[torchcodec] refactor test utils into its own library (#25)

Summary:
Pull Request resolved: #25

Refactors utility functions that were defined directly in test source files into a library that can be shared among tests. This diff:

1. Creates a Python test utility library that can be imported by Python tests.
2. Moves the test resources to the test top-level.
3. Defines a C++ target for those resources.

This is in preparation for adding new tests that will need this library.

Differential Revision: D58530481

test/decoders/VideoDecoderOpsTest.cpp

@@ -17,8 +17,7 @@ namespace facebook::torchcodec {
 
 std::string getResourcePath(const std::string& filename) {
 #ifdef FBCODE_BUILD
-  std::string filepath =
-      "pytorch/torchcodec/test/decoders/resources/" + filename;
+  std::string filepath = "pytorch/torchcodec/test/resources/" + filename;
   filepath = build::getResourcePath(filepath).string();
 #else
   std::filesystem::path dirPath = std::filesystem::path(__FILE__);

test/decoders/VideoDecoderTest.cpp

@@ -24,8 +24,7 @@ namespace facebook::torchcodec {
 
 std::string getResourcePath(const std::string& filename) {
 #ifdef FBCODE_BUILD
-  std::string filepath =
-      "pytorch/torchcodec/test/decoders/resources/" + filename;
+  std::string filepath = "pytorch/torchcodec/test/resources/" + filename;
   filepath = build::getResourcePath(filepath).string();
 #else
   std::filesystem::path dirPath = std::filesystem::path(__FILE__);

test/decoders/manual_smoke_test.py

@@ -4,7 +4,7 @@
 import torchcodec
 
 decoder = torchcodec.decoders._core.create_from_file(
-    os.path.dirname(__file__) + "/resources/nasa_13013.mp4"
+    os.path.dirname(__file__) + "../resources/nasa_13013.mp4"
 )
 torchcodec.decoders._core.add_video_stream(decoder, stream_index=3)
 frame = torchcodec.decoders._core.get_frame_at_index(

test/decoders/video_decoder_ops_test.py

@@ -10,7 +10,6 @@
 import pytest
 
 import torch
-import torchvision.transforms as transforms
 from PIL import Image
 
 from torchcodec.decoders._core import (
@@ -26,42 +25,14 @@
     seek_to_pts,
 )
 
-torch._dynamo.config.capture_dynamic_output_shape_ops = True
-IN_FBCODE = os.environ.get("IN_FBCODE_TORCHCODEC") == "1"
-
-
-# TODO: Eventually move that as a common test util
-def assert_equal(*args, **kwargs):
-    torch.testing.assert_close(*args, **kwargs, atol=0, rtol=0)
-
-
-# TODO: Eventually move that as a common test util
-def get_video_path(filename: str) -> pathlib.Path:
-    if IN_FBCODE:
-        resource = (
-            importlib.resources.files(__package__)
-            .joinpath("resources")
-            .joinpath(filename)
-        )
-        with importlib.resources.as_file(resource) as path:
-            return path
-    else:
-        return pathlib.Path(__file__).parent / "resources" / filename
-
-
-# TODO: make this a fixture or wrap with @functools.lru_cache to avoid
-# re-computing?
-def load_tensor_from_file(filename: str) -> torch.Tensor:
-    file_path = get_video_path(filename)
-    return torch.load(file_path)
-
-
-def get_reference_video_path() -> pathlib.Path:
-    return get_video_path("nasa_13013.mp4")
-
+from ..test_utils import (
+    assert_equal,
+    get_reference_audio_path,
+    get_reference_video_path,
+    load_tensor_from_file,
+)
 
-def get_reference_audio_path() -> pathlib.Path:
-    return get_video_path("nasa_13013.mp4.audio.mp3")
+torch._dynamo.config.capture_dynamic_output_shape_ops = True
 
 
 class ReferenceDecoder:

test/samplers/video_clip_sampler_test.py

@@ -15,23 +15,7 @@
     VideoClipSampler,
 )
 
-
-# TODO: move this to a common util
-IN_FBCODE = os.environ.get("IN_FBCODE_TORCHCODEC") == "1"
-
-
-# TODO: Eventually rely on common util for this
-@pytest.fixture()
-def nasa_13013() -> torch.Tensor:
-    if IN_FBCODE:
-        video_path = importlib.resources.path(__package__, "nasa_13013.mp4")
-    else:
-        video_path = (
-            Path(__file__).parent.parent / "decoders" / "resources" / "nasa_13013.mp4"
-        )
-    arr = np.fromfile(video_path, dtype=np.uint8)
-    video_tensor = torch.from_numpy(arr)
-    return video_tensor
+from ..test_utils import assert_equal, nasa_13013  # noqa: F401; see nasa_13013 use
 
 
 @pytest.mark.parametrize(
@@ -51,13 +35,13 @@ def nasa_13013() -> torch.Tensor:
         ),
     ],
 )
-def test_sampler(sampler_args, nasa_13013):
+def test_sampler(sampler_args, nasa_13013):  # noqa: F811; linter does not see this as a use
     torch.manual_seed(0)
     desired_width, desired_height = 320, 240
     video_args = VideoArgs(desired_width=desired_width, desired_height=desired_height)
     sampler = VideoClipSampler(video_args, sampler_args)
     clips = sampler(nasa_13013)
-    assert len(clips) == sampler_args.clips_per_video
+    assert_equal(len(clips), sampler_args.clips_per_video)
     clip = clips[0]
     if isinstance(sampler_args, TimeBasedSamplerArgs):
         # TODO FIXME: Looks like we have an API inconsistency.

test/test_utils.py

@@ -0,0 +1,52 @@
+import importlib
+import os
+import pathlib
+
+import numpy as np
+import pytest
+
+import torch
+
+
+def in_fbcode() -> bool:
+    return os.environ.get("IN_FBCODE_TORCHCODEC") == "1"
+
+
+IN_FBCODE = in_fbcode()
+
+
+def assert_equal(*args, **kwargs):
+    torch.testing.assert_close(*args, **kwargs, atol=0, rtol=0)
+
+
+def get_video_path(filename: str) -> pathlib.Path:
+    if IN_FBCODE:
+        resource = (
+            importlib.resources.files(__spec__.parent)
+            .joinpath("resources")
+            .joinpath(filename)
+        )
+        with importlib.resources.as_file(resource) as path:
+            return path
+    else:
+        return pathlib.Path(__file__).parent / "resources" / filename
+
+
+def get_reference_video_path() -> pathlib.Path:
+    return get_video_path("nasa_13013.mp4")
+
+
+def get_reference_audio_path() -> pathlib.Path:
+    return get_video_path("nasa_13013.mp4.audio.mp3")
+
+
+def load_tensor_from_file(filename: str) -> torch.Tensor:
+    file_path = get_video_path(filename)
+    return torch.load(file_path)
+
+
+@pytest.fixture()
+def nasa_13013() -> torch.Tensor:
+    arr = np.fromfile(get_reference_video_path(), dtype=np.uint8)
+    video_tensor = torch.from_numpy(arr)
+    return video_tensor