Moved many tests from notebook to test_resampler

resampler/__init__.py

@@ -15,7 +15,7 @@
 import functools
 import itertools
 import math
-import multiprocessing
+import os
 import sys
 import types
 import typing
@@ -237,7 +237,17 @@ def func(array: _NDArray) -> _NDArray:
         print(f'Creating numba jit-wrapper for {signature}.')
       jitted_function = numba.njit(func, parallel=True, fastmath=True, cache=True)
       self._jitted_function[signature] = jitted_function
-    result = jitted_function(a)
+
+    try:
+      result = jitted_function(a)
+    except RuntimeError:
+      message = (
+          'resampler: This runtime error may be due to a corrupt resampler/__pycache__;'
+          ' try deleting that directory.'
+      )
+      print(message, file=sys.stderr, flush=True)
+      raise
+
     return result[..., 0] if array.ndim == 2 else result
 
 
@@ -439,7 +449,7 @@ def premult_with_sparse(
     assert self.array.ndim == sparse.ndim == 2 and sparse.shape[1] == self.array.shape[0]
     # Empirically faster than with default numba.config.NUMBA_NUM_THREADS (e.g., 24).
     if using_numba:
-      num_threads2 = min(6, multiprocessing.cpu_count()) if num_threads == 'auto' else num_threads
+      num_threads2 = min(6, os.cpu_count()) if num_threads == 'auto' else num_threads
       src = np.ascontiguousarray(self.array)  # Like .ravel() in _mul_multivector().
       dtype = np.result_type(sparse.dtype, src.dtype)
       dst = np.empty((sparse.shape[0], src.shape[1]), dtype)
@@ -3794,14 +3804,14 @@ def _find_closest_filter(filter: str, resizer: Callable[..., Any]) -> str:
           skimage_transform_resize: 'box',
           tf_image_resize: 'trapezoid',
           torch_nn_resize: 'trapezoid',
-      }.get(resize, 'box')
+      }.get(resizer, 'box')
     case 'cubic_like':
       return {
           cv_resize: 'sharpcubic',
           scipy_ndimage_resize: 'cardinal3',
           skimage_transform_resize: 'cardinal3',
           torch_nn_resize: 'sharpcubic',
-      }.get(resize, 'cubic')
+      }.get(resizer, 'cubic')
     case 'high_quality':
       return {
           pil_image_resize: 'lanczos3',