apply ruff to experiments, examples, benchmarks, scripts as well

CONTRIBUTING.md

@@ -71,20 +71,30 @@ conda install isort black ruff
 pip install isort black ruff
 ```
 
-These tools are used to auto-format the Python code in the repository. Additionally, there is a CI check in place to enforce that the committed code follows our standards. You can use the tools to automatically format your python code by running:
+These tools are used to auto-format the Python code in the repository. Additionally, there is a CI check in place to enforce that the committed code follows our standards. To run only for the python/cucim folder, change to that folder and run
+
+```bash
+isort .
+black .
+ruff .
+```
+
+To also check formatting in top-level folders like `benchmarks`, `examples` and `experiments`, these tools can also be run from the top level of the repository as follows:
 
 ```bash
 isort --settings-path="python/cucim/pyproject.toml" .
-ruff python/cucim
 black --config python/cucim/pyproject.toml .
+ruff .
 ```
 
-In addition to these tools, [codespell]() can be used to help diagnose and interactively fix spelling errors in both Python and C++ code. It can also be run from the top level of the repository in interactive mode using:
+In addition to these tools, [codespell](https://github.com/codespell-project/codespell) can be used to help diagnose and interactively fix spelling errors in both Python and C++ code. It can also be run from the top level of the repository in interactive mode using:
 
 ```bash
 codespell --toml python/cucim/pyproject.toml . -i 3 -w
 ```
 
+If codespell is finding false positives in newly added code, the `ignore-words-list` entry of the `tool.codespell` section in `pyproject.toml` can be updated as needed.
+
 ### Get libcucim Dependencies
 
 Compiler requirements:

benchmarks/skimage/_image_bench.py

@@ -33,7 +33,6 @@ def __init__(
         fixed_kwargs={},
         var_kwargs={},
         index_str=None,  # extra string to append to dataframe index
-        # set_args_kwargs={},  # for passing additional arguments to custom set_args method
         module_cpu=scipy.ndimage,
         module_gpu=cupyx.scipy.ndimage,
         function_is_generator=False,
@@ -181,7 +180,7 @@ def run_benchmark(self, duration=3, verbose=True):
                 df.at[index, "dtype"] = np.dtype(dtype).name
                 df.at[index, "ndim"] = len(self.shape)
 
-                if self.run_cpu == True:
+                if self.run_cpu is True:
                     perf = repeat(
                         self.func_cpu, self.args_cpu, kw_cpu, **rep_kwargs_cpu
                     )

benchmarks/skimage/bench_convolve.py

@@ -3,10 +3,9 @@
 """
 import cupy as cp
 import cupyx.scipy.ndimage as ndi
-import pytest
 from cupyx.profiler import benchmark
 
-from cucim.skimage._vendored.ndimage import (
+from cucim.skimage._vendored.ndimage import (  # noqa: F401
     convolve1d,
     correlate1d,
     gaussian_filter,
@@ -134,7 +133,6 @@ def _compare_implementations_other(
         if output_dtype is None:
             output_dtype = image.dtype
         output1 = cp.empty(image.shape, dtype=output_dtype)
-        output2 = cp.empty(image.shape, dtype=output_dtype)
         kwargs.update(dict(output=output1))
         perf1 = benchmark(
             function_ref,
@@ -177,10 +175,10 @@ def _compare_implementations_other(
 
 print("\n\n")
 print(
-    "function | shape | dtype | mode | kernel size | preallocated | axis | dur (ms), CuPy | dur (ms), cuCIM | acceleration "
+    "function | shape | dtype | mode | kernel size | preallocated | axis | dur (ms), CuPy | dur (ms), cuCIM | acceleration "  # noqa: E501
 )
 print(
-    "---------|-------|-------|------|-------------|--------------|------|----------------|-----------------|--------------"
+    "---------|-------|-------|------|-------------|--------------|------|----------------|-----------------|--------------"  # noqa: E501
 )
 for function in [convolve1d]:
     for shape in [(512, 512), (3840, 2160), (64, 64, 64), (256, 256, 256)]:
@@ -203,15 +201,15 @@ def _compare_implementations_other(
