Python - meshio：如何为仅点数据定义顶点单元？

Question

问题

我有一个 2D 时间序列数据，我想使用 meshio 将其保存为 XMDF。 问题是，我的网格只是一个带有关联点数据的点数组，我没有定义任何单元格。 因此，我尝试使用"vertex"单元格类型，它是一个单点单元格，但它不起作用。 Meshio 的文档有点缺乏，所以我被卡住了。

代码

在他们的 Github 页面上的两个示例之后，我执行了以下操作。 我不确定如何正确定义单元格，因为 meshio 没有正确记录这一点。

# generate some data on a 10x10 mesh with 20 time steps (tested, works)
ts = np.arange(20)
x, y = np.meshgrid(np.arange(10), np.arange(10))
data = np.empty((20, 10, 10))
for i, t in enumerate(ts):
    data[i] = np.sin((x + y) * t)
# data is a 3D NumPy array now with dimensions (20,10,10)

# generate list of points (tested, works)
points = [list(p) for p in zip(*(x.flat, y.flat,))]

# won't use cells, so define vertex cell (1 point per cell) <-- ???
cells = [("vertex", [i,]) for i in range(len(points))]

# as seen in meshio's documentation, write time series data
filename = "test.xdmf"
with meshio.xdmf.TimeSeriesWriter(filename) as writer:
    writer.write_points_cells(points, cells)
    for i, t in enumerate(ts):
        writer.write_data(t, point_data={"sin_city": data[i]})

错误

上面的脚本产生以下错误：

Traceback (most recent call last):
  File "/home/ezio/Codes/gfield/_temp.py", line 103, in <module>
    writer.write_points_cells(points, cells)
  File "/home/ezio/anaconda3/envs/radpolpy/lib/python3.8/site-packages/meshio/xdmf/time_series.py", line 284, in write_points_cells
    self.points(grid, points)
  File "/home/ezio/anaconda3/envs/radpolpy/lib/python3.8/site-packages/meshio/xdmf/time_series.py", line 340, in points
    if points.shape[1] == 2:
AttributeError: 'list' object has no attribute 'shape'

我尝试了将一些 arrays 转换为 NumPy 阵列的不同组合，但我找不到原因。 我请求你的帮助。

更新：

在将每个使用的数字数组更改为 NumPy arrays （注释）之后 - 即在定义点之后直接插入points points = np.array(points) ，并将单元生成器行更改为cells = [("vertex", np.array([i,])) for i in range(len(points))] - 我仍然有不同的错误：

Traceback (most recent call last):
  File "/home/ezio/Codes/gfield/_temp.py", line 105, in <module>
    writer.write_points_cells(points, cells)
  File "/home/ezio/anaconda3/envs/radpolpy/lib/python3.8/site-packages/meshio/xdmf/time_series.py", line 285, in write_points_cells
    self.cells(cells, grid)
  File "/home/ezio/anaconda3/envs/radpolpy/lib/python3.8/site-packages/meshio/xdmf/time_series.py", line 409, in cells
    [
  File "/home/ezio/anaconda3/envs/radpolpy/lib/python3.8/site-packages/meshio/xdmf/time_series.py", line 411, in <listcomp>
    np.insert(
  File "<__array_function__ internals>", line 5, in insert
  File "/home/ezio/anaconda3/envs/radpolpy/lib/python3.8/site-packages/numpy/lib/function_base.py", line 4527, in insert
    axis = normalize_axis_index(axis, ndim)
numpy.AxisError: axis 1 is out of bounds for array of dimension 1

（我还注意到文档在示例中没有使用 NumPy arrays。）

Answer 1

问题是：

1. 我应该在任何地方都使用 NumPy arrays ：即使meshio的示例使用 Python 列表，xmdf 模块显然无法处理这些； 和
2. 我也应该展平数据（显然）。 此外，应该以更好的方式定义单元格，尽管原始概念也有效。

我的代码的工作版本是：

# generate some data on a 10x10 mesh with 20 time steps (tested, works)
ts = np.arange(20)
x, y = np.meshgrid(np.arange(10), np.arange(10))
data = np.empty((20, 10, 10))
for i, t in enumerate(ts):
    data[i] = np.sin((x + y) * t)
# data is a 3D NumPy array now with dimensions (20,10,10)

# generate list of points (tested, works)
points = [list(p) for p in zip(*(x.flat, y.flat,))]
points = np.array(points)  # add this

# won't use cells, so define vertex cell (1 point per cell)
cells = [("vertex", np.array([[i,] for i in range(len(points)])))]
# instead of cells = [("vertex", [i,]) for i in range(len(points))]

# as seen in meshio's documentation, write time series data
filename = "test.xdmf"
with meshio.xdmf.TimeSeriesWriter(filename) as writer:
    writer.write_points_cells(points, cells)
    for i, t in enumerate(ts):
        # here data[i] also should be flattened
        writer.write_data(t, point_data={"sin_city": data[i].flatten})