Make EstimateFlow's config class dataclasses_json compatible.

connectomics/common/array.py

@@ -25,9 +25,8 @@
 
 from collections import abc
 import numbers
-from typing import Any, List, Tuple, TypeVar, Type, Union
+from typing import Any, List, Tuple, TypeVar, Union
 
-from connectomics.common import array_mixins
 import numpy as np
 import numpy.typing as npt
 
@@ -43,12 +42,11 @@
 IndexExpOrPointLookups = Union[ArbitrarySlice, PointLookups]
 CanonicalSliceOrPointLookups = Union[CanonicalSlice, PointLookups]
 
-ArrayLike = Union[npt.ArrayLike, 'ImmutableArray', 'MutableArray']
+ArrayLike = npt.ArrayLike
 Tuple3f = Tuple[float, float, float]
 Tuple3i = Tuple[int, int, int]
 Tuple4i = Tuple[int, int, int, int]
-ArrayLike3d = Union[npt.ArrayLike, 'ImmutableArray', 'MutableArray', Tuple3f,
-                    Tuple3i]
+ArrayLike3d = Union[npt.ArrayLike, Tuple3f, Tuple3i]
 
 
 def is_point_lookup(ind: IndexExpOrPointLookups) -> bool:
@@ -120,75 +118,22 @@ def process_slice_ind(slice_ind: ArbitrarySlice, limit: int) -> slice:
   return slice(*slice_ind.indices(limit))
 
 
-# TODO(timblakely): Make these typed by using Generic[T]
-class ImmutableArray(array_mixins.ImmutableArrayMixin, np.ndarray):
-  """Strongly typed, immutable NumPy NDArray."""
-
-  def __new__(cls: Type['ImmutableArray'],
-              input_array: 'ArrayLike',
-              *args,
-              zero_copy=False,
-              **kwargs) -> 'ImmutableArray':
-    if zero_copy:
-      obj = np.asanyarray(input_array, *args, **kwargs).view(cls)
-    else:
-      obj = np.array(input_array, *args, **kwargs).view(cls)
-    obj.flags.writeable = False
-    return obj
-
-  def __init__(self, *args, zero_copy=False, **kwargs):
-    # Needed for mixin construction.
-    super().__init__()  # pylint: disable=no-value-for-parameter
-
-  def __array_ufunc__(self, ufunc, method, *inputs, **kwargs):
-    out = kwargs.get('out', ())
-
-    # Defer to the implementation of the ufunc on unwrapped values.
-    inputs = tuple(_to_np_compatible(x) for x in inputs)
-    if out:
-      kwargs['out'] = tuple(_to_np_compatible(x) for x in out)
-    result = getattr(ufunc, method)(*inputs, **kwargs)
-
-    if method == 'at':
-      # no return value
-      return None
-
-    return MutableArray(result)
-
-  def copy(self, *args, **kwargs) -> 'MutableArray':
-    return MutableArray(np.asarray(self).copy(*args, **kwargs))
-
-  def __str__(self):
-    return np.ndarray.__repr__(self)
-
-
-class MutableArray(array_mixins.MutableArrayMixin, ImmutableArray):
-  """Strongly typed mutable version of np.ndarray."""
-
-  def __new__(cls: Type['MutableArray'],
-              input_array: 'ArrayLike',
-              *args,
-              zero_copy=False,
-              **kwargs) -> 'MutableArray':
-    if zero_copy:
-      obj = np.asanyarray(input_array, *args, **kwargs).view(cls)
-    else:
-      obj = np.array(input_array, *args, **kwargs).view(cls)
-    obj.flags.writeable = True
-    return obj
+# Kludge to get around the fact that Jax uses its own ArrayImpl class, which
+# doesn't fulfill either the abc.Sequence or np.ndarray contracts.
+def _is_indexable(obj: Any) -> bool:
+  return hasattr(obj, '__getitem__') and hasattr(obj, '__setitem__')
 
 
 def is_arraylike(obj):
   # Technically sequences, but this is intended to check for numeric sequences.
   if isinstance(obj, str) or isinstance(obj, bytes):
     return False
-  return isinstance(obj, abc.Sequence) or isinstance(
-      obj, np.ndarray) or isinstance(obj, ImmutableArray) or isinstance(
-          obj, MutableArray)
+  return (
+      _is_indexable(obj)
+      or isinstance(obj, abc.Sequence)
+      or isinstance(obj, np.ndarray)
+  )
 
 
 def _to_np_compatible(array_like) -> np.ndarray:
-  if isinstance(array_like, ImmutableArray) or isinstance(
-      array_like, MutableArray):
-    return np.asarray(array_like)
-  return array_like
+  return np.asarray(array_like)