                             t_elem = perf1.gpu_times * 1000.0
                             t_shared = perf2.gpu_times * 1000.0
                             print(
-                                f"{function.__name__} | {shape} | {cp.dtype(dtype).name} | {mode} | {kernel_size=} | prealloc={output_preallocated} | {axis=} | {t_elem.mean():0.3f} +/- {t_elem.std():0.3f}  | {t_shared.mean():0.3f} +/- {t_shared.std():0.3f} | {t_elem.mean() / t_shared.mean():0.3f}"
+                                f"{function.__name__} | {shape} | {cp.dtype(dtype).name} | {mode} | {kernel_size=} | prealloc={output_preallocated} | {axis=} | {t_elem.mean():0.3f} +/- {t_elem.std():0.3f}  | {t_shared.mean():0.3f} +/- {t_shared.std():0.3f} | {t_elem.mean() / t_shared.mean():0.3f}"  # noqa: E501
                             )
 
 
 print(
-    "function | kwargs | shape | dtype | mode | preallocated | dur (ms), CuPy | dur (ms), cuCIM | acceleration "
+    "function | kwargs | shape | dtype | mode | preallocated | dur (ms), CuPy | dur (ms), cuCIM | acceleration "  # noqa: E501
 )
 print(
-    "---------|--------|-------|-------|------|--------------|----------------|-----------------|--------------"
+    "---------|--------|-------|-------|------|--------------|----------------|-----------------|--------------"  # noqa: E501
 )
 for function, func_kwargs in [
     # (gaussian_filter1d, dict(sigma=1.0, axis=0)),
@@ -245,5 +243,5 @@ def _compare_implementations_other(
                     t_elem = perf1.gpu_times * 1000.0
                     t_shared = perf2.gpu_times * 1000.0
                     print(
-                        f"{function.__name__} | {func_kwargs} | {shape} | {cp.dtype(dtype).name} | {mode} | {output_preallocated} | {t_elem.mean():0.3f} +/- {t_elem.std():0.3f}  | {t_shared.mean():0.3f} +/- {t_shared.std():0.3f} | {t_elem.mean() / t_shared.mean():0.3f}"
+                        f"{function.__name__} | {func_kwargs} | {shape} | {cp.dtype(dtype).name} | {mode} | {output_preallocated} | {t_elem.mean():0.3f} +/- {t_elem.std():0.3f}  | {t_shared.mean():0.3f} +/- {t_shared.std():0.3f} | {t_elem.mean() / t_shared.mean():0.3f}"  # noqa: E501
                     )

benchmarks/skimage/cucim_color_bench.py

@@ -124,8 +124,6 @@ def main(args):
 
     all_colorspaces = False
 
-    ndim = len(shape)
-
     for function_name in func_name_choices:
         if function_name != args.func_name:
             continue
@@ -224,7 +222,7 @@ def main(args):
     all_results.to_csv(fbase + ".csv")
     all_results.to_pickle(pfile)
     try:
-        import tabular
+        import tabular  # noqa: F401
 
         with open(fbase + ".md", "wt") as f:
             f.write(all_results.to_markdown())

benchmarks/skimage/cucim_exposure_bench.py

@@ -83,8 +83,6 @@ def main(args):
         if function_name != args.func_name:
             continue
 
-        ndim = len(shape)
-
         if function_name == "match_histograms":
             channel_axis = -1 if shape[-1] in [3, 4] else None
 
@@ -106,7 +104,8 @@ def main(args):
                 continue
 
             if function_name == "equalize_adapthist":
-                #  TODO: fix equalize_adapthist for size (3840, 2160) and kernel_size = [16, 16]
+                #  TODO: fix equalize_adapthist for size (3840, 2160)
+                #        and kernel_size = [16, 16]
                 size_factors = [4, 8, 16]
                 kernel_sizes = []
                 for size_factor in size_factors:

benchmarks/skimage/cucim_feature_bench.py

@@ -38,9 +38,9 @@ def set_args(self, dtype):
             locs = np.unravel_index(offsets, img.shape)
             for loc in zip(*locs):
                 loc = tuple(
-                    min(l, s - es) for l, s, es in zip(loc, img.shape, e.shape)
+                    min(p, s - es) for p, s, es in zip(loc, img.shape, e.shape)
                 )
-                sl = tuple(slice(l, l + es) for l, es in zip(loc, e.shape))
+                sl = tuple(slice(p, p + es) for p, es in zip(loc, e.shape))
                 img[sl] = e
         else:
