TypeError：“numpy.uint32”对象在使用 numpy.full 函数时不可迭代

Question

在一个巨大的 python 脚本中的某些行中，有一个由 74537000 个元素组成的 3d 数组，这些元素都是浮点数：

import numpy as np
prop_shape = [74537000, 3]  # to produce this however a long script is written and this will be printed (just before the controversial line below) in the first run of the code assuming it's a vector quantity like position.
prop_in_dtype = np.float32  # position array with the shape (74537000, 3) is made up of float32 values. In general, however, prop_in_dtype is changing across the script due to the nature of the array

当我们试图创建一个元素都是 -1 的相同数组时，我们可以使用 numpy 如下：

.
.
.
my_array = np.full(tuple(prop_shape), -1, prop_in_dtype)
.
.
.

运行此程序时，我收到以下错误消息：

Traceback (most recent call last):
  File "DLA_DM.py", line 19, in <module>
    settings_centroid.init()
  File "/usr5/username/settings_centroid.py", line 43, in init
    part=gizmo.io.Read.read_snapshots(species, snapshot_value_kind, snapshot_number, simulation_directory='.', snapshot_directory='output/', simulation_name='', properties=properties, element_indices=None, particle_subsample_factor=0, separate_dark_lowres=True, sort_dark_by_id=False, convert_float32=True, host_number=1, assign_host_coordinates=True, assign_host_principal_axes=False, assign_host_orbits=False, assign_formation_coordinates=False, assign_pointers=False, check_properties=True)
  File "/usr5/username/simulation/gizmo/gizmo_io.py", line 649, in read_snapshots
    element_indices, convert_float32, header)
  File "/usr5/username/simulation/gizmo/gizmo_io.py", line 1164, in read_particles
    part[spec_name][prop] = np.full(tuple(prop_shape), - 1, prop_in_dtype) 
TypeError: 'numpy.uint32' object is not iterable

Answer 1

prop_shape 在原始代码中进行迭代。有时它是一个整数（对于标量，例如 74537000），有时是一个二元素列表（对于向量，例如 [74537000, 3]）。我假设它被固定为一个二元素列表（因此是一个向量）。这就是为什么如果标量不是float32 dtype 的单个标量的元组会给出这样的错误。为了避免这种情况，我创建了一个if statement，每次在创建数组之前检查变量的dtype：

if isinstance(prop_shape, list):
    my_array = np.full(tuple(prop_shape), -1, prop_in_dtype)
else:
    my_array = np.full(prop_shape, -1, prop_in_dtype)

但是，正如我在对 Warren 建议的评论中所说，我最终得到如下 MemoryError：

Traceback (most recent call last):
  File "DLA_DM.py", line 96, in <module>
    gizmo.plot.Image.plot_image(settings_centroid.part, 'dark', 'mass', settings_centroid.image_kind, [0,1], settings_centroid.dimensions_select, settings_centroid.distance_max, settings_centroid.distance_bin_width_generic, distance_bin_number=settings_centroid.distance_bin_number, part_indices=part_indices, write_plot=write_plot, plot_directory=plot_directory, background_color=background_color, use_column_units=False, center_position=settings_centroid.center_position, rotation=settings_centroid.rotation, property_select={'mass.bound':[1e9,1e13]}, subsample_factor=subsample_factor, image_limits=settings_centroid.image_limits_xy, hal=settings_centroid.hal, hal_indices=settings_centroid.hal_indices, hal_position_kind=hal_position_kind, hal_radius_kind=hal_radius_kind, return_halo_info=settings_centroid.return_halo_info_value_dark)
  File "/usr5/username/simulation/gizmo/gizmo_plot.py", line 825, in plot_image
    settings_centroid.init()
  File "/usr5/username/settings_centroid.py", line 43, in init
    part=gizmo.io.Read.read_snapshots(species, snapshot_value_kind, snapshot_number, simulation_directory='.', snapshot_directory='output/', simulation_name='', properties=properties, element_indices=None, particle_subsample_factor=0, separate_dark_lowres=True, sort_dark_by_id=False, convert_float32=True, host_number=1, assign_host_coordinates=True, assign_host_principal_axes=False, assign_host_orbits=False, assign_formation_coordinates=False, assign_pointers=False, check_properties=True)
  File "/usr5/username/simulation/gizmo/gizmo_io.py", line 649, in read_snapshots
    element_indices, convert_float32, header)
  File "/usr5/username/simulation/gizmo/gizmo_io.py", line 1171, in read_particles
    part[spec_name][prop] = np.full(prop_shape, -1, prop_in_dtype)
  File "/usr/local/anaconda3/lib/python3.5/site-packages/numpy/core/numeric.py", line 309, in full
    a = empty(shape, dtype, order)
MemoryError

Answer 2

我可以通过以下方式重现错误消息：

In [174]: tuple(np.uint32(1))                                                        
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-174-3420675a1788> in <module>
----> 1 tuple(np.uint32(1))

TypeError: 'numpy.uint32' object is not iterable

所以我猜prop_shape 变量有问题。