = calculate_valid_crop_size(256, upscale_factor)
+        # pylint: disable=no-member
+        my_dataset = pkg.mixed.img(asa=dataset(
+            include=is_image,
+            node_parser=node_parser,
+            input_transform=input_transform(crop_size, upscale_factor),
+            target_transform=target_transform(crop_size)
+        ))
+        assert isinstance(my_dataset, Dataset), \
+            'expected type {}, got {}'.format(type(Dataset), type(my_dataset))
+
+        assert my_dataset.__len__() == 2, \
+            'expected two images in mixed.img, got {}'.format(my_dataset.__len__())
+
+        for i in range(my_dataset.__len__()):
+            tens = my_dataset.__getitem__(i)
+            assert all((isinstance(x, Tensor) for x in tens)), \
+                'Expected all torch.Tensors in tuple, got {}'.format(tens)

compiler/quilt/test/test_import.py

@@ -5,23 +5,16 @@
 import platform
 import time
 
-# the following two lines must happen first
-import matplotlib as mpl
-mpl.use('Agg') # specify a backend so renderer doesn't barf
-# pylint: disable=wrong-import-position
-from PIL import Image
 import numpy as np
 import pandas as pd
 import pytest
-from matplotlib import pyplot as plt
 from six import string_types
 
 from quilt.tools import command
 from quilt.nodes import DataNode, GroupNode
 from quilt.tools.const import PACKAGE_DIR_NAME
 from quilt.tools.package import Package
 from quilt.tools.store import PackageStore, StoreException
-from quilt.asa.img import plot
 from .utils import patch, QuiltTestCase
 
  # pylint: disable=protected-access
@@ -492,160 +485,6 @@ def test_lambda(node, hashes):
         assert pkg.dataframes(asa=test_lambda) is testdata
         assert pkg(asa=test_lambda) is testdata
 