             raise NotImplementedError("unsupported ndim")
@@ -82,7 +82,8 @@ def main(args):
             True,
         ),
         ("canny", dict(sigma=1.8), dict(), False, False),
-        # reduced default rings, histograms, orientations to fit daisy at (3840, 2160) into GPU memory
+        # reduced default rings, histograms, orientations to fit daisy at
+        # (3840, 2160) into GPU memory
         (
             "daisy",
             dict(step=4, radius=15, rings=2, histograms=5, orientations=4),
@@ -152,8 +153,6 @@ def main(args):
         else:
             continue
 
-        # if function_name in ["corner_peaks", "peak_local_max"] and np.prod(shape) > 1000000:
-        # skip any large sizes that take too long
         ndim = len(shape)
         run_cpu = not args.no_cpu
         if not allow_nd:

benchmarks/skimage/cucim_filters_bench.py

@@ -62,7 +62,8 @@ def main(args):
         # lpi_filter.py
         # TODO: benchmark wiener
         # ridges.py
-        # TODO: had to set meijering, etc allow_nd to False just due to insufficient GPU memory
+        # TODO: Had to set meijering, etc allow_nd to False just due to
+        #       insufficient GPU memory
         (
             "meijering",
             dict(sigmas=range(1, 10, 2), alpha=None),

benchmarks/skimage/cucim_measure_bench.py

@@ -239,7 +239,7 @@ def main(args):
 
             for contiguous_labels in [True, False]:
                 if contiguous_labels:
-                    index_str = f"contiguous"
+                    index_str = "contiguous"
                 else:
                     index_str = None
                 B = Tester(

benchmarks/skimage/cucim_morphology_bench.py

@@ -280,7 +280,7 @@ def main(args):
     all_results.to_csv(fbase + ".csv")
     all_results.to_pickle(pfile)
     try:
-        import tabular
+        import tabular  # noqa: F401
 
         with open(fbase + ".md", "wt") as f:
             f.write(all_results.to_markdown())

benchmarks/skimage/cucim_registration_bench.py

@@ -144,7 +144,7 @@ def main(args):
     all_results.to_csv(fbase + ".csv")
     all_results.to_pickle(pfile)
     try:
-        import tabular
+        import tabular  # noqa: F401
 
         with open(fbase + ".md", "wt") as f:
             f.write(all_results.to_markdown())

benchmarks/skimage/cucim_restoration_bench.py

@@ -179,7 +179,7 @@ def main(args):
     all_results.to_csv(fbase + ".csv")
     all_results.to_pickle(pfile)
     try:
-        import tabular
+        import tabular  # noqa: F401
 
         with open(fbase + ".md", "wt") as f:
             f.write(all_results.to_markdown())

benchmarks/skimage/cucim_segmentation_bench.py

@@ -11,7 +11,7 @@
 from _image_bench import ImageBench
 
 import cucim.skimage
-from cucim.skimage import data, exposure, measure, segmentation
+from cucim.skimage import data, measure
 
 
 class LabelBench(ImageBench):
@@ -262,6 +262,7 @@ def main(args):
                 var_kwargs=var_kwargs,
                 module_cpu=skimage.segmentation,
                 module_gpu=cucim.skimage.segmentation,
+                run_cpu=run_cpu,
             )
             results = B.run_benchmark(duration=args.duration)
             all_results = pd.concat([all_results, results["full"]])
@@ -285,6 +286,7 @@ def main(args):
                 var_kwargs=var_kwargs,
                 module_cpu=skimage.segmentation,
                 module_gpu=cucim.skimage.segmentation,
+                run_cpu=run_cpu,
             )
             results = B.run_benchmark(duration=args.duration)
             all_results = pd.concat([all_results, results["full"]])
@@ -293,7 +295,7 @@ def main(args):
     all_results.to_csv(fbase + ".csv")
     all_results.to_pickle(pfile)
     try:
-        import tabular
+        import tabular  # noqa: F401
 
         with open(fbase + ".md", "wt") as f:
             f.write(all_results.to_markdown())

benchmarks/skimage/cucim_transform_bench.py

@@ -154,6 +154,7 @@ def main(args):
             module_cpu=skimage.transform,
             module_gpu=cucim.skimage.transform,
