使用PyVista可视化3D数据时遇UniformGrid属性错误及值相关问题求助

问题描述 #

运行以下PyVista可视化代码时出现AttributeError: module 'pyvista' has no attribute 'UniformGrid'错误（使用最新版PyVista）：

from openpmd_viewer import OpenPMDTimeSeries
import pyvista

# Open the simulation outputs using openPMD viewer
ts = OpenPMDTimeSeries('./sim_outputs/diags/hdf5')

# Create the PyVista plotter
plotter = pyvista.Plotter()
plotter.set_background("white")

# Retrieve the rho field from the simulation
# The theta=None argument constructs a 3D cartesian grid from the cylindrical data
rho, meta = ts.get_field("rho", iteration=ts.iterations[-1], theta=None)

# Create the grid on which PyVista can deposit the data
grid = pyvista.UniformGrid()
grid.dimensions = rho.shape
grid.origin = [meta.xmin * 1e6, meta.ymin * 1e6, meta.zmin * 1e6]
grid.spacing = [meta.dx * 1e6, meta.dy * 1e6, meta.dz * 1e6]
grid.point_data['values'] = -rho.flatten(order='F')

# Add the grid to the plotter
# Use a cutoff for rho via the clim argument since otherwise it shows only a small density spike
plotter.add_volume(grid, clim=(0, 4e6), opacity='sigmoid',
              cmap='viridis', mapper='gpu', show_scalar_bar=False)

# A good starting camera position - the three values are the camera position,
# the camera focus, and the up vector of the viewport
plotter.camera_position = [(-74, 32, 51), (0, 0, 88), (0, 1, 0)]

plotter.show()

将UniformGrid替换为StructuredGrid后，又出现与values相关的维度不匹配错误。

错误原因 #

1. UniformGrid报错：最新版PyVista中UniformGrid类是存在的，报错大概率是导入路径冲突（如环境中存在同名pyvista模块）、PyVista版本实际未更新到位，或者导入方式问题。
2. StructuredGrid的values错误：StructuredGrid与UniformGrid的初始化逻辑不同，前者需要显式生成网格点坐标，不能直接通过origin和spacing设置，若跳过这一步会导致点数据的维度不匹配。

解决方案 #

方案1：修复UniformGrid的使用 #

先确认PyVista版本，再调整导入方式：

from openpmd_viewer import OpenPMDTimeSeries
# 直接导入UniformGrid，避免模块属性查找问题
from pyvista import Plotter, UniformGrid

ts = OpenPMDTimeSeries('./sim_outputs/diags/hdf5')

plotter = Plotter()
plotter.set_background("white")

rho, meta = ts.get_field("rho", iteration=ts.iterations[-1], theta=None)

# 正确初始化UniformGrid
grid = UniformGrid()
grid.dimensions = rho.shape
grid.origin = [meta.xmin * 1e6, meta.ymin * 1e6, meta.zmin * 1e6]
grid.spacing = [meta.dx * 1e6, meta.dy * 1e6, meta.dz * 1e6]
grid.point_data['values'] = -rho.flatten(order='F')

plotter.add_volume(grid, clim=(0, 4e6), opacity='sigmoid',
              cmap='viridis', mapper='gpu', show_scalar_bar=False)
plotter.camera_position = [(-74, 32, 51), (0, 0, 88), (0, 1, 0)]

plotter.show()

若仍报错，执行pip install --upgrade pyvista确保版本为最新，同时检查环境中是否有其他同名模块干扰。

方案2：正确使用StructuredGrid #

若需使用StructuredGrid，需先生成网格点坐标：

from openpmd_viewer import OpenPMDTimeSeries
import pyvista
import numpy as np

ts = OpenPMDTimeSeries('./sim_outputs/diags/hdf5')

plotter = pyvista.Plotter()
plotter.set_background("white")

rho, meta = ts.get_field("rho", iteration=ts.iterations[-1], theta=None)

# 生成坐标轴数组
x = np.linspace(meta.xmin * 1e6, meta.xmax * 1e6, rho.shape[0])
y = np.linspace(meta.ymin * 1e6, meta.ymax * 1e6, rho.shape[1])
z = np.linspace(meta.zmin * 1e6, meta.zmax * 1e6, rho.shape[2])

# 创建结构化网格的点坐标（保持与rho的维度顺序一致）
xx, yy, zz = np.meshgrid(x, y, z, indexing='ij')
points = np.column_stack((xx.ravel(order='F'), yy.ravel(order='F'), zz.ravel(order='F')))

# 初始化StructuredGrid并设置属性
grid = pyvista.StructuredGrid()
grid.points = points
grid.dimensions = rho.shape
grid.point_data['values'] = -rho.flatten(order='F')

plotter.add_volume(grid, clim=(0, 4e6), opacity='sigmoid',
              cmap='viridis', mapper='gpu', show_scalar_bar=False)
plotter.camera_position = [(-74, 32, 51), (0, 0, 88), (0, 1, 0)]

plotter.show()

这里通过numpy.meshgrid生成与rho维度匹配的点坐标，确保point_data['values']的长度与点数量一致，避免维度错误。

内容的提问来源于stack exchange，提问作者Dinkar Mishra