[FIX] do not panic on unreadable files (#95)

* skip files that cannot be read by openslide

* add func test_issue_94

* reformat with black

* add torch and torchvision to install_requires

setup.cfg

@@ -30,9 +30,6 @@ classifiers =
 packages = find:
 python_requires = >= 3.7
 install_requires =
-    # We require torch and torchvision but do not include them here because their
-    # installation varies by platform and hardware.
-    # See https://pytorch.org/get-started/locally/.
     click>=8.0,<9
     h5py
     # OpenSlide and TIFF readers should handle all images we will encounter.
@@ -43,6 +40,11 @@ install_requires =
     pillow
     pyyaml
     timm
+    # The installation fo torch and torchvision can differ by hardware. Users are
+    # advised to install torch and torchvision for their given hardware and then install
+    # wsinfer. See https://pytorch.org/get-started/locally/.
+    torch>=1.7
+    torchvision
     tqdm
 
 [options.extras_require]

tests/test_all.py

@@ -53,7 +53,6 @@ def tiff_image(tmp_path: Path) -> Path:
 
 
 def test_cli_list(tmp_path: Path):
-
     from wsinfer.cli.cli import cli
 
     runner = CliRunner()
@@ -967,6 +966,7 @@ def test_jit_compile(model_name: str, weights_name: str):
 
 
 def test_issue_89():
+    """Do not fail if 'git' is not installed."""
     from wsinfer.cli.infer import _get_info_for_save
 
     w = get_model_weights("resnet34", "TCGA-BRCA-v1")
@@ -988,3 +988,34 @@ def test_issue_89():
         assert d["runtime"]["git"] is None
     finally:
         os.environ["PATH"] = orig_path  # reset path
+
+
+def test_issue_94(tmp_path: Path, tiff_image: Path):
+    """Gracefully handle unreadable slides."""
+    from wsinfer.cli.cli import cli
+
+    # We have a valid tiff in 'tiff_image.parent'. We put in an unreadable file too.
+    badpath = tiff_image.parent / "bad.svs"
+    badpath.touch()
+
+    runner = CliRunner()
+    results_dir = tmp_path / "inference"
+    result = runner.invoke(
+        cli,
+        [
+            "run",
+            "--wsi-dir",
+            str(tiff_image.parent),
+            "--model",
+            "resnet34",
+            "--weights",
+            "TCGA-BRCA-v1",
+            "--results-dir",
+            str(results_dir),
+        ],
+    )
+    # Important part is that we run through all of the files, despite the unreadble
+    # file.
+    assert result.exit_code == 0
+    assert results_dir.joinpath("model-outputs").joinpath("purple.csv").exists()
+    assert not results_dir.joinpath("model-outputs").joinpath("bad.csv").exists()

wsinfer/_patchlib/create_patches_fp.py

@@ -118,7 +118,6 @@ def seg_and_patch(
     process_list=None,
     patch_spacing=None,
 ):
-
     slides = sorted(os.listdir(source))
     slides = [slide for slide in slides if os.path.isfile(os.path.join(source, slide))]
     if process_list is None:
@@ -160,7 +159,7 @@ def seg_and_patch(
         df.to_csv(os.path.join(save_dir, "process_list_autogen.csv"), index=False)
         idx = process_stack.index[i]
         slide = process_stack.loc[idx, "slide_id"]
-        print("\n\nprogress: {:.2f}, {}/{}".format(i / total, i, total))
+        print("\n\nprogress: {:.1%}, {}/{}".format(i / total, i + 1, total))
         print("processing {}".format(slide))
 
         df.loc[idx, "process"] = 0
@@ -173,7 +172,12 @@ def seg_and_patch(
 
         # Inialize WSI
         full_path = os.path.join(source, slide)
-        WSI_object = WholeSlideImage(full_path)
+        # Some slide files might be malformed and unreadable. Skip them.
+        try:
+            WSI_object = WholeSlideImage(full_path)
+        except Exception:
+            print(f"Failed to load slide, skipping {full_path}")
+            continue
 
         if use_default_params:
             current_vis_params = vis_params.copy()

wsinfer/_patchlib/utils/utils.py

@@ -182,7 +182,6 @@ def generate_split(
             all_val_ids.extend(val_ids)
 
             if custom_test_ids is None:  # sample test split
-
                 test_ids = np.random.choice(remaining_ids, test_num[c], replace=False)
                 remaining_ids = np.setdiff1d(remaining_ids, test_ids)
                 all_test_ids.extend(test_ids)

wsinfer/_patchlib/wsi_core/WholeSlideImage.py

@@ -28,7 +28,6 @@
 
 class WholeSlideImage(object):
     def __init__(self, path):
-
         """
         Args:
             path (str): fullpath to WSI file
@@ -242,7 +241,6 @@ def visWSI(
         seg_display=True,
         annot_display=True,
     ):
-
         downsample = self.level_downsamples[vis_level]
         scale = [1 / downsample[0], 1 / downsample[1]]
 
@@ -456,7 +454,6 @@ def _getPatchGenerator(
         count = 0
         for y in range(start_y, stop_y, step_size_y):
             for x in range(start_x, stop_x, step_size_x):
-
                 if not self.isInContours(
                     cont_check_fn,
                     (x, y),

wsinfer/_patchlib/wsi_core/wsi_utils.py

@@ -204,7 +204,6 @@ def sample_rois(
     top_left=None,
     bot_right=None,
 ):
-
     if len(scores.shape) == 2:
         scores = scores.flatten()