-    # pylint: disable=no-member
-    def test_asa_plot(self):
-        mydir = os.path.dirname(__file__)
-        build_path = os.path.join(mydir, './build_img.yml')
-        command.build('foo/imgtest', build_path)
-        pkg = command.load('foo/imgtest')
-        # expect no exceptions on root
-        pkg(asa=plot())
-        # expect no exceptions on GroupNode with only DF children
-        pkg.dataframes(asa=plot())
-        # expect no exceptions on GroupNode with mixed children
-        pkg.mixed(asa=plot())
-        # expect no exceptions on dir of images
-        pkg.mixed.img(asa=plot())
-        pkg.mixed.img(asa=plot(formats=['jpg', 'png']))
-        # assert images != filtered, 'Expected only .jpg and .png images'
-        # expect no exceptions on single images
-        pkg.mixed.img.sf(asa=plot())
-        pkg.mixed.img.portal(asa=plot())
-
-    def _are_similar(self, ima, imb, error=0.01):
-        """predicate to see if images differ by less than
-        the given error; uses mean squared error; see also
-        https://www.pyimagesearch.com/2014/09/15/python-compare-two-images/
-
-        ima, imb: PIL.Image instances
-        """
-        ima_ = np.array(ima).astype('float')
-        imb_ = np.array(imb).astype('float')
-        assert ima_.shape == imb_.shape, 'ima and imb must have same shape'
-        for x, y, _ in (ima_.shape, imb_.shape):
-            assert x > 0 and y > 0, 'unexpected image dimension: {}'.format(shape)
-        # sum of normalized channel differences squared
-        error_ = np.sum(((ima_ - imb_)/255) ** 2)
-        # normalize by total number of samples
-        error_ /= float(ima_.shape[0] * imb_.shape[1])
-
-        return error_ < error
-
-    # pylint: disable=no-member
-    def test_asa_plot_output(self):
-        mydir = os.path.dirname(__file__)
-        build_path = os.path.join(mydir, 'build_img.yml')
-        command.build('foo/imgtest', build_path)
-        pkg = command.load('foo/imgtest')
-
-        outfile = os.path.join('.', 'temp-plot.png')
-        pkg.mixed.img(asa=plot(figsize=(10, 10)))
-        # size * dpi = 1000 x 1000 pixels
-        plt.savefig(outfile, dpi=100, format='png', transparent=False)
-
-        ref_path = os.path.join(mydir, 'data', 'ref-asa-plot.png')
-
-        ref_img = Image.open(ref_path)
-        tst_img = Image.open(outfile)
-
-        assert self._are_similar(ref_img, tst_img), \
-            'render differs from reference: {}'.format(ref_img)
-
-    # pylint: disable=no-member
-    def test_asa_plot_formats_output(self):
-        mydir = os.path.dirname(__file__)
-        build_path = os.path.join(mydir, 'build_img.yml')
-        command.build('foo/imgtest', build_path)
-        pkg = command.load('foo/imgtest')
-
-        outfile = os.path.join('.', 'temp-formats-plot.png')
-        pkg.mixed.img(asa=plot(figsize=(10, 10), formats=['png']))
-        # size * dpi = 1000 x 1000 pixels
-        plt.savefig(outfile, dpi=100, format='png', transparent=False)
-
-        ref_path = os.path.join(mydir, 'data', 'ref-asa-formats.png')
-
-        ref_img = Image.open(ref_path)
-        tst_img = Image.open(outfile)
-
-        assert self._are_similar(ref_img, tst_img), \
-            'render differs from reference: {}'.format(ref_img)
-
-
-    @pytest.mark.xfail(platform.system() in ['Windows'], reason=(
-      "infeasible to install pytorch on appveyor (even with conda)"
-    ))
-    def test_asa_pytorch(self):
-        """test asa.torch interface by converting a GroupNode with asa="""
-        from torchvision.transforms import Compose, CenterCrop, ToTensor, Resize
-        from torch.utils.data import Dataset
-        from torch import Tensor
-
-        from quilt.asa.pytorch import dataset
-        # pylint: disable=missing-docstring
-        # helper functions to simulate real pytorch dataset usage
-        def calculate_valid_crop_size(crop_size, upscale_factor):
-            return crop_size - (crop_size % upscale_factor)
-
-        def node_parser(node):
-            path = node()
-            if isinstance(path, string_types):
-                img = Image.open(path).convert('YCbCr')
-                chan, _, _ = img.split()
-                return chan
-            else:
-                raise TypeError('Expected string path to an image fragment')
-
-        def input_transform(crop_size, upscale_factor):
-            return Compose([
-                CenterCrop(crop_size),
-                Resize(crop_size // upscale_factor),
-                ToTensor(),
-            ])
-
-        def target_transform(crop_size):
-            def _inner(img):
-                img_ = img.copy()
-                return Compose([
-                    CenterCrop(crop_size),
-                    ToTensor(),
-                ])(img_)
-            return _inner
-
-        def is_image(node):
-            """file extension introspection on Quilt nodes"""
-            if isinstance(node, DataNode):
-                filepath = node._meta.get('_system', {}).get('filepath')
-                if filepath:
-                    return any(
-                        filepath.endswith(extension)
-                        for extension in [".png", ".jpg", ".jpeg"])
-        # end helper functions
-
-        mydir = os.path.dirname(__file__)
-        build_path = os.path.join(mydir, 'build_img.yml')
-        command.build('foo/torchtest', build_path)
-        pkg = command.load('foo/torchtest')
-
-        upscale_factor = 3
-        crop_size = calculate_valid_crop_size(256, upscale_factor)
-        my_dataset = pkg.mixed.img(asa=dataset(
-            include=is_image,
-            node_parser=node_parser,
-            input_transform=input_transform(crop_size, upscale_factor),
-            target_transform=target_transform(crop_size)
-        ))
-        assert isinstance(my_dataset, Dataset), \
-            'expected type {}, got {}'.format(type(Dataset), type(my_dataset))
-
-        assert my_dataset.__len__() == 2, \
-            'expected two images in mixed.img, got {}'.format()
-
-        for i in range(my_dataset.__len__()):
-            tens = my_dataset.__getitem__(i);
-            assert all((isinstance(x, Tensor) for x in tens)), \
-                'Expected all torch.Tensors in tuple, got {}'.format(tens)
-
     def test_memory_only_datanode_asa(self):
         testdata = "justatest"
         def test_lambda(node, hashes):