             function_is_generator=function_is_generator,
+            run_cpu=run_cpu,
         )
         results = B.run_benchmark(duration=args.duration)
         all_results = pd.concat([all_results, results["full"]])
@@ -162,7 +163,7 @@ def main(args):
     all_results.to_csv(fbase + ".csv")
     all_results.to_pickle(pfile)
     try:
-        import tabular
+        import tabular  # noqa: F401
 
         with open(fbase + ".md", "wt") as f:
             f.write(all_results.to_markdown())

benchmarks/skimage/cupyx_scipy_ndimage_morphology_bench.py

@@ -1,4 +1,3 @@
-import math
 import os
 import pickle
 

examples/python/gds_whole_slide/benchmark_read.py

@@ -244,4 +244,4 @@
         duration (kvikio-pread (n_buffer=16) (gds_enabled=True)) = 0.9027577819824221
         duration (kvikio-pread (n_buffer=64) (gds_enabled=True)) = 0.7827104492187501
         duration (kvikio-pread (n_buffer=256) (gds_enabled=True)) = 0.756464599609375
-"""
+"""  # noqa: E501

examples/python/gds_whole_slide/benchmark_round_trip.py

@@ -158,7 +158,7 @@ def round_trip(
                     "tiled" if apply_kernel_tilewise else "global"
                 )
                 gds_description = "with GDS" if gds_enabled else "without GDS"
-                label = f"{computation=}, {kernel_description}, chunk_shape={zarr_chunk_shape}, {gds_description}"
+                label = f"{computation=}, {kernel_description}, chunk_shape={zarr_chunk_shape}, {gds_description}"  # noqa: E501
                 print(f"Duration ({label}): {t.mean()} s +/- {t.std()} s")
 
                 times.append(t.mean())
@@ -242,4 +242,4 @@ def round_trip(
     Duration (computation='sobel', tiled, chunk_shape=(4096, 4096, 3), without GDS): 5.430656982421875 s +/- 0.01277270507812478 s
     Duration (computation='sobel', tiled, chunk_shape=(4096, 4096, 3), with GDS): 3.1940713500976563 s +/- 0.19852391299904837 s
 
-"""
+"""  # noqa: E501

examples/python/gds_whole_slide/benchmark_zarr_write.py

@@ -98,7 +98,7 @@
                 write_time_means[i, j, k, m] = t.mean()
                 write_time_stds[i, j, k, m] = t.std()
                 print(
-                    f"Duration ({cp.dtype(dtype).name} write, {chunk_shape=}, {backend=}, {gds_enabled=}): "
+                    f"Duration ({cp.dtype(dtype).name} write, {chunk_shape=}, {backend=}, {gds_enabled=}): "  # noqa: E501
                     f"{t.mean()} s +/- {t.std()} s"
                 )
 out_name = "write_times.npz"

examples/python/gds_whole_slide/benchmark_zarr_write_lz4_via_dask.py

@@ -15,8 +15,6 @@
 from lz4_nvcomp import LZ4NVCOMP
 from tifffile import TiffFile
 
-from cucim.core.operations.color import image_to_absorbance
-
 data_dir = os.environ.get("WHOLE_SLIDE_DATA_DIR", os.path.dirname("__file__"))
 fname = os.path.join(data_dir, "resize.tiff")
 if not os.path.exists(fname):
@@ -55,7 +53,7 @@
 
 # benchmark writing these CuPy outputs to Zarr with various chunk sizes
 # Note: nvcomp only supports integer and unsigned dtypes.
-# https://github.com/rapidsai/kvikio/blob/b0c6cedf43d1bc240c3ef1b38ebb9d89574a08ee/python/kvikio/nvcomp.py#L12-L21  # noqa
+# https://github.com/rapidsai/kvikio/blob/b0c6cedf43d1bc240c3ef1b38ebb9d89574a08ee/python/kvikio/nvcomp.py#L12-L21  # noqa: E501
 
 dtypes = ["uint16"]
 chunk_shapes = [
@@ -107,7 +105,7 @@
                 write_time_means[i, j, k, m] = t.mean()
                 write_time_stds[i, j, k, m] = t.std()
                 print(
-                    f"Duration ({cp.dtype(dtype).name} write, {chunk_shape=}, {compressor=}, {gds_enabled=}): "
+                    f"Duration ({cp.dtype(dtype).name} write, {chunk_shape=}, {compressor=}, {gds_enabled=}): "  # noqa: E501
                     f"{t.mean()} s +/- {t.std()} s"
                 )
 
@@ -171,4 +169,4 @@
 Duration (uint16 write, chunk_shape=(4096, 4096, 3), compressor=LZ4NVCOMP, gds_enabled=False): 0.8166022583007813 s +/- 0.19712617258152443 s
 Duration (uint16 write, chunk_shape=(4096, 4096, 3), compressor=LZ4NVCOMP, gds_enabled=True): 0.7342889607747396 s +/- 0.025796103217999546 s
 
-"""
+"""  # noqa: E501